Looking for a few good reductionists/strong AI
Jacques Maurice Mallah
Hello. As the title says, are there some knowledgeable people
here who believe in computational reductive functionalism with regard to
consciousness? It apppears that a serious problem may have been
discovered with the approach, and I want to discuss it with such people.
- - - - - - -
Jacques Mallah (mal...@acf2.nyu.edu)
Graduate Student / Many Worlder / Devil's Advocate
"I know what no one else knows" - 'Runaway Train', Soul Asylum
http://pages.nyu.edu/~jqm1584/
Jacques Maurice Mallah
Jacques Maurice Mallah (mal...@acf2.nyu.edu) wrote:
: Hello. As the title says, are there some knowledgeable people
Jacques Maurice Mallah
I have recieved a couple of emails, and have decided to start the
discussion here on c.a.p.
I find it plausible that the so called "hard problem" of
consciousness can be solved by explaining qualia as an illusion
encountered in the course of a computation.
Ironically, there seems to be a problem with another, crucial but
sadly neglected, aspect of reductive computational functionalism with
regard to consciousness (RCFC).
In order for RCFC to make sense, there *must* be an *objective*
criterion for under what conditions a given physical system implements or
does not implement a given computation. It appears that a suitable
criterion may not be possible, throwing the notion of computation in doubt.
The starting point for this discussion is a paper by David
Chalmers entitled "Does a rock implement every computation?" This paper
is available on the web at http://www1.artsci.wustl.edu/~chalmers
under Foundations of Artificial Intelligence. Chalmers is not a
reductionist, but he is a computational functionalist, which is why he
has worked on relevent issues.
In this paper, Chalmers first shows that a certain class of
"false implementations" of computations by a rock, proposed by Putnam,
can be ruled out with a suitable definition of computation.
He then goes on to study another class of "false implementations"
which is much harder to rule out.
This involves mapping the states of a clock and a dial to the
formal states of a computation.
Recall that the usual definition of a computation talks of states
which have a certain causal relationship between them. How to relate
such formal states to physical systems? Computations are implemented by
circuits, by arrangements of levers, or just about anything. It is clear
that formal states can by assigned by using functionals of the physical
system to map system configurations to formal states. Such functionals
should be restricted so that mutually exclusive formal states are
mutually exclusive, but otherwise have few limitations.
So it turns out that the readings of the clock and the dial can
be mapped so that any arbritrary computation looks like it is
implemented. The clock assures the causal relationships are right (if A
then B happens next, etc.) while the dial supplies counterfactuals (if
the state had been D, then E would have happened next).
Chalmers "solves" the problem by introducing the Combinatorial
State Automaton (CSA). Unlike its predecessor, the Finite State
Automaton (FSA), the CSA has states that have structure, such as a vector
of bits. In order to prevent the distinction from being merely a matter of
nomenclature, which would leave the problem unsolved, Chalmers proposes
that the individual parameters (e.g. bits) must be "independent in some
strong sense."
He suggests that a suitable sense is that the functionals depend
on spacially seperated regions. He believes that this is only one
example of a more general way of defining "independence", but does not go
any further. There remains one class of "false implementations" even
with "independence", but he is not too worried about it because it
requires a computer exponentially large in the size of the computation,
so the universe will contain no such false implementations of a
computation complex enough to be concious.
I looked at the idea of "independence". After some false starts,
it became clear to me that the idea seems impossible to generalize in a
suitable way. For obvious reasons his present definition is unsuitable:
it invokes a **preferred** quantity, namely position, that has no
mathematical properties, other than being *a priori* identified as
preferred, that can not be simulated by appropriately mapping the states
of the clock and the dial;
it is far from clear how to generalize it to the case of quantum
mechanics, where all atoms spacially overlap, and one of the famous
features of quantum mechanics is the ability to work in any basis, such
as position or momentum; also as I am a many-worlder, I seek a resolution
to the problem that the MWI seems to require a preferred basis;
finally, it is possible to concieve of computers where the bits
clearly do overlap spacially. For example, bits could be stored as
amplitudes of sound waves at different frequencies in the same object.
Concievably, one might not need the idea of "independence" if one
could replace it with some other restriction on the functionals, for
example, that they be sufficiently "simple" in contrast to the
gerrymander of the clock and dial. However, that also requires a
preferred basis to take as the starting point, and it is far from clear
what "simple" would mean.
The problem for RCFC, then, is that merely defining
implementation of computation seems to require a quantity that is a prioi
specified as preferred, such as position.
To put it in the clearest way, there seemingly must exist a new,
"compu-physical law", that specifies the preferred quantity. Clearly, we
are no longer talking about a *reductionist* theory. Since the precise
defintion of implementation was only sought as a way to explain
conciousness, it is clear that such a law is a variety of
"psycho-physical law" of the sort usually associated with dualists.
Even then, with implementation defined in terms of independence,
there would still be a class of "false implementations" possible in
principle if not in practice. Added to the previous problem, the whole
idea of computation as a meaningful, objective thing, comes into question.
Can you solve this problem? Good luck, and I mean it!
Neil Rickert
> Ironically, there seems to be a problem with another, crucial but
>sadly neglected, aspect of reductive computational functionalism with
>regard to consciousness (RCFC).
> In order for RCFC to make sense, there *must* be an *objective*
>criterion for under what conditions a given physical system implements or
>does not implement a given computation. It appears that a suitable
>criterion may not be possible, throwing the notion of computation in doubt.
This is a problem that Searle raises, at least implicitly, in his
Chinese Room argument. He raises is explicitly in his "Rediscovery
of Mind." And Putnam raises the problem in the appendix to his
"Representation and Reality."
I think it is not so clear that objective criteria are required. In
some sense, we don't have objective criteria for anything. Perhaps
we can reduce matter to molecules, molecules to atoms, and atoms to
quarks. But do we have objective criteria for quarks? Do we even
have objective criteria for objective criteria?
My preference is to think of an algorithm as an explanatory theory
for the operation of a computer. There are alternative explanatory
theories such as those which explain in terms of electron motions.
The computational theory does not account for everything physical
that the computer does. For example, it does not account for the
noise produced by the fan designed to cool the processor chip. Yet
we still accept the algorithmic description as a reduction of the
action of the computer, even though it is not uniquely determined and
does not explain everything.
The general view is that a reduction of cognition would involve two
steps:
step1: reduce cognition to computation.
step2: reduce computation to physics.
It seems to me that the question of objective criteria for
computation is really a question which mainly concerns step2. Yet,
ironically, it is mainly brought up to challenge the possibility of
step1, and most people think step2 is a solved problem.
> Chalmers "solves" the problem by introducing the Combinatorial
>State Automaton (CSA). Unlike its predecessor, the Finite State
>Automaton (FSA), the CSA has states that have structure, such as a vector
>of bits. In order to prevent the distinction from being merely a matter of
>nomenclature, which would leave the problem unsolved, Chalmers proposes
>that the individual parameters (e.g. bits) must be "independent in some
>strong sense."
> He suggests that a suitable sense is that the functionals depend
>on spacially seperated regions.
Presumably this approach requires making assumptions about space as
initial a priori committments. But if one does that, one makes it
difficult to study to what extent neural computation leads us to
discover the nature of space.
Jacques Maurice Mallah
Neil Rickert (ric...@cs.niu.edu) wrote:
: > Ironically, there seems to be a problem with another, crucial but
: >regard to consciousness (RCFC).
: > In order for RCFC to make sense, there *must* be an *objective*
: >criterion for under what conditions a given physical system implements or
: >does not implement a given computation. It appears that a suitable
: >criterion may not be possible, throwing the notion of computation in doubt.
: I think it is not so clear that objective criteria are required. In
: some sense, we don't have objective criteria for anything. Perhaps
: we can reduce matter to molecules, molecules to atoms, and atoms to
: quarks. But do we have objective criteria for quarks? Do we even
: have objective criteria for objective criteria?
:
: My preference is to think of an algorithm as an explanatory theory
: for the operation of a computer. There are alternative explanatory
: theories such as those which explain in terms of electron motions.
: The computational theory does not account for everything physical
: that the computer does. For example, it does not account for the
: noise produced by the fan designed to cool the processor chip. Yet
: we still accept the algorithmic description as a reduction of the
: action of the computer, even though it is not uniquely determined and
: does not explain everything.
I don't see your point. *Of course* objective criteria are
required if computation is to be the foundation of consciousness. If
not, there is no sense in which "the world contains human beings thinking
about computers" is a different kind of statement than "the world
contains pop up toasters thinking about computers".
Not only must there be an objective criterion, but also,
perceptions sort of like ours must be fairly typical in some sense among
the thoughts in the world. If these two conditions are not met,
computational reductive functionalism **is disproved**. Period. CRFC is
a philosophically / mathematically disprovable kind of belief, and that
is how it should be.
Why don't we need objective criteria for electrons, without
invoking a preferred basis? Two reasons. First, electrons are physical
objects, not mathematical ones the way computations are. Second, and
more importantly, whether electrons exist or whether there is some other
reality that just gives the appearance of electrons, plays no fundamental
role in what we percieve. It just plain doesn't matter.
Consciousness is the opposite of that. It is the only thing of
which what we percieve consists. *All* of what we know, relates to it
and exclusively to it, by definition. It is what we must explain to
understand the world we percieve. It and nothing else.
Now, we assume that a physical world exists, and has a certain
mathematical description, because such as assumption seems to allow us to
make sense of patterns in what we percieve. CRFC is the belief,
basically, that that physical world is sufficient to account not only for
what we percieve, but for the fact that we percieve it, in terms of
computation. Clearly an objective criterion is required.
If computation is undefined except as something some conscious
person can use to organize data, it clearly can play no explanatory
role. Nothing in our model of the world would give rise to consciousness
in the first place.
: > Chalmers "solves" the problem by introducing the Combinatorial
: >State Automaton (CSA). Unlike its predecessor, the Finite State
: >Automaton (FSA), the CSA has states that have structure, such as a vector
: >of bits. In order to prevent the distinction from being merely a
matter of
: >nomenclature, which would leave the problem unsolved, Chalmers proposes
: >that the individual parameters (e.g. bits) must be "independent in some
: >strong sense."
: > He suggests that a suitable sense is that the functionals depend
: >on spacially seperated regions.
:
: Presumably this approach requires making assumptions about space as
: initial a priori committments. But if one does that, one makes it
: difficult to study to what extent neural computation leads us to
: discover the nature of space.
I'm not following.
It does seem like "compu-physical laws" may be inevitable. That
would not be reductive, but presumably if we see it's needed, we're
learning something.
H. M. Hubey
ric...@cs.niu.edu (Neil Rickert) writes:
> step1: reduce cognition to computation.
> step2: reduce computation to physics.
>It seems to me that the question of objective criteria for
>computation is really a question which mainly concerns step2. Yet,
>ironically, it is mainly brought up to challenge the possibility of
>step1, and most people think step2 is a solved problem.
Sometimes I get the feeling that some people (not you) think
of things like consciousness as something like a mode that
a computer is in, e.g. supervisor mode, user mode, or I/O mode,
or memory_fetch mode.
Maybe the simulation programs should explicitly have such
flags which can be analogues of human things like; awake, asleep,
listening, thinking, absent-minded (lost in thought), watches_but_does_not_see,
listens_but_cannot_hear, memory_fetch, nostalgic_mode, anger_mode,
etc.
At least I get the feeling that they are looking for some
switches, neural_correlates, cell_assemblies firing at
certain rates, a certain number of times, etc so that we can\
then make the brain behave like a computer, or think of it
that way because we have become so fond of the digital computer
explanation.
--
Regards, Mark
http://www.csam.montclair.edu/~hubey hu...@pegasus.montclair.edu
Gary Forbis
Jacques Maurice Mallah wrote:
> I find it plausible that the so called "hard problem" of
> consciousness can be solved by explaining qualia as an illusion
> encountered in the course of a computation.
I know much of what I write grates on others nerves so this may be
a bit inappropriate. None the less, "qualia as an illusion" grates
on my nerves. I can make no sense of it. What in the heck is an
"illusion" without some sense of "qualia?"
I have been approaching AI as a computationalist understanding of the
mapping between human input and human output. Such a map will let us
build many useful machines but it need not explain "qualia" or
"consciousness."
There is a dualist nature tied to the word "implement" in that an
implementation must be true to the nature of the medium of the
implmentation and the function implmented. Much of this relies upon
the level of the description and the reliability of the behavior,
for instance a light switch has a reliable map between functional
state and switch position, input and pressure, and state change and
position change.
I think the major problem with using computationalist explanations
of behavior to talk about qualia is that the reliable mapping to
behavior needn't necessarily involve qualia and there is yet to be
found an objective link between qualia and behavior or even conceptual
framework upon which to establish such an objective link.
-- gary for...@accessone.com
Jacques Maurice Mallah
: is available on the web at http://www1.artsci.wustl.edu/~chalmers
: under Foundations of Artificial Intelligence.
The url is now http://www.artsci.wustl.edu/~chalmers
Oliver Sparrow
(Jacques Maurice Mallah) writes, of NR:
> I don't see your point. *Of course* objective criteria are
> required if computation is to be the foundation of consciousness.
Why? You talk of computations as "objects" when they are, in fact,
processes. NR makes the point that there are an infinite number
of answers to the question 'why?' - why, say, is that flower pink .
The answer could be pitched in terms of human aesthetics, recent
developments in floral retailing, spectral reflectance, plant breeding,
Mrs Jone's personal preferences, Japanese cultural influence.... One
could also audit photons, spectra, retinal pigments and grope for the
origins of the pink qualia in primary data, in neurophysiology or what have
you for a more integrative approach. There is no Answer, just useful ways
of thinking, which intersect and offer a rounded view of what is going on.
NR's view of the 'best' description of what a computer is doing is
expressed in the algorithm which it is running. The switches are being
driven by a set of instructions which are designed to give rise to a process.
CAs and related systems are interesting in that the role of design is set
one stage back - to designing the CA - and any regularities which emerge
were not put in there by a master mind. That is to say, the algorithm is a
good description because (a) it is, at the moment, in the driving seat and
(b) as we make such things, if we understand this instance, then we can
understand the overall intent which gave it existence. That is what fulfills
one meaning of the word 'explanation'. In other words, we are back to
multi-level explanations, whereby the program is adequately mapped into a
sea of otherthings that we know, such that the fit is comfortable. We know,
more or less, how computers and made and retailed, broadly how software
gets written, how MPUs go about their business, so out focus lies upon
issues which this instance of computing may raise, perhaps by failing to
conform to our expetations. We then find and 'explanation' as to why this is
so, altering out views at some level until we feel that we once again have
understanding.
This elastic, connected and non-hierarchical way of organising percept
and cognition denies 'objective criteria', save in relationship to other
more or less fixed points that are representative of constellations of concepts
that are well-embedded, closely intercoupled and able to rub along together.
Out there in the phenomenal world there may be fixed stars - quarks and gluons,
although don't count on it - but this is a poor model for how cognitive processes
seem to work. The decoupling of emergent systems from their substrate
components implies tha some absolute systems, beginning in number theory or
particle physics, which builds out to encompass all of cognition, life and physical
reality is a C19th dream which holds us back.
_________________________________________________
Oliver Sparrow
oh...@chatham.demon.co.uk
Jacques Maurice Mallah
Gary Forbis (for...@accessone.com) wrote:
: What in the heck is an "illusion" without some sense of "qualia?"
I don't really want to discuss this; as you may have noticed,
computational reductive functionalism has a big problem which is
unrelated, and that is what I want to discuss. However I will make a
couple of comments.
The thing people usually have trouble understanding about
consciousness is that it seems to have a "qualitative" aspect. It also
has quantitative aspects, such as the shapes of things we see, and it is
often considered plausible that computation could be the explanation for
the quantitative aspects. Colors, however, can in no way be described
mathematically; they are qualia. So, I assume they have no real
existance, but that our brains have a tendency to compute statements like
"I see colors" nonetheless.
You may find yourself unable to believe in computational
reductionism. However, even if you are a dualist, you can still be a
compuational functionalist, as Chalmers is. That is, he believes that
there is a law of nature that attaches real qualia to appropriate
computations. Therefore he too is concerned with the issue of
implementation of computations.
Neil Rickert
In <4uqodp$v...@news.nyu.edu> mal...@acf2.nyu.edu (Jacques Maurice Mallah) writes:
>Neil Rickert (ric...@cs.niu.edu) wrote:
>: I think it is not so clear that objective criteria are required. In
>: some sense, we don't have objective criteria for anything. Perhaps
>: we can reduce matter to molecules, molecules to atoms, and atoms to
>: quarks. But do we have objective criteria for quarks? Do we even
>: have objective criteria for objective criteria?
>: My preference is to think of an algorithm as an explanatory theory
>: for the operation of a computer. There are alternative explanatory
>: theories such as those which explain in terms of electron motions.
>: The computational theory does not account for everything physical
>: that the computer does. For example, it does not account for the
>: noise produced by the fan designed to cool the processor chip. Yet
>: we still accept the algorithmic description as a reduction of the
>: action of the computer, even though it is not uniquely determined and
>: does not explain everything.
> I don't see your point. *Of course* objective criteria are
>required if computation is to be the foundation of consciousness. If
>not, there is no sense in which "the world contains human beings thinking
>about computers" is a different kind of statement than "the world
>contains pop up toasters thinking about computers".
Imagine an ant. Perhaps it is a super intelligent ant. Is there any
sense to this ant in saying that "the world contains pop up toasters
thinking about computers" is a different kind of statement that "the
world contains humans thinking about computers?" Whether there are
ants thinking about computers might have a different kind of
significance, however.
In his argument against reductionism, John Searle claims that
intentionality is an intrinsic property of reality, but computation
is only an interpretation (in his "Rediscovery of mind"). You want
to disagree with Searle as to whether computation is intrinsic or an
interpretation. I prefer to disagree with Searle as to whether
intentionality is intrinsic or an interpretation.
> Not only must there be an objective criterion, but also,
>perceptions sort of like ours must be fairly typical in some sense among
>the thoughts in the world. If these two conditions are not met,
>computational reductive functionalism **is disproved**. Period.
No. At most this would prove that there could not be a functionalism
which fits your own preconceived notions. The history of science
suggests that preconceived notions are the enemy of progress. One
must take ones preconceived notions as merely tentative, and subject
to revision. You cannot rule out a reduction which does not come in
the form in which you expect it.
> Why don't we need objective criteria for electrons, without
>invoking a preferred basis? Two reasons. First, electrons are physical
>objects, not mathematical ones the way computations are.
This is not uniformly agreed. There are some philosophers of science
who take an anti-realist position with respect to entities such as
electrons. Bas van Fraassen ("The Scientific Image") argues that
there is no need to believe that unobservables such as electrons are
actually real entities.
> Second, and
>more importantly, whether electrons exist or whether there is some other
>reality that just gives the appearance of electrons, plays no fundamental
>role in what we percieve. It just plain doesn't matter.
Except to the programmer, whether a computer is a digital
computational device, or simply an electic appliance, makes no
fundamental difference to the physical input-output relations that we
perceive.
> Consciousness is the opposite of that. It is the only thing of
>which what we percieve consists. *All* of what we know, relates to it
>and exclusively to it, by definition. It is what we must explain to
>understand the world we percieve. It and nothing else.
Except that no two people can exactly agree as to what consciousness
is.
> Now, we assume that a physical world exists, and has a certain
>mathematical description, because such as assumption seems to allow us to
>make sense of patterns in what we percieve. CRFC is the belief,
>basically, that that physical world is sufficient to account not only for
>what we percieve, but for the fact that we percieve it, in terms of
>computation. Clearly an objective criterion is required.
Or there is Berkeley's approach. Roughly speaking, Berkeley can be
interpreted as arguing that a computational reduction is possible
because even reality can be reduced to computation. (I mention this
example, simply to illustrate that reduction need not fit your
preconceptions.)
> If computation is undefined except as something some conscious
>person can use to organize data, it clearly can play no explanatory
>role. Nothing in our model of the world would give rise to consciousness
>in the first place.
That is not at all clear.
>: > Chalmers "solves" the problem by introducing the Combinatorial
>: >State Automaton (CSA). Unlike its predecessor, the Finite State
>: >Automaton (FSA), the CSA has states that have structure, such as a vector
>: >of bits. In order to prevent the distinction from being merely a
>matter of
>: >nomenclature, which would leave the problem unsolved, Chalmers proposes
>: >that the individual parameters (e.g. bits) must be "independent in some
>: >strong sense."
>: > He suggests that a suitable sense is that the functionals depend
>: >on spacially seperated regions.
>:
>: Presumably this approach requires making assumptions about space as
>: initial a priori committments. But if one does that, one makes it
>: difficult to study to what extent neural computation leads us to
>: discover the nature of space.
> I'm not following.
> It does seem like "compu-physical laws" may be inevitable. That
>would not be reductive, but presumably if we see it's needed, we're
>learning something.
Is it in our genes, that we live in a 3-dimensional world, with a 4th
dimension of time? Or is this something that the neurons discover?
If the latter, then any investigation which presupposes the
3-dimensional world as a base concept on which to build a theory will
have difficulty accounting for the neurons discovering that which is
presupposed.
Jacques Maurice Mallah
Neil Rickert (ric...@cs.niu.edu) wrote:
: > I don't see your point. *Of course* objective criteria are
: >not, there is no sense in which "the world contains human beings thinking
: >about computers" is a different kind of statement than "the world
: >contains pop up toasters thinking about computers".
:
: Imagine an ant. Perhaps it is a super intelligent ant. Is there any
: sense to this ant in saying that "the world contains pop up toasters
: thinking about computers" is a different kind of statement that "the
: world contains humans thinking about computers?"
The answer is obviously yes.
: Whether there are
: ants thinking about computers might have a different kind of
: significance, however.
Not from the statement about humans, no.
Besides, the phrase "intelligent ant" is simply undefined without
a criterion for implementation of computations.
In addition, the ant would have no way to distinguish between
toasters and her fellow mutant giant phaser wielding six legged friends.
: In his argument against reductionism, John Searle claims that
: intentionality is an intrinsic property of reality, but computation
: is only an interpretation (in his "Rediscovery of mind"). You want
: to disagree with Searle as to whether computation is intrinsic or an
: interpretation. I prefer to disagree with Searle as to whether
: intentionality is intrinsic or an interpretation.
I've read Searle, and was not impressed. His "Chinese room"
argument is particularly stupid. His non-computational brand of
reductionism does not seem to have anything plausible about it or have
any explanatory power. To me, intentionality is basically like qualia.
: > Not only must there be an objective criterion, but also,
: >perceptions sort of like ours must be fairly typical in some sense among
: >the thoughts in the world. If these two conditions are not met,
: >computational reductive functionalism **is disproved**. Period.
:
: No. At most this would prove that there could not be a functionalism
: which fits your own preconceived notions. The history of science
: suggests that preconceived notions are the enemy of progress. One
: must take ones preconceived notions as merely tentative, and subject
: to revision. You cannot rule out a reduction which does not come in
: the form in which you expect it.
At this point I will give you two pieces of advice:
Read before you type. I said "computational reductive
functionalism" specifically. If you wish to make a case for a
non-computational reductionism not being disproved by those 2 things,
that has no bearing on my statement.
Secondly, drop phrases like "your preconcieved notions" from your
vocabulary. They serve no purpose other than to irritate the other
person. Of course everyone comes to a debate with certain beliefs.
That's not bad, because otherwise they would have to reinvent the wheel
every time.
The problem comes when people don't take new information, which may
lead to revised beliefs, into account. I have never been guilty of
that. I do rule out a reduction of a sort which I consider implausible,
but you have not presented any new arguments in favor of it.
Chalmers makes several arguments in favor of computationalism. I
suggest you read those. Personally, if a computerized robot acts just
like a human, and uses the same decision making algorithms as a human, I
can't believe he is not concious just because his mathematical
description is nonetheless different from a human's.
This kind of thing is why I specifically asked for input from
computational reductive functionalist / strong AI supporters. You don't
fit into that category. If "we" are forced to abandon our position, "we"
would want to decide among ourselves which direction to go in.
: > Second, and
: >more importantly, whether electrons exist or whether there is some other
: >reality that just gives the appearance of electrons, plays no fundamental
: >role in what we percieve. It just plain doesn't matter.
:
: Except to the programmer, whether a computer is a digital
: computational device, or simply an electic appliance, makes no
: fundamental difference to the physical input-output relations that we
: perceive.
It makes a difference if the computer itself is conscious, or
not, depending on if it is really a computer.
: > Consciousness is the opposite of that. It is the only thing of
: >which what we percieve consists. *All* of what we know, relates to it
: >and exclusively to it, by definition. It is what we must explain to
: >understand the world we percieve. It and nothing else.
:
: Except that no two people can exactly agree as to what consciousness
: is.
That is an irrelevent exaggeration.
: > Now, we assume that a physical world exists, and has a certain
: >mathematical description, because such as assumption seems to allow us to
: >make sense of patterns in what we percieve. CRFC is the belief,
: >basically, that that physical world is sufficient to account not only for
: >what we percieve, but for the fact that we percieve it, in terms of
: >computation. Clearly an objective criterion is required.
:
: Or there is Berkeley's approach. Roughly speaking, Berkeley can be
: interpreted as arguing that a computational reduction is possible
: because even reality can be reduced to computation.
What does that mean? That the laws of physics explicitly involve
the mathematical objects known as computations? I am in fact considering
such an idea, but don't consider it reductive in the traditional sense.
Your rough "explanation" is simply unclear.
: >: > Chalmers "solves" the problem by introducing the Combinatorial
: >: >State Automaton (CSA). Unlike its predecessor, the Finite State
: >: >Automaton (FSA), the CSA has states that have structure, such as a
vector
: >: >of bits. In order to prevent the distinction from being merely a
: >matter of
: >: >nomenclature, which would leave the problem unsolved, Chalmers proposes
: >: >that the individual parameters (e.g. bits) must be "independent in some
: >: >strong sense."
: >: > He suggests that a suitable sense is that the functionals depend
: >: >on spacially seperated regions.
: >:
: >: Presumably this approach requires making assumptions about space as
: >: initial a priori committments. But if one does that, one makes it
: >: difficult to study to what extent neural computation leads us to
: >: discover the nature of space.
:
: > I'm not following.
: > It does seem like "compu-physical laws" may be inevitable. That
: >would not be reductive, but presumably if we see it's needed, we're
: >learning something.
:
: Is it in our genes, that we live in a 3-dimensional world, with a 4th
: dimension of time? Or is this something that the neurons discover?
In our genes? What is that supposed to mean? If it is true,
then I would certainly hope the neurons would discover it. If it's not
true, I hope they would not be so fooled.
: If the latter, then any investigation which presupposes the
: 3-dimensional world as a base concept on which to build a theory will
: have difficulty accounting for the neurons discovering that which is
: presupposed.
On the contrary. As usual (by now), you have it completely
wrong. If you have a theory of reality, it not only can but must describe
your own place in it. If your theory has a 3D world with neurons in it,
then it seems straightforward to suppose that the neurons could perform
experiments to discover that. Then your theory is self-consistent.
If on the other hand your theory assumes that there is no such 3D
world, but then your theory predicts that your neurons will think there
is, then your theory is inconsistent because in fact your neurons are the
one assuming that there is no such world.
Likewise, if your theory leaves the existence or nonexistence of
such a world as something that could be either way with no observable
consequences, then it had better predict that your neurons would come to
the same conclusion.
Neil Rickert
In <4v56bt$u...@news.nyu.edu> mal...@acf2.nyu.edu (Jacques Maurice Mallah) writes:
>Neil Rickert (ric...@cs.niu.edu) wrote:
> At this point I will give you two pieces of advice:
> Read before you type.
I do that.
> I said "computational reductive
>functionalism" specifically. If you wish to make a case for a
>non-computational reductionism not being disproved by those 2 things,
>that has no bearing on my statement.
I am a supporter of computational reduction. I most certainly am not
making a case for non-computational reduction. However, the term
"computation" is itself difficult to define, and perhaps I construe
the term more broadly than you.
> Secondly, drop phrases like "your preconcieved notions" from your
>vocabulary. They serve no purpose other than to irritate the other
>person.
Given the comments you have just made to me, you are in no position
to complain about irritation.
> Chalmers makes several arguments in favor of computationalism. I
>suggest you read those.
I am familiar with Chalmers position on computationalism. My major
disagreement with him is with regard to his position on zombies.
> This kind of thing is why I specifically asked for input from
>computational reductive functionalist / strong AI supporters. You don't
>fit into that category.
I'm sorry that you see it that way. Irritating, or not, I'll have to
assume that your preconceived notions as to what constitutes
computation are too restrictive.
>: Except to the programmer, whether a computer is a digital
>: computational device, or simply an electic appliance, makes no
>: fundamental difference to the physical input-output relations that we
>: perceive.
> It makes a difference if the computer itself is conscious, or
>not, depending on if it is really a computer.
The expressions "the computer itself is conscious" and "is really a
computer" lack the clarity of a precise definition.
>: Or there is Berkeley's approach. Roughly speaking, Berkeley can be
>: interpreted as arguing that a computational reduction is possible
>: because even reality can be reduced to computation.
> What does that mean? That the laws of physics explicitly involve
>the mathematical objects known as computations? I am in fact considering
>such an idea, but don't consider it reductive in the traditional sense.
> Your rough "explanation" is simply unclear.
I had assumed you were familiar with Berkeley, given that he was a
famous philosopher. His position is sometimes expressed as
perception is existence. Paraphrasing his position in modern
terminology, he argued that all that exists is a stream of
information on which we base our perceptions. According to Berkeley,
the objects talked about by physics are then no more than ideas that
we create out of our perceptions of the information stream. Such a
position reduces the objects of physics to information processing, so
amounts to a computational reduction of physics.
Anders N Weinstein
In article <4v25bi$c...@ux.cs.niu.edu>, Neil Rickert <ric...@cs.niu.edu> wrote:
>
>Is it in our genes, that we live in a 3-dimensional world, with a 4th
>dimension of time? Or is this something that the neurons discover?
>If the latter, then any investigation which presupposes the
>3-dimensional world as a base concept on which to build a theory will
>have difficulty accounting for the neurons discovering that which is
>presupposed.
What if the meaning or content of the neuronal states is
context-dependent? So that they only count as representing spatial
relations because they function in the context of a body that moves in
such and such a space? Then you couldn't account for what the neurons
do in informational terms without presupposing the reality of such and
such a space. Yet you still might have a kind of account of neural
adapatation.
I would of course also prefer to say the neurons adapt themselves to
the structure of the environment then that they discover it. They can
adapt themselves to an environmental feature without representing it,
as a fly-ball catching routine, perhaps, might be adapted to parabolic
trajectories without representing them.
In general I think any ascription of content to neural states must
depend on presuppositions about the normal environment in which they
are embedded, so I don't see that this is in any way a bad thing --
more or less unavoidable I would have thought.
Jacques Maurice Mallah
Neil Rickert (ric...@cs.niu.edu) wrote:
: mal...@acf2.nyu.edu (Jacques Maurice Mallah) writes:
: >Neil Rickert (ric...@cs.niu.edu) wrote:
: > I said "computational reductive
: >non-computational reductionism not being disproved by those 2 things,
: >that has no bearing on my statement.
: I am a supporter of computational reduction. I most certainly am not
: making a case for non-computational reduction. However, the term
: "computation" is itself difficult to define, and perhaps I construe
: the term more broadly than you.
You implied that the subjectivity of computation is a point on
which you agree with Searle.
: > Secondly, drop phrases like "your preconcieved notions" from your
: >vocabulary. They serve no purpose other than to irritate the other
: >person.
:
: Given the comments you have just made to me, you are in no position
: to complain about irritation.
On the contrary, they prove that I had been irritated.
I have no intention of dishing out less than I am forced to take.
: > Chalmers makes several arguments in favor of computationalism. I
: >suggest you read those.
:
: I am familiar with Chalmers position on computationalism. My major
: disagreement with him is with regard to his position on zombies.
Then you would agree with him, and me, that an objective
criterion for computation is needed! That seems out of step with your
previous post. Maybe you should read your own post.
Since my notions of what is needed for computations are not so
different from Chalmers, just what notions of mine were you calling
restrictive?
: > This kind of thing is why I specifically asked for input from
: >computational reductive functionalist / strong AI supporters. You don't
: >fit into that category.
:
: I'm sorry that you see it that way. Irritating, or not, I'll have to
: assume that your preconceived notions as to what constitutes
: computation are too restrictive.
So what are your preconcieved notions of what constitutes one?
: >: Except to the programmer, whether a computer is a digital
: >: computational device, or simply an electic appliance, makes no
: >: fundamental difference to the physical input-output relations that we
: >: perceive.
:
: > It makes a difference if the computer itself is conscious, or
: >not, depending on if it is really a computer.
:
: The expressions "the computer itself is conscious" and "is really a
: computer" lack the clarity of a precise definition.
Exactly the problem I want to solve, as you would know if you
were paying attention.
Regardless of that, you know what I mean.
If a satifactory definition is impossible, then computationalism is
false.
BTW Chalmers has some ideas but still hasn't found a good
definition. (I have exchanged emails with him on the subject).
: >: Or there is Berkeley's approach. Roughly speaking, Berkeley can be
: >: interpreted as arguing that a computational reduction is possible
: >: because even reality can be reduced to computation.
:
: > What does that mean? That the laws of physics explicitly involve
: >the mathematical objects known as computations? I am in fact considering
: >such an idea, but don't consider it reductive in the traditional sense.
: > Your rough "explanation" is simply unclear.
:
: I had assumed you were familiar with Berkeley, given that he was a
: famous philosopher. His position is sometimes expressed as
: perception is existence. Paraphrasing his position in modern
: terminology, he argued that all that exists is a stream of
: information on which we base our perceptions. According to Berkeley,
: the objects talked about by physics are then no more than ideas that
: we create out of our perceptions of the information stream. Such a
: position reduces the objects of physics to information processing, so
: amounts to a computational reduction of physics.
As you would know if you'd looked at my home page, I am a
physicist; my interest in this kind of philosophy is due to the need to
use it in a particular application, interpretation of quantum mechanics.
While I may not know too much about famous philosophers, from
what I see, I don't want to. It is no coincidence that Chalmers'
original background is in math, not philosophy.
Your description of Berkeley makes him sound like what I call a
positivist, someone who believes that reality consists only of conscious
perceptions. Therefore, he is reversing the usual purpose of reduction,
which is to reduce consciousness to the more understandable level of
physics.
I suppose he combines that with the idea that consciousness
consists of computations, giving a mathematical description of such a
reality; that is basically like my compuphysical laws idea, but mine has
the advantages that not all computations need be conscious, and that
ordinary physics can be retained to serve as a guide for the computations.
Neil Rickert
> You implied that the subjectivity of computation is a point on
>which you agree with Searle.
I made no such implication, although perhaps you read such an
implication into what I wrote.
>: I am familiar with Chalmers position on computationalism. My major
>: disagreement with him is with regard to his position on zombies.
> Then you would agree with him, and me, that an objective
>criterion for computation is needed! That seems out of step with your
>previous post. Maybe you should read your own post.
On the contrary, I think we have a pretty good idea of what
computation is, and there is no need to precisely define it. We
cannot precisely define length or force, but we still manage to reach
an agreeable consensus. Why should we need to precisely define
computation?
> Since my notions of what is needed for computations are not so
>different from Chalmers, just what notions of mine were you calling
>restrictive?
Your notions are not sufficiently clear to me. Some people identify
computation with that which can be exactly modelled on a Turing
machine. I think such a view of computation is too restrictive, and
leads to the type of confusion shown by Searle, Penrose and others.
I don't expect the theory of Turing machines, or of recursive
functions, to be very important in the question finding a
computational basis for human cognition.
>: > It makes a difference if the computer itself is conscious, or
>: >not, depending on if it is really a computer.
>: The expressions "the computer itself is conscious" and "is really a
>: computer" lack the clarity of a precise definition.
> Exactly the problem I want to solve, as you would know if you
>were paying attention.
> Regardless of that, you know what I mean.
> If a satifactory definition is impossible, then computationalism is
>false.
Should we say that chemistry, and in particular the theory of
combustion is false, simply because we cannot give a satisfactory
definition of "phlogiston." Surely we have the option of simply
saying the "phlogiston" was an unsuitable concept. In much the same
way, further research might persuade us that "consciousness" is an
unsuitable concept.
> As you would know if you'd looked at my home page, I am a
>physicist;
Sorry, I don't have the time to look at the home pages of every
author I read on usenet.
> While I may not know too much about famous philosophers, from
>what I see, I don't want to. It is no coincidence that Chalmers'
>original background is in math, not philosophy.
I assure you that my background is in mathematics, not philosophy.
And I have not drifted nearly as far into philosophy as has
Chalmers.
Anders N Weinstein
In article <4vcqch$h...@ux.cs.niu.edu>, Neil Rickert <ric...@cs.niu.edu> wrote:
>
>Agreed. And if we assume this is possible, then surely the `what' of
>neural representations is irrelevant to the `how'. That is, the
>neurons need to have the ability (the `how') to represent whatever
>type of world they happen to be in.
Not obvious to me our neurons could adapt suitably to every possible
environment that we might imagine. The neurons themselves considered
as learning machines evolved out of simpler stuff in *this* world.
Who knows how they'd do in other possible worlds.
In his early explorations of inductive logic, Carnap wrote a monograph
on the "continuum of inductive methods". As I recall it involved in
effect an infinite set of learning rules which varied in certain
parameters. I also recall the suggestion that which one was optimal
depended entirely on which world it was in. I'd have to look it up.
>Because of just such considerations, I want to avoid making too many
>assumptions about the `what'. Unfortunately, when I look at matters
>in this manner, some people (both Longley and you yourself) suggest
>that I am toying with solipsism.
If you personify the neurons -- take the brain's eye view -- then you
will depict them in exactly the hopeless predicament of empiricist
epistemology. They can only learn of the external world mediately,
through proximal effects. They can indeed only learn whether their
changes are beneficial or not through a proximal training signal of
some sort (the pleasure or pain of traditional empiricism). Indeed
whatever is beyond the interface is irrelevant to the local process
except insofar as it makes some impression on the interface. This is a
kind of solipsistic predicament.
I think Lupton and perhaps, in this message, you, undertake the heroic
task of explaining the possibility of knowledge of the external world
from the "epistemological initial position" of a system locked up
inside the skull behind an interface. Call this the heroic view of the
machine learning project. To me it recalls the early researches of
Reichenbach and Carnap on induction. I am pessimistic about this heroic
project -- I think that as soon as one assumes this initial position,
knowledge of the world becomes impossible.
I am not sure how much *you* really want to undertake the heroic
project, if, like Gibson and like externalists in philosophy, you
concede that the informational content of the states is determined by
the actual environment. The other danger however is that you will be
impelled to some kind of "narrow content" which does not depend on the
environment. Clearly from the "brain's eye view" the environment --
indeed the informational content being transmitted itself -- is wholly
irrelevant to the functioning of the system. But then it is not
explained as representing anything outside itself, and so does not
arrive at any knowledge of the irrelevant external world.
Anyway I am taking as read (from existentialists, if not elsewhere)
that the problem of traditional modern epistemology is hopeless, and
that the way out is to reject the empiricist starting point. In
particular to reject the idea of an interface of thin experience
mediating between the epistemological subject and the world cognized.
This is why I think stressing the personal/sub-personal distinction is
so absolutely crucial to understanding the significance of sub-personal
cognitive science. If we personify it, then the brain *is* in
precisely that hopeless Humean predicament, but it is not an organism,
not a literal subject of phenomenological predicates at all. Whereas
the organism that can be immediately conscious of the world outside
itself is simply *not* in that predicament at all and does not acquire
its view of the world via induction from sense data.
Neil Rickert
In <4vcvuk$f...@usenet.srv.cis.pitt.edu> ande...@pitt.edu (Anders N Weinstein) writes:
>In article <4vcqch$h...@ux.cs.niu.edu>, Neil Rickert <ric...@cs.niu.edu> wrote:
>>Agreed. And if we assume this is possible, then surely the `what' of
>>neural representations is irrelevant to the `how'. That is, the
>>neurons need to have the ability (the `how') to represent whatever
>>type of world they happen to be in.
>Not obvious to me our neurons could adapt suitably to every possible
>environment that we might imagine. The neurons themselves considered
>as learning machines evolved out of simpler stuff in *this* world.
>Who knows how they'd do in other possible worlds.
Perhaps. However the neurons carry out some form of information
processing. The same information processing procedures could quite
possibly be carried out in different worlds, even if the physical
implementation is different from that in our neurons.
>If you personify the neurons -- take the brain's eye view -- then you
>will depict them in exactly the hopeless predicament of empiricist
>epistemology. They can only learn of the external world mediately,
>through proximal effects.
Surely that is exactly the problem to be solved. I don't see it as
hopeless at all.
> They can indeed only learn whether their
>changes are beneficial or not through a proximal training signal of
>some sort (the pleasure or pain of traditional empiricism).
What a surprising comment. For several months you have been
criticizing behaviorism. But now you assert that the behaviorist
learning paradigm is the only one possible.
If one were concerned only with learning by individual neurons,
perhaps there would be a point to your pessimistic comment. But the
neurons form a natural community, and can interoperate
cooperatively.
>I think Lupton and perhaps, in this message, you, undertake the heroic
>task of explaining the possibility of knowledge of the external world
>from the "epistemological initial position" of a system locked up
>inside the skull behind an interface. Call this the heroic view of the
>machine learning project.
But that is the problem that must be solved. Our predicament is one
of the brain in a vat. For sure, the vat is a bony skull, and the
virtual reality machine connected to it is reality itself. But the
predicament is the same. Any valid information processing theory of
what the brain is doing ought to be as valid for the brain in the vat
as it is for the person situated in the world. If a theory cannot
get off the ground without making enormous assumptions about the
nature of reality, then we ought to be highly suspicious of such a
theory.
Note that calling on innate knowledge does not help here, unless you
are a creationist. Any innate knowledge would have needed to
evolve. And the conditions of access to information available to
evolutionary forces are at least as restrictive as those available to
the brain in a vat. If you are an evolutionist, then there is a
serious information processing problem to be solved which cannot be
predicated on a priori knowledge.
> To me it recalls the early researches of
>Reichenbach and Carnap on induction. I am pessimistic about this heroic
>project -- I think that as soon as one assumes this initial position,
>knowledge of the world becomes impossible.
I am also pessimistic about induction, mainly because it depends too
much on having a priori knowledge of the concepts involved in the
hypothesis being put to inductive test.
>I am not sure how much *you* really want to undertake the heroic
>project, if, like Gibson and like externalists in philosophy, you
>concede that the informational content of the states is determined by
>the actual environment.
It is not so much a problem that I want to undertake. Rather, it is
a problem that I have solved. The problem I have to undertake is
that of finding a way of communicating this to a world so skeptical
of its possibility that I cannot get past peer review.
> The other danger however is that you will be
>impelled to some kind of "narrow content" which does not depend on the
>environment.
That does not seem to have been a problem.
> Clearly from the "brain's eye view" the environment --
>indeed the informational content being transmitted itself -- is wholly
>irrelevant to the functioning of the system.
Not at all. That would only be the case if the brain were internally
self sufficient. But if the brain requires external nutrients, then
information relating to those nutrients is highly relevant.
> But then it is not
>explained as representing anything outside itself, and so does not
>arrive at any knowledge of the irrelevant external world.
The problem is one of dealing with how one starts with the tiny
amount of external information that is directly related to nutrition,
and bootstraps it into a comprehensive knowledge of the external
world, much of which is only indirectly related to nutrition.
>Anyway I am taking as read (from existentialists, if not elsewhere)
>that the problem of traditional modern epistemology is hopeless, and
>that the way out is to reject the empiricist starting point.
The only alternative I see is creationism. But if one is a
creationist, there should be no need for science at all. Theology
should suffice.
Neil Rickert
In <4vde7e$i...@usenet.srv.cis.pitt.edu> ande...@pitt.edu (Anders N Weinstein) writes:
>>What a surprising comment. For several months you have been
>>criticizing behaviorism. But now you assert that the behaviorist
>>learning paradigm is the only one possible.
>Not exactly. Just the fact that a local learning procedure cannot make
>any use of an objective standard of goodness or badness that might be
>different from how that standard is proximally represented. What
>matters to neural learning is not whether they are actually doing well
>but whether it "seems" to them -- that is, as far as the information
>that reaches them is concerned -- that they are doing well. They might
>just as well be brains in a vat and all their "beliefs" false but
>useful in predicting the stimuli that reach them.
Yet it seems to me that you are still imposing a behaviorist
standard. You are requiring that the neurons act in a way that
`seems to them that they are doing well.' That is, you are insisting
on a subjective standard for neural activity, and then discovering
that they cannot thereby acquire objective knowledge. Why must the
neurons only act according to a subjective standard?
For a comparison, consider an automobile. Is the engine acting in a
way that `seems to it to be doing well'? Or is the action of the
engine primarily a response to the external information (arising from
tire friction, pressure on the gas pedal, etc). An engine can be
highly responsive to received information, even when it has no
knowledge of the nature or source of that information. Similarly, it
should be possible for the actions of a neuron to be highly
responsive to the information it is processing, even though the
neuron has no knowledge of the source of that information. We would
not expect the actions of the neuron to be strongly contrary to
acting so it `seems to it to be doing well.' But the actions of the
neuron could nevertheless be largely orthogonal to the requirements
of that standard, if those actions are highly responsive to the
information stream.
>>If one were concerned only with learning by individual neurons,
>>perhaps there would be a point to your pessimistic comment. But the
>>neurons form a natural community, and can interoperate
>>cooperatively.
>OK, but the issue concerns how news of success/failure reaches the
>system.
Again, when you reduce it to success/failure, you are imposing a
behaviorist standard. Why should the actions of be neuron be such
that a binary standard of success/failure applies?
>> Any valid information processing theory of
>>what the brain is doing ought to be as valid for the brain in the vat
>>as it is for the person situated in the world.
>1. That is largely why I do not think of mental states as states
>of an information processing theory of the what the brain is doing, but
>something at a higher, supervenient level of description.
I have no problem with that way of talking of `mental states'. Talk
of `mental states' is, after all, only a way of talking as part of a
folk psychological metaphor of uncertain scientific validity.
>2. You talk about "information", presumably a semantic notion. Do you
>think the information is the same in each case? What is this
>informational content then, if not something that answers for its
>correctness to the state of the external world?
I have serious problems with such correspondence theories of truth.
I don't see that they can ever get off the ground. Correctness is,
after all, an attribute of representations. We cannot use the state
of the world as a standard for evaluating representations, unless
there is already a known mapping from states of the world to
representations. But if there is already such a known mapping, then
there is no need for a discipline of science.
It is not as if we know the state of the world, and can use it to
evaluate the correctness of our representations. The problem for the
scientist is the other way around. We must use our representations
of received information so as to attempt to determine the state of
the world.
>I don't see that a functional theory can refrain from assumptions about
>the sort of creature it is in and the sort of environment it has to
>operate in. The story of what neurons do is not an autonomous level of
>description you can study in isolation, it plays a role in the life of
>a creature in an environment. If some neural circuitry functions to
>enable me to catch fly balls you could not discover this by drawing a
>dotted line around the skull and studying only what's inside.
We are talking at cross purpose here. Of course any discussion of
the ability to throw fly balls will depend on the nature of the
environment. But if we are talking about the processes an organism
uses to discover whether the environment is suitable for throwing fly
balls, then we need no such assumptions about the environment.
>>Note that calling on innate knowledge does not help here, unless you
>>are a creationist. Any innate knowledge would have needed to
>>evolve. And the conditions of access to information available to
>>evolutionary forces are at least as restrictive as those available to
>>the brain in a vat. If you are an evolutionist, then there is a
>>serious information processing problem to be solved which cannot be
>>predicated on a priori knowledge.
>How about having knowledge inherited from the culture? Which
>is subject to its own distinctive form of evolution?
Knowledge which arises from the culture is still knowledge that the
individual must acquire by means of information received from the
environment, albeit a cultural environment.
>>It is not so much a problem that I want to undertake. Rather, it is
>>a problem that I have solved.
>Perhaps if I could see the solution I could explain why you have
>not solved *that* problem :-) (or change my tune).
Given your reaction so far, I don't think I can easily present the
solution in a way that you would find persuasive. However you may
take my comments below about astronomy as at least weak hints as to
the direction I am taking.
>>> The other danger however is that you will be
>>>impelled to some kind of "narrow content" which does not depend on the
>>>environment.
>>That does not seem to have been a problem.
>But you suggested that content must be the same in the case of a brain
>in a vat. When does the information get to be about the external world,
>and how does this come about?
The problem for the brain in the vat is not to pretend that its world
does not exist. Surely the problem is to find what is the best
description of its world on the basis of the information received.
It is not much different from the problem of the scientists trying to
discover the best description of their world.
You seem to think this is all impossible. But what do you say of
astronomy? Here we are confined to this solar system. Until very
recently we were confined to this planet. Yet we have been able to
construct a picture of a cosmos with huge galaxies composed of stars
illuminated by the energy from hydrogen fusion. Our picture includes
white dwarfs, red giants, supernovae, gas clouds, etc. We have built
this picture on the basis of local information -- photons detected
within the solar system, and mostly on earth. Presumably the exact
same evidence could have been produced by a cloud of angels shining
flashlights at us. But we have a rather strong conviction that the
picture in terms of stars is a far more likely one than the picture
based on a cloud of angels.
I think, to a large extent, that the progress of astronomy is a
counter example to the claims that empiricism cannot succeed. And
the methods of astronomy provide at least a plausible example of the
type of methodology whereby knowledge can grow based on local
information.
Incidently, I think the Gibsonian metaphor works rather well for
astronomy. The astronomers mainly work by trying to find which parts
of the received signals (photons) can be considered information, and
which can be discarded as noise. They build large telescopes,
interferometers with large baselines, and other types of equipment.
These can all be considered transducers. The astronomers often have
to delay determining what the signals inform them about, until well
after they have built the transducers which allow them to recover
that information.
>>The only alternative I see is creationism. But if one is a
>>creationist, there should be no need for science at all. Theology
>>should suffice.
>The traditional theo-centric view has it that the concepts required for
>objective knowledge derive from those ideas in the mind of god.
In that case, scientists should not study atoms and quarks. They
would be better advised to study theology so that they could come to
comprehend the mind of god.
> The
>modern individualistic view has it that they derive from innate ideas
>in the mind of the individual.
But if this is the case, scientists again should not study atoms and
quarks. They would be better advised to study genetics. For if the
knowledge is innate, then we should study the carriers of that innate
knowledge.
> But you should acknowledge the
>culturalist alternative of which Dewey, Wittgenstein, Heidegger,
>Sellars, Vygotsky, Thomas Kuhn are exemplars: they are embodied in
>social practices, customs or institutions of using symbols which are
>transmitted from generation to generation through socialization.
In that case, scientists should not study atoms and quarks. Instead
they should study social relations.
The fact is, the physical sciences have been highly productive by
ignoring theology, inherited innate knowledge, and sociology. At
least in the case of astronomy, the methodology cannot be much
different from that which you claim cannot succeed. Yet these same
physical sciences have been highly successful, while the successes of
theology, sociology, or a study of inherited knowledge have had far
less success. I think your pessimistic argument does not work.
Anders N Weinstein
In article <4vd3pg$i...@ux.cs.niu.edu>, Neil Rickert <ric...@cs.niu.edu> wrote:
>In <4vcvuk$f...@usenet.srv.cis.pitt.edu> ande...@pitt.edu (Anders N Weinstein) writes:
>> They can indeed only learn whether their
>>changes are beneficial or not through a proximal training signal of
>>some sort (the pleasure or pain of traditional empiricism).
>
>What a surprising comment. For several months you have been
>criticizing behaviorism. But now you assert that the behaviorist
>learning paradigm is the only one possible.
Not exactly. Just the fact that a local learning procedure cannot make
any use of an objective standard of goodness or badness that might be
different from how that standard is proximally represented. What
matters to neural learning is not whether they are actually doing well
but whether it "seems" to them -- that is, as far as the information
that reaches them is concerned -- that they are doing well. They might
just as well be brains in a vat and all their "beliefs" false but
useful in predicting the stimuli that reach them.
If that's a consequence of your theory then I continue to believe they
never really attain knowledge of the actual environment.
>If one were concerned only with learning by individual neurons,
>perhaps there would be a point to your pessimistic comment. But the
>neurons form a natural community, and can interoperate
>cooperatively.
OK, but the issue concerns how news of success/failure reaches the
system.
>>I think Lupton and perhaps, in this message, you, undertake the heroic
>>task of explaining the possibility of knowledge of the external world
>>from the "epistemological initial position" of a system locked up
>>inside the skull behind an interface. Call this the heroic view of the
>>machine learning project.
>
>But that is the problem that must be solved. Our predicament is one
>of the brain in a vat. For sure, the vat is a bony skull, and the
>virtual reality machine connected to it is reality itself. But the
>predicament is the same.
I disagree entirely. It is our neuron's "predicament". *We* are beings
in a world.
> Any valid information processing theory of
>what the brain is doing ought to be as valid for the brain in the vat
>as it is for the person situated in the world.
1. That is largely why I do not think of mental states as states
of an information processing theory of the what the brain is doing, but
something at a higher, supervenient level of description.
2. You talk about "information", presumably a semantic notion. Do you
think the information is the same in each case? What is this
informational content then, if not something that answers for its
correctness to the state of the external world?
>get off the ground without making enormous assumptions about the
>nature of reality, then we ought to be highly suspicious of such a
>theory.
I don't see that a functional theory can refrain from assumptions about
the sort of creature it is in and the sort of environment it has to
operate in. The story of what neurons do is not an autonomous level of
description you can study in isolation, it plays a role in the life of
a creature in an environment. If some neural circuitry functions to
enable me to catch fly balls you could not discover this by drawing a
dotted line around the skull and studying only what's inside.
>Note that calling on innate knowledge does not help here, unless you
>are a creationist. Any innate knowledge would have needed to
>evolve. And the conditions of access to information available to
>evolutionary forces are at least as restrictive as those available to
>the brain in a vat. If you are an evolutionist, then there is a
>serious information processing problem to be solved which cannot be
>predicated on a priori knowledge.
How about having knowledge inherited from the culture? Which
is subject to its own distinctive form of evolution?
>It is not so much a problem that I want to undertake. Rather, it is
>a problem that I have solved.
Perhaps if I could see the solution I could explain why you have
not solved *that* problem :-) (or change my tune).
>> The other danger however is that you will be
>>impelled to some kind of "narrow content" which does not depend on the
>>environment.
>
>That does not seem to have been a problem.
But you suggested that content must be the same in the case of a brain
in a vat. When does the information get to be about the external world,
and how does this come about?
>>Anyway I am taking as read (from existentialists, if not elsewhere)
>>that the problem of traditional modern epistemology is hopeless, and
>>that the way out is to reject the empiricist starting point.
>
>The only alternative I see is creationism. But if one is a
>creationist, there should be no need for science at all. Theology
>should suffice.
The traditional theo-centric view has it that the concepts required for
objective knowledge derive from those ideas in the mind of god. The
modern individualistic view has it that they derive from innate ideas
in the mind of the individual. But you should acknowledge the
Neil Rickert
In <4vakfo$6...@usenet.srv.cis.pitt.edu> ande...@pitt.edu (Anders N Weinstein) writes:
>>Is it in our genes, that we live in a 3-dimensional world, with a 4th
>>dimension of time? Or is this something that the neurons discover?
>>If the latter, then any investigation which presupposes the
>>3-dimensional world as a base concept on which to build a theory will
>>have difficulty accounting for the neurons discovering that which is
>>presupposed.
>What if the meaning or content of the neuronal states is
>context-dependent? So that they only count as representing spatial
>relations because they function in the context of a body that moves in
>such and such a space? Then you couldn't account for what the neurons
>do in informational terms without presupposing the reality of such and
>such a space. Yet you still might have a kind of account of neural
>adapatation.
Well, ok. But then you are concerned with a different problem than
am I. Your concern is with what the neuronal states represent. My
concern is with how they represent.
>I would of course also prefer to say the neurons adapt themselves to
>the structure of the environment then that they discover it.
Agreed. And if we assume this is possible, then surely the `what' of
neural representations is irrelevant to the `how'. That is, the
neurons need to have the ability (the `how') to represent whatever
type of world they happen to be in.
Because of just such considerations, I want to avoid making too many
Peter_...@uk.ibm.com
In <4vcvuk$f...@usenet.srv.cis.pitt.edu>, ande...@pitt.edu (Anders N Weinstein) writes:
<snipetty-snip>
>If you personify the neurons -- take the brain's eye view -- then you
>will depict them in exactly the hopeless predicament of empiricist
>epistemology. They can only learn of the external world mediately,
>through proximal effects. They can indeed only learn whether their
>changes are beneficial or not through a proximal training signal of
>some sort (the pleasure or pain of traditional empiricism). Indeed
>whatever is beyond the interface is irrelevant to the local process
>except insofar as it makes some impression on the interface. This is a
>kind of solipsistic predicament.
>
>I think Lupton and perhaps, in this message, you, undertake the heroic
>task of explaining the possibility of knowledge of the external world
>from the "epistemological initial position" of a system locked up
>inside the skull behind an interface. Call this the heroic view of the
>machine learning project. To me it recalls the early researches of
>Reichenbach and Carnap on induction. I am pessimistic about this heroic
>project -- I think that as soon as one assumes this initial position,
>knowledge of the world becomes impossible.
Since you mention my name, I'll respond to what you say. It seems to me
that what's missing (and has been missing for a very long time) is an
adequate account of ontological change. Why is it and how is it that ontology
is extended? Given that such matters are either taken for granted (in which
case the question of heroism doesn't arise) or subject to very restrictive rules
(such as first-order definability ala Carnap) are exploited, it is no wonder
that the question is considered so heroic.
If we take the heroic stance to be the attempt to show, perhaps deductively,
or to proceed with certainty, or even to prove that from the internal one can
get to the external, that strikes me as obviously impossible. Once one's terms
are internal, not amount of deductive jiggery-pokery is going to change that.
My observation, for what it's worth, is that parsimony produces a higgledy-
piggledy motley crew of ontological structures with no regard for anything
like first-order definability. There is, further, no concern with monotonicity
of the deductive sort. Quite simply all these logicist considerations are
nowhere to be found in structures arrived at through parsimony: whatever
fits, goes. And so far as I can tell, the step from internal processes alone to
external processes causing internal processes is plainly enormously
parsimonious in the relevant way, and that's because the sensory/motor
boundary is, from a parsimony point of view, very much a bleeding edge.
My guess is that's why we products of evolution happen to have this as the
theory we do. Now I don't know how one should see this wrt the heroic task
as Anders sees it, since I don't know whether the heroism presupposes a
deductive/certain/strict definitions procedure for getting inside out, as it were.
So I think that a certain amount of reflection about exactly why there
is this presumed impossibility from some initial position is worth considering.
And I think one will find that what is actually doing the work is ontological
conservatism, a form of conservatism entirely out of place with what I
take parsimony to be.
>I am not sure how much *you* really want to undertake the heroic
The '*you*' above is Neil, not me. I've been pegged as a 'heroic fool'
long ago.
>project, if, like Gibson and like externalists in philosophy, you
>concede that the informational content of the states is determined by
>the actual environment. The other danger however is that you will be
>impelled to some kind of "narrow content" which does not depend on the
>environment. Clearly from the "brain's eye view" the environment --
^^^^^^^^
>indeed the informational content being transmitted itself -- is wholly
>irrelevant to the functioning of the system. But then it is not
^^^^^^^^^
>explained as representing anything outside itself, and so does not
>arrive at any knowledge of the irrelevant external world.
The 'But then' is a non-sequitur and, I believe, false. This means that
the 'Clearly' is also false. From the brain's-eye view, there is a rather
enormous bleeding edge - sensory input, which interlocks with motor
responses in interesting and exciting ways. The question of what is
*explained* and *how* is concerned with whether explanation per
se is associated with parsimony or whether explanation follows
monotonic/deductive/strict definitional restrictions. *That* presumption
counts, in my book, as egregious question-begging.
So if explanation is on the agenda and this involves ontological excesses
such as an external world, the environment (and the fact that sensory data
being information about the environment) will not be irrelevant to how
the system functions nor to or how the system represents its function to
itself, should it, as a system, be able to represent itself as a functioning
system.
Cheers,
Pete Lupton
Anders N Weinstein
In article <4vf60j$h...@zen.hursley.ibm.com>, <Peter_...@uk.ibm.com> wrote:
>In <4vcvuk$f...@usenet.srv.cis.pitt.edu>, ande...@pitt.edu (Anders N Weinstein) writes:
>
>fits, goes. And so far as I can tell, the step from internal processes alone to
>external processes causing internal processes is plainly enormously
>parsimonious in the relevant way, and that's because the sensory/motor
>boundary is, from a parsimony point of view, very much a bleeding edge.
You are saying, I think, that it can be advantageous for a system that
is trapped behind an interface to introduce the hypothesis of external
objects. This as a way, perhaps, of systematizing the stream of
sensations received at the interface. This was basically Quine's view
too -- if physical objects did not exist we would have to invent them.
It is a bit ironic, in that the cognitive system must as it were
overreach its surfaces in order to economically predict what will
happen there.
But I am not really buying. One question is what it even means for such
a system to posit an external object if it really is something whose
significance is exhausted by its role in predicting proximal input, as
Quine supposes. In that case the posit will be equally useful or well
supported whether or not an external object exists, as long as the
stream of proximal input is as predicted. This is so whether or not the
system is in a body on Earth, twin-earth, or a disembodied brain in a
vat. The content of the "posit" would seem to have to be the same in
all cases if that is what it's function is.
In that case, I continue to think we can't legitimately see it as
succeeding in casting its referential purport out to any reality beyond
its surface. Like Quine (on occasion) it seems more appropriate to
think of it as a holistic phenomenalism, in which the meaning of chair
is not connected to real chairs in the world, but as far as the
cognitive system is concerned, is simply one part in a system of
permanent possibilities of sensory input.
>> The other danger however is that you will be
>>impelled to some kind of "narrow content" which does not depend on the
>>environment. Clearly from the "brain's eye view" the environment --
> ^^^^^^^^
>>indeed the informational content being transmitted itself -- is wholly
>>irrelevant to the functioning of the system. But then it is not
> ^^^^^^^^^
>>explained as representing anything outside itself, and so does not
>>arrive at any knowledge of the irrelevant external world.
>
>The 'But then' is a non-sequitur and, I believe, false. This means that
>the 'Clearly' is also false. From the brain's-eye view, there is a rather
>enormous bleeding edge - sensory input, which interlocks with motor
>responses in interesting and exciting ways. The question of what is
>*explained* and *how* is concerned with whether explanation per
>se is associated with parsimony or whether explanation follows
>monotonic/deductive/strict definitional restrictions. *That* presumption
>counts, in my book, as egregious question-begging.
Don't see it. The question is whether your explanation is hanging
informational content on the brain states and, if so, how (and, I guess,
why). Is this informational content individualistically determined, so
that it is the same on Earth and Twin Earth and for a brain in a vat?
If so, it both fails to match our intuitions and also fails to connect
the representations to their world.
If not, then you should not pretend that you are theorizing from the
brain's eye view. You should also acknowledge that the solution to the
problem of objective reference is all smuggled into the notion of
information content which you are hanging on the processes in your
account, not at all from the processes internal parsimony-seeking
structure. I.e. you are somewhere just presupposing, not explaining,
that the system has the ability of doing something I called positing an
external object, which act has the unexplained power of possesing a
content which answers to something external. Or so it seems to me.
Anders N Weinstein
In article <4vf82b$j...@ux.cs.niu.edu>, Neil Rickert <ric...@cs.niu.edu> wrote:
>In <4vdu6k$j...@usenet.srv.cis.pitt.edu> ande...@pitt.edu (Anders N Weinstein) writes:
> As for external
>standards, you could say that our own standard of doing well derives
>from the externally imposed evolutionary requirement of successful
>propogation of the species.
One could say this, but I don't. Doing well for creatures with our
capacities is, in very broad terms, something like leading a full and
productive life. I think this ought to include due regard for the
propagation of the species, but it hardly seems like the supreme goal
of a human life to me. Maybe it is of a fruit-fly's.
>>But this information content is context-dependent, it can be different
>>in the case of a brain on Earth and on Twin-earth and in the case of a
>>brain in a vat the neural activity might have no information content at
>>all. Is that acceptable to you?
>
>As far as I am concerned, the pure scientist is concerned first and
>foremost with information. The content of that information cannot be
>determined until the information is being collected.
I am trying to understand the concept of information at issue, in particular
whether it is the same for a brain on Twin-earth, on earth, in a vat.
I didn't see an answer to that reasonably clear question.
If it is not, then, as I say, a lot of the solution to the problem of
the brain's eye view is packed into the externalist notion of information
content.
>Fair enough. But you are describing the problems of the applied
>scientist. For the applied scientist, there is indeed a
>correspondence available to test the correctness of statements about
>the real world. The problem is different for the pure scientist, who
>is discovering and describing the nature of the real world in the
>first place. In some sense, the pure scientist is constructing the
>correspondence that the applied scientist can use. If the pure
I would say constructing the language or concepts which allow
possibly new features of the world to show up for us.
>>But on a more phenomenological sort of view, we do not necessarily have
>>two things such that a question about correspondence can significantly
>>be raised. That is, we might think of mental states instead as all
>>semantics and no syntax, characterized exhaustively by intentional
>>content alone and unmediated by any representational vehicle, neural or
>>otherwise.
>
>That sounds almost solipsistic. It suggests that a person is a
>disembodied spirit with semantics but no physical body in which to
>represent that semantics.
I think the person *is* a body, so that is not a problem for me. But
also the subject of certain intentional predicates. Perhaps these can
be classed as non-physical, even spiritual. But they only get a grip on
living hence bodily things on my view.
I don't think we use our bodies as representational vehicles, however
(unless we get tatoos or the like). Or, if the states have realizations
in our bodies, the actual structure of the realization drops out of
the account as irrelevant -- all that matters in reason-giving explanation
is content, and public language expression.
>>The way out is to insist there is no way of adequately characterizing such
>>an intentional state that does not presuppose the real existence of the
>>world and of its object.
>
>As a philosopher, it may be important to you to characterize
>intentional states. As a scientist, I am not so concerned with
>them. If I can get everything else right, they will take care of
>themselves.
Sounds like wishful thinking, if you are not in fact concerned with them.
>> And we must insist there is no world-independent way
>>to characterize this state -- it is essentially, not externally, tied
>>to desk d1, the one I pointed to.
>
>I don't see that you are saying much more than that mental states are
>not the same things as brain states. I don't have any disagreement
>with that.
Well I was talking further about how to characterize the nature of the
mental state. Many people who agree with what you concede still disagree
on that.
>The problem is that you seem to be denying the possibility of
>science. In essence you are saying that the scientist cannot begin
>his work until he knows just about everything about the world. But
>if you take that position, there is little left for the scientist to
>do.
Again, I am not mainly talking about science. I am more interested in
the metaphysics of common-sense experience, in what must be possible
for knowledge of the world to exist. I think whatever can be said here
is *also* a pre-requisite for specially scientific knowledge.
>It is my contention that we come about our everyday knowledge of the
>observable world by working as amateur intuitive scientists, doing
>the same sort of thing the scientist does, although perhaps on a
>smaller scale.
This I disagree with. I think that if you formulate a hypothesis you
must have observational data on which to base it. I am interested in what
is required for one to have observational data at all.
>I don't think this distinction between theoretical and actual objects
>will get you very far. If our scientific picture is right, then the
>table in front of you is mostly empty space, but with a sparse array
>of atoms occupying a small portion of that space. If so, then the
>table you see is in some sense a theoretical construct chosen to fit
>your causal relations with the table.
I would question what the data is that the theory of the table is
posited to explain. You mention causal relations -- what does that
mean? Do I observe these relations and then hypothesize a table?
>Why do you want to argue for one over the other? Surely they can be
>considered to be descriptions at different levels. There is no
You miss my point. In arguing that common-sense is like theoretical
science, some philosophers have resorted to an analogy:
sense-data : physical object statements
::
physical-object statement : theoretical statements
I.e. in each case the second term is supposed to involve hypotheses posited to
account for the first. But I take the critique of the given to rule out
the first pair, I don't think the relation could be the same in both cases.
So I don't mind the second pair but resist applying that model to
physical-object statements themselves.
>Fair enough. But you are the one who wants to classify some types of
>knowledge as theoretical. I don't see any important differences
>between perceptual knowledge and instrumental knowledge. Why should
>it matter whether or not the measuring instrument happens to sit in
>our eye sockets?
It matters whether there is an inferential step we could perform.
It matters whether there is an intermediary that could itself be
an object for us to which we might apply an interpretation.
>It doesn't matter whether a scientist studies computational learning
>theory. It matters only that the scientist be able to learn.
And, one might say, it doesn't matter whether a scientist studies
sociology of knowledge. It matters only that the (normal) scientist
should be able to conform to the standards imposed by his or her
paradigm.
>On the contrary, I consider the role of culture to be very
>important. But I don't see that it does what you want it to do. In
>a sense the cultural environment is just a bigger vat, and the
>members of the culture collectively constitute a bigger brain in that
>bigger vat. You are arguing that the brain in a vat approach won't
>work, but then you want to replace it with a bigger brain in a bigger
>vat. If there is a problem with the brain in a vat, then your
>approach does not solve that problem.
I don't understand this at all. Cultural activity is embodied and takes
place in a material environment as well. To take a simple example,
consider the slab game from the beginning of Wittgenstein's
Philosophical Investigations. Well, this is played by embodied
creatures in space and time who operate with noises and large pieces of
rock on the surface of the earth. We can talk about a kind of
non-physical *aspect* of it when we say: the practice involves
conforming to the rule for responding to the order "Slab!" by bringing
a slab. A rule like that is not material or physical, but is not in an
autonomous immaterial world either -- it is immanent, if you like, or
situated, to coin a phrase.
I always think of meanings or rules on that model. As *aspects* --
non-physical aspects, if one likes -- of the uses of signs in the
activity of embodied creatures in the world.
Anders N Weinstein
In article <4vftr9$j...@ux.cs.niu.edu>, Neil Rickert <ric...@cs.niu.edu> wrote:
>
>>> As for external
>>>standards, you could say that our own standard of doing well derives
>>>from the externally imposed evolutionary requirement of successful
>>>propogation of the species.
>
>>One could say this, but I don't. Doing well for creatures with our
>>capacities is, in very broad terms, something like leading a full and
>>productive life.
>
>derives from rather than what our standard is. You seem to have
>confused the two.
I am unclear as to how such a standard could derive from "the requirement
of of successful propogation of the species" if living well by such
a standard is largely superfluous to that end. If it derives from
anything, it would seem to derive from what it is to be a human being
or something like that.
But this is a digression.
>>I am trying to understand the concept of information at issue, in particular
>>whether it is the same for a brain on Twin-earth, on earth, in a vat.
>>I didn't see an answer to that reasonably clear question.
>
>Let me answer with an analogy. In an earlier message I suggested
Let me try to extrapolate to a straight answer. Your conception of
information is individualistic. The information content must be identical
in all three cases. Information content is not individuated by reference to
the world at all, but only by the causal effects of the world on the
surface of the nervous system. Accurate?
In that case the brain's thoughts do not refer to, say, water in the world
any more than to twin-water on Twin-earth or to virtual water in the
simulated vat-world.
I think this is (1) not how we individuate mental content in everyday
practice; (2) not even how cognitive science determines informational
content (see Tyler Burge, "Individualism and Psychology" for a defense
of this claim); and, most important, (3) dooms you to solipsism since
reference to the extramental world is not achieved. It is also a bit
mysterious how communication over a common world of objects is
possible, so science as an intersubjective process also becomes
mysterious.
You seem to be thinking of science as basically a kind of measurement
process. I am wondering if this abets your apparent individualism --
measurement is a process which might be effected by a module in an
individual brain alone , functioning as an instrument. But this is a
quite bizarre to me. I think of scientific measurement as something
that takes place in the world with objects, that requires an awful lot
of symbolic mediation which only training in language and mathematics
can provide. If these capacities are in a brain it is only because
someone taught the person how to manipulate symbols in accordance with
rules (or the person was a genius who invented a new symbolic
technique).
With all this context in place, we can speak of scientific discoveries
as akin to measurement, but a spectrometer can't do science, and the
brain of a scientist is not just a measuring device which arrives at
knowledge by simple detection.
>that all of the astronomical evidence we have could be produced by a
>cloud of angels shining flashlights (with some radio beacons thrown
>in). Let's suppose that I were to discover that in fact that is the
>source of light from the evening sky. Perhaps I learned this via a
>direct pipeline to god.
>
>I claim that this changes nothing. I should still say that the
>heavens are full of stars, nebulae, galaxies, supernovae, etc.
>Describing the heavens in that way allows me to predict future
>observations. It allows me to use the light to align my telescope,
>or for terrestrial navigation. However, if instead I describe the
>heavens in terms of the angels, then I have no clear basis for
>prediction, telescope alignment, or terrestrial navigation. Thus I
>claim the information is of stars, galaxies, nebulae, etc, and
>whether the light is actually produced by a cloud of angels is
>irrelevant.
You can't have it both ways -- that is you can't have us grant by
hypothesis that you know your theory is false and then go on to say
it changes nothing or disproves the correspondence theory. The case just
shows that a false theory might be useful, as the flat earth theory is.
> If I understand you correctly, under the same
>circumstances you would prefer to say that the information is of
>angels with flashlights.
I don't think of finding out about the heavens as an information
pickup process. The main issue concerns the meaning of the claims one
makes. I think that ex hypothesi you have described a case where the
claims of the astronomer have a meaning that is literally false but
nonetheless useful, since there really are no stars as you normally
mean the term.
>If I am right in my understanding of your position, then I see you as
>taking the content of information to be something metaphysical and
>unknowable, while I am taking it to be something useful and
>knowable.
I reject the assumption that basic knowledge of the world is mediated by
interpretation of cues or proximal sense data so I don't think I
have any problem with unknowability. That only applies when one must
make an inference from some datum to a theory, not in consciousness of
the world which is non-inferential, detection of information itself.
>From my perspective, that is a way of saying that you are concerned
>about the unknowable, and that you are insisting that knowledge of
>the unknowable is a prerequisite for scientific knowledge. I would
>rather abandon the unknowable as being unknowable, and concentrate on
>what we can know.
But I also abandon the unknowable. I would like to reject the assumption
which makes it seem to you as if I am talking about something
unknowable, the assumption that knowledge of the world is in all
cases necessarily mediated by representations.
>> I am interested in what
>>is required for one to have observational data at all.
>
>Sensory organs receiving external signals, at least as I am using the
>term.
Necessary but not sufficient, of course. One can't pickup the information
that a postbox affords mailing if one doesn't know what a postal system
is. The information pickup dimension is only the "input" side of the
story, not the whole story (as your measurement view sometimes seems
to have it). One also needs a certain cultivated background as well.
>>I would question what the data is that the theory of the table is
>>posited to explain. You mention causal relations -- what does that
>>mean? Do I observe these relations and then hypothesize a table?
>
>Again, I think you are putting too much emphasis on hypotheses. Your
>neurons discovered these relations, and used them as the basis for
>constructing a table transducer.
Even if this is true I am not sure it counts as itself a rational or
epistemic process. And in any case it does not show that my epistemic
standing vis a vis the table is theoretical, merely that it requires
certain antecedent changes in my neurons which are somewhat akin
to a real scientist making a discovery.
> discovery of patterns of neural signals:
> construction of neural transducers
> ::
> discovery of patterns of scientific measurements:
> construction of scientific instruments
>
>>It matters whether there is an inferential step we could perform.
>>It matters whether there is an intermediary that could itself be
>>an object for us to which we might apply an interpretation.
>
>Alright. Perhaps I am questioning your view of the nature of
>science. Just because we write scientific research papers which
>describe forming hypotheses and making inferences, it does not follow
>that the discovery was actually made on the basis of such
>hypothesizing and inferring. The hypothesis and inference might
>often be a later reconstruction, intended to represent the work in a
>way that meets the expectations for scientific research papers.
That's not the main point. The main point is whether there is an
object for us which stands as intermediary which gets interpreted. If
the scientists plot a table of measurements on graph paper, that is
one sort of representation, and the knowledge they develop from it
is, in a broad sense another. You have two objects of awareness here; one
of which is interpreted and serves as a basis. In that broad sense there
is data and theory, whatever the actual process is.
It is not so when I am conscious of a desk. On a Gibsonian view there
is no more basic data from which a desk is inferred. The process which
constructed the transducer which makes this non- inferential knowledge
possible is not itself an *epistemic* link in my present justification
for belief in the desk.
>>I don't understand this at all. Cultural activity is embodied and takes
>>place in a material environment as well.
>
>Sure. But under your assumptions I don't see that it can explain the
>acquisition of other than cultural knowledge. That is, if activities
>within a culture can lead to discoveries about something outside the
>culture, then why could not activities within the culture of neurons
>in the brain (in a vat) lead to the discovery of something external
>to that brain?
It can, but not if the explanatory framework is individualistic,
limited to treating the brain as an autonomous system. I took it for
granted that our description of culture is not so. Real descriptions of
brain processes are not so either, but you seemed to be defending
individualism.
Jacques Maurice Mallah
:
: implication into what I wrote.
Let's look at what you wrote:
>Neil Rickert (ric...@cs.niu.edu) wrote:
>: I think it is not so clear that objective criteria are required.
>: My preference is to think of an algorithm as an explanatory theory
>: for the operation of a computer. There are alternative explanatory
>: theories such as those which explain in terms of electron motions.
>: The computational theory does not account for everything physical
>: that the computer does. For example, it does not account for the
>: noise produced by the fan designed to cool the processor chip. Yet
>: we still accept the algorithmic description as a reduction of the
>: action of the computer, even though it is not uniquely determined and
>: does not explain everything.
Implying you gave no special significance to computation, while
everything physical that it does might be more significant to you, as to
Searle.
You also wrote:
"In his argument against reductionism, John Searle claims that
intentionality is an intrinsic property of reality, but computation
is only an interpretation (in his "Rediscovery of mind"). You want
to disagree with Searle as to whether computation is intrinsic or an
interpretation. I prefer to disagree with Searle as to whether
intentionality is intrinsic or an interpretation."
Which clearly implies that you agree with him that computation is
only an interpretation.
> If computation is undefined except as something some conscious
>person can use to organize data, it clearly can play no explanatory
>role. Nothing in our model of the world would give rise to consciousness
>in the first place.
And you wrote:
"That is not at all clear."
implying that something other than computation could give rise to
consciousness.
: >: I am familiar with Chalmers position on computationalism. My major
: >: disagreement with him is with regard to his position on zombies.
:
: > Then you would agree with him, and me, that an objective
: >criterion for computation is needed! That seems out of step with your
: >previous post. Maybe you should read your own post.
:
: On the contrary, I think we have a pretty good idea of what
: computation is, and there is no need to precisely define it. We
: cannot precisely define length or force, but we still manage to reach
: an agreeable consensus. Why should we need to precisely define
: computation?
On what grounds do you say we can't precisely define physical
quantities?
Even if we couldn't, it would be an irrellevant example.
Computation is mathematical, not physical; are you saying that pi is not
precisely defined? And we obviously need to precisely define it if we
are going to use the definition to determine what mathematical structures
have conciousness, and which don't.
At the present precision of the definition, we have no grounds for
saying that a toaster has less consciousness than does a human or would an
advanced robot. Does that satisfy you? I would hardly describe such a
situation as saying we have at least reached "an agreeable consensus."
You seem to think that precision is an unattainable goal, while
anything less is just as good as anything else less. That's why debating
with you seems pointless; you seem to have no sense of what progress would
consist of.
: > Since my notions of what is needed for computations are not so
: >different from Chalmers, just what notions of mine were you calling
: >restrictive?
:
: Your notions are not sufficiently clear to me. Some people identify
: computation with that which can be exactly modelled on a Turing
: machine. I think such a view of computation is too restrictive, and
: leads to the type of confusion shown by Searle, Penrose and others.
: I don't expect the theory of Turing machines, or of recursive
: functions, to be very important in the question finding a
: computational basis for human cognition.
I don't see why noncomputable math would play any role. Let's
talk artificial intelligence. Could a digital computer be conscious? I
see absolutely no reason why it could not be just as conscious as a
human.
What kind of computation do you have in mind?
: >: > It makes a difference if the computer itself is conscious, or
: >: >not, depending on if it is really a computer.
:
: >: The expressions "the computer itself is conscious" and "is really a
: >: computer" lack the clarity of a precise definition.
:
: > Exactly the problem I want to solve, as you would know if you
: >were paying attention.
: > Regardless of that, you know what I mean.
: > If a satifactory definition is impossible, then computationalism is
: >false.
:
: Should we say that chemistry, and in particular the theory of
: combustion is false, simply because we cannot give a satisfactory
: definition of "phlogiston." Surely we have the option of simply
: saying the "phlogiston" was an unsuitable concept. In much the same
: way, further research might persuade us that "consciousness" is an
: unsuitable concept.
You said you were a computationalist. If you want to say that
computation is an unsuitable concept, you clearly are not.
We could certainly say that the phlogistonist theory is false,
just as we might have to say that computationalism is false.
I don't know what you mean about "consciousness". Maybe you
meant to say "computation"?
Neil Rickert
In <4vdu6k$j...@usenet.srv.cis.pitt.edu> ande...@pitt.edu (Anders N Weinstein) writes:
>>For a comparison, consider an automobile. Is the engine acting in a
>>way that `seems to it to be doing well'?
>It is not a learning system, and the standard of doing well derives
>from something external.
I agree it is not a learning system. I was mainly using it to point
out that a system can be highly responsive to the information,
without having to support any individual goals. As for external
standards, you could say that our own standard of doing well derives
from the externally imposed evolutionary requirement of successful
propogation of the species.
>But this information content is context-dependent, it can be different
>in the case of a brain on Earth and on Twin-earth and in the case of a
>brain in a vat the neural activity might have no information content at
>all. Is that acceptable to you?
As far as I am concerned, the pure scientist is concerned first and
foremost with information. The content of that information cannot be
determined until the information is being collected.
I don't see that the brain in a vat would have no information
content. The content would derive from whatever virtual reality
machine was feeding information to the neurons. I don't see much
hope for a disconnected brain in a vat with no source of information,
and thus undergoing total sensory deprivation.
>>I have no problem with that way of talking of `mental states'. Talk
>>of `mental states' is, after all, only a way of talking as part of a
>>folk psychological metaphor of uncertain scientific validity.
>I wonder why you call it "metaphor".
>"Talk of tables and chairs is only part of a folk furniture metaphor
>of uncertain scientific validity.
My skepticism is not toward talk of tables and chairs. It is toward
the usefulness of talk about there being mental states involved in
discussions of tables and chairs.
>>I have serious problems with such correspondence theories of truth.
>I'm not actually sure I meant the correspondence theory. Suppose I
>am conscious of my desk. Well consciousness-of requires consciousness-
>that -- say I am conscious *that* this desk (here I point at desk d1) is
>brown.
>On the assumption of the representational theory of the mind we must
>have two things, a proximal representational *vehicle* -- something
>with non-semantic as well as semantic properties -- which bears this
>intentional content, and the distal world which it may or may not
>accurately represent. Perhaps the vehicle is a neural state, I don't
>know. But on the assumption of representations (vehicles) a question
>about correspondence can be raised.
Fair enough. But you are describing the problems of the applied
scientist. For the applied scientist, there is indeed a
correspondence available to test the correctness of statements about
the real world. The problem is different for the pure scientist, who
is discovering and describing the nature of the real world in the
first place. In some sense, the pure scientist is constructing the
correspondence that the applied scientist can use. If the pure
scientist is constructing that correspondence, then it cannot be used
in a truth evaluation of the work of the pure scientist. A truth by
correspondence could not be a standard for the pure scientist, and
perhaps this explains why Kuhn found that scientists could not be
easily described as searching for truth. I am not suggesting that
there is no standard for science -- only that the standard cannot be
based on correspondence.
>But on a more phenomenological sort of view, we do not necessarily have
>two things such that a question about correspondence can significantly
>be raised. That is, we might think of mental states instead as all
>semantics and no syntax, characterized exhaustively by intentional
>content alone and unmediated by any representational vehicle, neural or
>otherwise.
That sounds almost solipsistic. It suggests that a person is a
disembodied spirit with semantics but no physical body in which to
represent that semantics.
>The way out is to insist there is no way of adequately characterizing such
>an intentional state that does not presuppose the real existence of the
>world and of its object.
As a philosopher, it may be important to you to characterize
intentional states. As a scientist, I am not so concerned with
them. If I can get everything else right, they will take care of
themselves.
> And we must insist there is no world-independent way
>to characterize this state -- it is essentially, not externally, tied
>to desk d1, the one I pointed to.
I don't see that you are saying much more than that mental states are
not the same things as brain states. I don't have any disagreement
with that.
>In one way the idea is to recognize that the unit of epistemological
>explanation is not the mind all by itself but an organism-world system,
>and that we have no real conception of a mind as a detachable thing
>that could exist even if there were no world.
The problem is that you seem to be denying the possibility of
science. In essence you are saying that the scientist cannot begin
his work until he knows just about everything about the world. But
if you take that position, there is little left for the scientist to
do.
>*Science* is something a little more special, I freely grant -- it is highly
>mediated by theoretical postualation. I am not mainly concerned with
>scientific knowledge but with everyday knowledge of the observable world.
It is my contention that we come about our everyday knowledge of the
observable world by working as amateur intuitive scientists, doing
the same sort of thing the scientist does, although perhaps on a
smaller scale.
>I think of it this way: we use consciousness of the macro- or
>commonsense world as a base from which to hypothesize theories about
>the unseen micro- (or distant or very large) objects. The scientific
>knowledge is theoretical, justified via inference from what we can
>observe.
I don't think this distinction between theoretical and actual objects
will get you very far. If our scientific picture is right, then the
table in front of you is mostly empty space, but with a sparse array
of atoms occupying a small portion of that space. If so, then the
table you see is in some sense a theoretical construct chosen to fit
your causal relations with the table.
> The empiricist thinks *also* that the common-sense world is a
>hypothesis to account for our "thin" experience -- the stream of
>sensations. But that I reject, since you can't base an inference on a
>sensory state. Rather, I think you don't have experience in the
>full-bodied sense until you are coached into some understanding of a
>common-sense world.
Why do you want to argue for one over the other? Surely they can be
considered to be descriptions at different levels. There is no
reason that the description at the common sense level need be
identical to the description at the fundamental particle level. And
I see no reason to say that one is more theoretical than the other.
Each description is appropriate to the problems being considered at
that level.
>Indeed there is a response to external world skepticism that concedes
>that we are trapped behind a veil of ideas but says we can make an
>"inference to the best explanation" that there is a world beyond.
I see explanation as mainly filling a social and education role. But
I see them as not fundamental to what scientists do. Scientists need
explanations to communicate their discoveries to their culture. But
the discovery is more fundamental than the explanation.
>I think this is deceptive, and I don't like the depiction of
>perceptual knowledge as theoretical.
Fair enough. But you are the one who wants to classify some types of
knowledge as theoretical. I don't see any important differences
between perceptual knowledge and instrumental knowledge. Why should
it matter whether or not the measuring instrument happens to sit in
our eye sockets?
>>The fact is, the physical sciences have been highly productive by
>>ignoring theology, inherited innate knowledge, and sociology. At
>True but irrelevant (and they ignore computational learning theory as
>well).
It doesn't matter whether a scientist studies computational learning
theory. It matters only that the scientist be able to learn.
At this stage I should perhaps comment on terminology. I have been
careful to talk of computational learning theory, rather than machine
learning. The latter suggest learning in a disembodied Turing
machine. I see the prospects for that as quite limited. What I need
is a system with a rich collection of information inputs, a rich
collection of outputs, complex causal relations between the inputs
and outputs, and considerable redundancy in both input and output
information. You can't get that with a disembodied Turing machine.
As you might express it, learning is about something, and without
suitable sources of information there would be nothing for the
learning to be about. You could, at least in principle, meet these
requirements with a brain in a vat suitably connected to some complex
virtual reality machine, although I doubt that it is possible in
practice.
>Anyway I mainly meant to point out an alternative that you do not seem
>to acknowledge -- that the standards or rules which constitute a
>scientific paradigm have their being in social space, not primarily
>in individual minds or the mind of god.
Neil Rickert
In <4vfn59$p...@usenet.srv.cis.pitt.edu> ande...@pitt.edu (Anders N Weinstein) writes:
>In article <4vf82b$j...@ux.cs.niu.edu>, Neil Rickert <ric...@cs.niu.edu> wrote:
>> As for external
>>standards, you could say that our own standard of doing well derives
>>from the externally imposed evolutionary requirement of successful
>>propogation of the species.
>One could say this, but I don't. Doing well for creatures with our
>capacities is, in very broad terms, something like leading a full and
>productive life.
Of course, I agree. But I carefully wrote about what our standard
derives from rather than what our standard is. You seem to have
confused the two.
>>>But this information content is context-dependent, it can be different
>>>in the case of a brain on Earth and on Twin-earth and in the case of a
>>>brain in a vat the neural activity might have no information content at
>>>all. Is that acceptable to you?
>>As far as I am concerned, the pure scientist is concerned first and
>>foremost with information. The content of that information cannot be
>>determined until the information is being collected.
>I am trying to understand the concept of information at issue, in particular
>whether it is the same for a brain on Twin-earth, on earth, in a vat.
>I didn't see an answer to that reasonably clear question.
Let me answer with an analogy. In an earlier message I suggested
that all of the astronomical evidence we have could be produced by a
cloud of angels shining flashlights (with some radio beacons thrown
in). Let's suppose that I were to discover that in fact that is the
source of light from the evening sky. Perhaps I learned this via a
direct pipeline to god.
I claim that this changes nothing. I should still say that the
heavens are full of stars, nebulae, galaxies, supernovae, etc.
Describing the heavens in that way allows me to predict future
observations. It allows me to use the light to align my telescope,
or for terrestrial navigation. However, if instead I describe the
heavens in terms of the angels, then I have no clear basis for
prediction, telescope alignment, or terrestrial navigation. Thus I
claim the information is of stars, galaxies, nebulae, etc, and
whether the light is actually produced by a cloud of angels is
irrelevant. If I understand you correctly, under the same
circumstances you would prefer to say that the information is of
angels with flashlights.
If I am right in my understanding of your position, then I see you as
taking the content of information to be something metaphysical and
unknowable, while I am taking it to be something useful and
knowable.
>>The problem is that you seem to be denying the possibility of
>>science. In essence you are saying that the scientist cannot begin
>>his work until he knows just about everything about the world. But
>>if you take that position, there is little left for the scientist to
>>do.
>Again, I am not mainly talking about science. I am more interested in
>the metaphysics of common-sense experience, in what must be possible
>for knowledge of the world to exist. I think whatever can be said here
>is *also* a pre-requisite for specially scientific knowledge.
From my perspective, that is a way of saying that you are concerned
about the unknowable, and that you are insisting that knowledge of
the unknowable is a prerequisite for scientific knowledge. I would
rather abandon the unknowable as being unknowable, and concentrate on
what we can know.
>>It is my contention that we come about our everyday knowledge of the
>>observable world by working as amateur intuitive scientists, doing
>>the same sort of thing the scientist does, although perhaps on a
>>smaller scale.
>This I disagree with.
Then we will have to agree to disagree.
> I think that if you formulate a hypothesis you
>must have observational data on which to base it.
There, I agree. But I expect we disagree as to what makes data
observational. I also think it is a mistake to describe science in
terms of the formulation of hypotheses. By the time the scientist is
ready to formulate a hypothesis, much of the research may already
have been done. If we ignore what goes on beforehand we will never
be able to completely understand the processes of science.
> I am interested in what
>is required for one to have observational data at all.
Sensory organs receiving external signals, at least as I am using the
term.
>>I don't think this distinction between theoretical and actual objects
>>will get you very far. If our scientific picture is right, then the
>>table in front of you is mostly empty space, but with a sparse array
>>of atoms occupying a small portion of that space. If so, then the
>>table you see is in some sense a theoretical construct chosen to fit
>>your causal relations with the table.
>I would question what the data is that the theory of the table is
>posited to explain. You mention causal relations -- what does that
>mean? Do I observe these relations and then hypothesize a table?
Again, I think you are putting too much emphasis on hypotheses. Your
neurons discovered these relations, and used them as the basis for
constructing a table transducer.
>>Why do you want to argue for one over the other? Surely they can be
>>considered to be descriptions at different levels. There is no
>You miss my point. In arguing that common-sense is like theoretical
>science, some philosophers have resorted to an analogy:
> sense-data : physical object statements
> ::
> physical-object statement : theoretical statements
>I.e. in each case the second term is supposed to involve hypotheses posited to
>account for the first.
I'm not much sold on such a role for hypotheses in either case. My
preference would be to say:
discovery of patterns of neural signals:
construction of neural transducers
::
discovery of patterns of scientific measurements:
construction of scientific instruments
>>Fair enough. But you are the one who wants to classify some types of
>>knowledge as theoretical. I don't see any important differences
>>between perceptual knowledge and instrumental knowledge. Why should
>>it matter whether or not the measuring instrument happens to sit in
>>our eye sockets?
>It matters whether there is an inferential step we could perform.
>It matters whether there is an intermediary that could itself be
>an object for us to which we might apply an interpretation.
Alright. Perhaps I am questioning your view of the nature of
science. Just because we write scientific research papers which
describe forming hypotheses and making inferences, it does not follow
that the discovery was actually made on the basis of such
hypothesizing and inferring. The hypothesis and inference might
often be a later reconstruction, intended to represent the work in a
way that meets the expectations for scientific research papers.
>>On the contrary, I consider the role of culture to be very
>>important. But I don't see that it does what you want it to do. In
>>a sense the cultural environment is just a bigger vat, and the
>>members of the culture collectively constitute a bigger brain in that
>>bigger vat. You are arguing that the brain in a vat approach won't
>>work, but then you want to replace it with a bigger brain in a bigger
>>vat. If there is a problem with the brain in a vat, then your
>>approach does not solve that problem.
>I don't understand this at all. Cultural activity is embodied and takes
>place in a material environment as well.
Sure. But under your assumptions I don't see that it can explain the
Anders N Weinstein
In article <4vdkke$i...@ux.cs.niu.edu>, Neil Rickert <ric...@cs.niu.edu> wrote:
>In <4vde7e$i...@usenet.srv.cis.pitt.edu> ande...@pitt.edu (Anders N Weinstein) writes:
What
>>matters to neural learning is not whether they are actually doing well
>>but whether it "seems" to them -- that is, as far as the information
>>that reaches them is concerned -- that they are doing well. They might
>>just as well be brains in a vat and all their "beliefs" false but
>>useful in predicting the stimuli that reach them.
>
>Yet it seems to me that you are still imposing a behaviorist
>standard. You are requiring that the neurons act in a way that
>`seems to them that they are doing well.' That is, you are insisting
>on a subjective standard for neural activity, and then discovering
>that they cannot thereby acquire objective knowledge. Why must the
>neurons only act according to a subjective standard?
I don't exactly *impose* any such requirement, I merely thought it
was a natural consequence of sticking to an individualistic perspective,
the brain's eye view, the system trapped inside the skull behind the
sensory surfaces.
>For a comparison, consider an automobile. Is the engine acting in a
>way that `seems to it to be doing well'? Or is the action of the
It is not a learning system, and the standard of doing well derives
from something external.
>engine primarily a response to the external information (arising from
>tire friction, pressure on the gas pedal, etc). An engine can be
>highly responsive to received information, even when it has no
>knowledge of the nature or source of that information. Similarly, it
>should be possible for the actions of a neuron to be highly
>responsive to the information it is processing, even though the
>neuron has no knowledge of the source of that information. We would
>not expect the actions of the neuron to be strongly contrary to
>acting so it `seems to it to be doing well.' But the actions of the
>neuron could nevertheless be largely orthogonal to the requirements
>of that standard, if those actions are highly responsive to the
>information stream.
OK. I guess the notion of information at issue is not individualistic,
not determined by what's in the brain alone. So the objective reference
is all packed into the conception of information which you are hanging
on the neuronal activity. And indeed then you do not have a problem.
But this information content is context-dependent, it can be different
in the case of a brain on Earth and on Twin-earth and in the case of a
brain in a vat the neural activity might have no information content at
all. Is that acceptable to you?
>>1. That is largely why I do not think of mental states as states
>>of an information processing theory of the what the brain is doing, but
>>something at a higher, supervenient level of description.
>
>I have no problem with that way of talking of `mental states'. Talk
>of `mental states' is, after all, only a way of talking as part of a
>folk psychological metaphor of uncertain scientific validity.
I wonder why you call it "metaphor".
"Talk of tables and chairs is only part of a folk furniture metaphor
of uncertain scientific validity. Talk of science and scientists, of
experiments, research traditions, professional journals, peer review,
professional conferences, university appointments is only part of a
folk metaphor of uncertain scientific validity. ... "
So what? Anyway I think it is in the mental states of common sense that
we find cognition of the world, not the hidden neural states of
scientific "psychology". People qua knowers are common-sense objects, not
scientific ones, on my view.
>>2. You talk about "information", presumably a semantic notion. Do you
>>think the information is the same in each case? What is this
>>informational content then, if not something that answers for its
>>correctness to the state of the external world?
>
>I have serious problems with such correspondence theories of truth.
I'm not actually sure I meant the correspondence theory. Suppose I
am conscious of my desk. Well consciousness-of requires consciousness-
that -- say I am conscious *that* this desk (here I point at desk d1) is
brown.
On the assumption of the representational theory of the mind we must
have two things, a proximal representational *vehicle* -- something
with non-semantic as well as semantic properties -- which bears this
intentional content, and the distal world which it may or may not
accurately represent. Perhaps the vehicle is a neural state, I don't
know. But on the assumption of representations (vehicles) a question
about correspondence can be raised.
But on a more phenomenological sort of view, we do not necessarily have
two things such that a question about correspondence can significantly
be raised. That is, we might think of mental states instead as all
semantics and no syntax, characterized exhaustively by intentional
content alone and unmediated by any representational vehicle, neural or
otherwise. This move by itself doesn't solve the problem -- witness
Searle, whose view is like this -- but I think it makes room for a
better view.
The way out is to insist there is no way of adequately characterizing such
an intentional state that does not presuppose the real existence of the
world and of its object. I did that already by characterizing it as
consciousness *of* my desk -- after that description, you can't go on to
raise questions about whether the desk or the spatial world exists or
not, for example. And we must insist there is no world-independent way
to characterize this state -- it is essentially, not externally, tied
to desk d1, the one I pointed to. (Perhaps I only pointed "in thought", but
pointing in thought is derivative from pointing in space, I think.) If you
want to know whether the claim made by my state is true, you have to look
at my situation to find out which desk to examine.
A brain in a vat simply couldn't get into such a mental state -- it
can't point to d1, and so can't perceive it under a demonstrative mode
of presentation. Only an embodied creature in the same world as the desk
could.
In one way the idea is to recognize that the unit of epistemological
explanation is not the mind all by itself but an organism-world system,
and that we have no real conception of a mind as a detachable thing
that could exist even if there were no world.
>I don't see that they can ever get off the ground. Correctness is,
>after all, an attribute of representations.
Ambiguous: it may be an attribute of intentional states conceived in
the phenomenologist's manner.
> We cannot use the state
>of the world as a standard for evaluating representations, unless
We seem to do it all the time. If you tell me a desk has been moved into my
office I might walk into my office and take a look to compare what you
said with reality, for example. I turn to it, and, perhaps, acquire the
objective information that a desk is in my office. Mundane things like
this are what "comparing representations with reality" is.
I have compared *your* representation -- the assertion you made -- with
the state of the world. But if I do in fact see the desk I do not
encounter in my consciousness another representation which needs to be
compared with reality, my state is rather one of consciousness of the
desk as a desk.
>there is already a known mapping from states of the world to
>representations. But if there is already such a known mapping, then
>there is no need for a discipline of science.
*Science* is something a little more special, I freely grant -- it is highly
mediated by theoretical postualation. I am not mainly concerned with
scientific knowledge but with everyday knowledge of the observable world.
>It is not as if we know the state of the world, and can use it to
>evaluate the correctness of our representations. The problem for the
>scientist is the other way around. We must use our representations
>of received information so as to attempt to determine the state of
>the world.
I think of it this way: we use consciousness of the macro- or
commonsense world as a base from which to hypothesize theories about
the unseen micro- (or distant or very large) objects. The scientific
knowledge is theoretical, justified via inference from what we can
observe. The empiricist thinks *also* that the common-sense world is a
hypothesis to account for our "thin" experience -- the stream of
sensations. But that I reject, since you can't base an inference on a
sensory state. Rather, I think you don't have experience in the
full-bodied sense until you are coached into some understanding of a
common-sense world.
So yes, in theoretical knowledge our knowledge of the commonsense world
provides data which we must interpret to construct a theory. But
in observational knowledge such as becoming conscious of my desk there
is no more basic data which must be interpreted, just consciousness
of my desk.
>>How about having knowledge inherited from the culture? Which
>>is subject to its own distinctive form of evolution?
>
>Knowledge which arises from the culture is still knowledge that the
>individual must acquire by means of information received from the
>environment, albeit a cultural environment.
Not necessarily, I think. It may be that the upshot of socialization is
to transform the information you can detect, but it's effect is not
itself transduced or mediated by representational processes.
>>Perhaps if I could see the solution I could explain why you have
>>not solved *that* problem :-) (or change my tune).
>
>Given your reaction so far, I don't think I can easily present the
>solution in a way that you would find persuasive. However you may
I am more sympathetic than I let on. But I am actually very interested
in understanding via concrete example (I can't really understand anything
without examples, I don't think many people really can)
>You seem to think this is all impossible. But what do you say of
>astronomy? Here we are confined to this solar system. Until very
>recently we were confined to this planet. Yet we have been able to
>construct a picture of a cosmos with huge galaxies composed of stars
>illuminated by the energy from hydrogen fusion. Our picture includes
>white dwarfs, red giants, supernovae, gas clouds, etc. We have built
>this picture on the basis of local information -- photons detected
Well see above on the difference between mediated and unmediated
knowledge.
Indeed there is a response to external world skepticism that concedes
that we are trapped behind a veil of ideas but says we can make an
"inference to the best explanation" that there is a world beyond.
I think this is deceptive, and I don't like the depiction of
perceptual knowledge as theoretical. I never formed the hypothesis
of my office to account for a sensation. In any case one can't really
base a theory of the common-sense world on the foundation of thin
experience anyway, given that thin experience doesn't stand in
a rational inferential relation to propositions.
>> But you should acknowledge the
>>culturalist alternative of which Dewey, Wittgenstein, Heidegger,
>>Sellars, Vygotsky, Thomas Kuhn are exemplars: they are embodied in
>>social practices, customs or institutions of using symbols which are
>>transmitted from generation to generation through socialization.
>
>In that case, scientists should not study atoms and quarks. Instead
>they should study social relations.
A gross fallacy. Physical scientists as such should stick to studying
atoms and quarks. On the culturalist view it is only meta-scientists,
those who want to understand the nature of scientific understanding who
would have to think about (a certain kind of) socially inculcated
normative structure.
Such facts, perhaps, as that the rules or standards for what could or
could not be a possible feature of atoms, indeed for what is a good
problem, or something in need of explanation at all, are seldom
invented ex nihilo by an individual brain, but largely inherited from
an ongoing explanation- seeking practice into which new members are
initiated through training, typically in graduate school.
I had thought you yourself had bumped up hard against the socially
inculcated and maintained pre-judgement of the more dominant scientific
community, which prescribes what counts as an acceptable explanation.
>The fact is, the physical sciences have been highly productive by
>ignoring theology, inherited innate knowledge, and sociology. At
True but irrelevant (and they ignore computational learning theory as
well). A culturalist does not advocate that physical scientists study
sociology as part of their scientific work. On the other hand, many
scientists are like Longley in being suckered by the myth that science,
by virtue of its special method, attains to an understanding that is
wholly free of culturally determined prejudice or "preconceived
notions". Appreciating that this is untrue -- indeed, probably
impossible in principle, given the need for the imposition of standards
on phenomena to cognize them at all -- might require a more reflective
inquiry. One in which the role of a socially inculcated background in
determining the bounds of what is and isn't a possible object might
come to the fore.
>least in the case of astronomy, the methodology cannot be much
>different from that which you claim cannot succeed. Yet these same
>physical sciences have been highly successful, while the successes of
>theology, sociology, or a study of inherited knowledge have had far
>less success. I think your pessimistic argument does not work.
If the culturalists are right at least some of the science of science
is necessarily a soft science. So what? Your response fails to
distinguish first-order scientific theorizing about the world from
reflective or meta-scientific analysis of first-order science.
Peter Lupton
In article: <4vfju7$p...@usenet.srv.cis.pitt.edu> ande...@pitt.edu (Anders N Weinstein) writes:
>
> In article <4vf60j$h...@zen.hursley.ibm.com>, <Peter_...@uk.ibm.com> wrote:
> >In <4vcvuk$f...@usenet.srv.cis.pitt.edu>, ande...@pitt.edu (Anders N Weinstein) writes:
> >
> >fits, goes. And so far as I can tell, the step from internal processes alone to
> >external processes causing internal processes is plainly enormously
> >parsimonious in the relevant way, and that's because the sensory/motor
> >boundary is, from a parsimony point of view, very much a bleeding edge.
>
> You are saying, I think, that it can be advantageous for a system that
> is trapped behind an interface to introduce the hypothesis of external
> objects. This as a way, perhaps, of systematizing the stream of
> sensations received at the interface. This was basically Quine's view
> too -- if physical objects did not exist we would have to invent them.
> It is a bit ironic, in that the cognitive system must as it were
> overreach its surfaces in order to economically predict what will
> happen there.
>
> But I am not really buying. One question is what it even means for such
> a system to posit an external object if it really is something whose
> significance is exhausted by its role in predicting proximal input, as
> Quine supposes. In that case the posit will be equally useful or well
> supported whether or not an external object exists, as long as the
> stream of proximal input is as predicted.
I *think* you're observing that I cannot, with certainty, exclude the
possibility that my body is literally the whole universe. And the surface
of that universe (which happens to include my sensory nerve-endings) just
happens to buzz just exactly *this* way. I think that case is impossible
to exclude, if one is concerned with absolute certainty. Externalism
doesn't help -- that just shifts uncertainty about whether what one means
is true to uncertainty about whether it is true that one means.
> ......This is so whether or not the
> system is in a body on Earth, twin-earth, or a disembodied brain in a
> vat. The content of the "posit" would seem to have to be the same in
> all cases if that is what it's function is.
Not necessarily. I take it that the sensory data are sufficiently
organised (in a way parsimony can pick out) for indexicality to
still be relevant here. That permits the same externalist manoeuvres
one is used to.
> In that case, I continue to think we can't legitimately see it as
> succeeding in casting its referential purport out to any reality beyond
> its surface.
Don't see this at all. It's all far too quick.
> Like Quine (on occasion) it seems more appropriate to
> think of it as a holistic phenomenalism, in which the meaning of chair
> is not connected to real chairs in the world, but as far as the
> cognitive system is concerned, is simply one part in a system of
> permanent possibilities of sensory input.
This may be fine for Quine, but not for me.
> >> The other danger however is that you will be
> >>impelled to some kind of "narrow content" which does not depend on the
> >>environment. Clearly from the "brain's eye view" the environment --
> > ^^^^^^^^
> >>indeed the informational content being transmitted itself -- is wholly
> >>irrelevant to the functioning of the system. But then it is not
> > ^^^^^^^^^
> >>explained as representing anything outside itself, and so does not
> >>arrive at any knowledge of the irrelevant external world.
> >
> >The 'But then' is a non-sequitur and, I believe, false. This means that
> >the 'Clearly' is also false. From the brain's-eye view, there is a rather
> >enormous bleeding edge - sensory input, which interlocks with motor
> >responses in interesting and exciting ways. The question of what is
> >*explained* and *how* is concerned with whether explanation per
> >se is associated with parsimony or whether explanation follows
> >monotonic/deductive/strict definitional restrictions. *That* presumption
> >counts, in my book, as egregious question-begging.
>
> Don't see it. The question is whether your explanation is hanging
> informational content on the brain states and, if so, how (and, I guess,
> why). Is this informational content individualistically determined, so
> that it is the same on Earth and Twin Earth and for a brain in a vat?
As far as I'm concerned, the externalist moves are fine and dandy.
They incorporate how things are into what meaning is. There's nothing
wrong with that, but one must distinguish such an account from what
I, qua individual, can distinguish using my own individualistic resources.
I can, as an individual, come to the view that if certain conditions aren't
met, then certain things I might think can't mean anything. Fine. But that
isn't to say I can know it. What happens is that I don't mean anything by
some things I think I mean, but as it happens I don't know that I don't
mean anything.
What I'm suggesting here is that the internalism/externalism debate isn't
relevant to this discussion. However, for what it's worth, I think that
parsimony will produce the result that internalism is an unstable state
which will inevitably drift towards externalism.
> If so, it both fails to match our intuitions and also fails to connect
> the representations to their world.
I don't think the 'If so' is realised.
> If not, then you should not pretend that you are theorizing from the
> brain's eye view.
Perhaps not. But then I don't know which step was not from that view
and since you haven't specified one, I guess you don't either. But I
think your problem is that you think 'internalism' equals 'brain's-
eye view'. That's what's at issue here. I'm not going to just lay down
and let you assert it, although I'll accept that historically this
connection has held. But that's history and not a piece of reasoning.
The problem for you is just this: what's to stop the brain's-eye view
from making the step to accepting meanings as involving truth? Only
one thing so far as I can tell: that the brain's-eye view doesn't
have a notion of truth at all. But that isn't actually a discussion
about internalism/exernalism: it's a discussion about the way a local
notion of fit (parsimony) permits the brain's-eye view to latch onto
and exploit non-local regularities as a matter of course. To put it
another way, even though parsimony is a local notion of fit, the
representational structures created are representational of non-local
structures. And we can see *that* they're representational of non-local
structure if for no other reason than that they represent the bleeding
edge of the brain's sensorimotor boundarary. But not now *as* a
bleeding edge, but as a part of a greater whole.
> ....You should also acknowledge that the solution to the
> problem of objective reference is all smuggled into the notion of
> information content which you are hanging on the processes in your
> account, not at all from the processes internal parsimony-seeking
> structure. I.e. you are somewhere just presupposing, not explaining,
So you say. But you don't say where or how. Nothing you say above relates
at all directly to what I've been saying. So I guess that if I do make
the mistake you attribute to me, you don't know where and how the
mistake is made. In short, you're just saying that.
> that the system has the ability of doing something I called positing an
> external object, which act has the unexplained power of possesing a
> content which answers to something external. Or so it seems to me.
These phrases 'positing an external object', 'possessing a content',
'answering to something external' - very important, but you haven't
related them at all to the structure of parsimony I'm advocating.
So my question is: what is it about positing an external object, say,
that makes such an activity impossible for a parsimony-guided
system to achieve? Now I've been thinking about this for some time,
and I haven't yet thought of anything. But if you can pick out
something firm (that is, something not question-begging) that a
parsimony-guided system can't do, I'd be very interested.
---
Cheers,
Pete Lupton
Neil Rickert
In <4vfjpf$q...@news.nyu.edu> mal...@acf2.nyu.edu (Jacques Maurice Mallah) writes:
>Neil Rickert (ric...@cs.niu.edu) wrote:
>: mal...@acf2.nyu.edu (Jacques Maurice Mallah) writes:
>: > You implied that the subjectivity of computation is a point on
>: >which you agree with Searle.
>: I made no such implication, although perhaps you read such an
>: implication into what I wrote.
> Let's look at what you wrote:
>>Neil Rickert (ric...@cs.niu.edu) wrote:
>>: I think it is not so clear that objective criteria are required.
>>: My preference is to think of an algorithm as an explanatory theory
>>: for the operation of a computer. There are alternative explanatory
>>: theories such as those which explain in terms of electron motions.
>>: The computational theory does not account for everything physical
>>: that the computer does. For example, it does not account for the
>>: noise produced by the fan designed to cool the processor chip. Yet
>>: we still accept the algorithmic description as a reduction of the
>>: action of the computer, even though it is not uniquely determined and
>>: does not explain everything.
> Implying you gave no special significance to computation, while
>everything physical that it does might be more significant to you, as to
>Searle.
Well, I didn't imply that there is no special significance to
computation. If I thought that, I would not be working in a computer
science department. You are reading things into what I wrote that
simply are not there.
What I suggested, is that there are multiple ways of describing the
operation of a computer. The computational description is one of
them, and surely the most useful given the type of work that I do.
> You also wrote:
>"In his argument against reductionism, John Searle claims that
>intentionality is an intrinsic property of reality, but computation
>is only an interpretation (in his "Rediscovery of mind"). You want
>to disagree with Searle as to whether computation is intrinsic or an
>interpretation. I prefer to disagree with Searle as to whether
>intentionality is intrinsic or an interpretation."
> Which clearly implies that you agree with him that computation is
>only an interpretation.
Right. So what is your problem. In front of me is a screen covered
with colored pixels. One interpretation is that the screen contains
words which you have written. The fact that it is an interpretation
does not imply there is anything subjective involved. I'm responding
with the aid of an editor. The editor results from the
interpretation of C-code by the C-compiler. The Pascal compiler
could not have made the same interpretation of that code. This does
not imply that there is anything subjective about there being an
editor.
>> If computation is undefined except as something some conscious
>>person can use to organize data, it clearly can play no explanatory
>>role. Nothing in our model of the world would give rise to consciousness
>>in the first place.
> And you wrote:
>"That is not at all clear."
> implying that something other than computation could give rise to
>consciousness.
Actually, it only implied that you wrote something that was not at
all clear. It does not obviously follow from "computation is
undefined" that "nothing in our model of the world would give rise to
consciousness."
It seems likely enough to me that computation can be used in ways
that give rise to consciousness. And it seems clear to me that
computation exists in our model of the world, even if it is
undefined.
>: > Then you would agree with him, and me, that an objective
>: >criterion for computation is needed! That seems out of step with your
>: >previous post. Maybe you should read your own post.
>: On the contrary, I think we have a pretty good idea of what
>: computation is, and there is no need to precisely define it. We
>: cannot precisely define length or force, but we still manage to reach
>: an agreeable consensus. Why should we need to precisely define
>: computation?
> On what grounds do you say we can't precisely define physical
>quantities?
I didn't say that we cannot precisely define quantities. I said that
we cannot precisely define length or force. We can define the
measuring conventions by means of which we quantitatively determine
length. But we cannot precisely define what it actually is that we
measure. We can define the quantities, but not the concepts which
those quantities refer to.
Prove me wrong by precisely defining the concept of length.
> Even if we couldn't, it would be an irrellevant example.
>Computation is mathematical, not physical; are you saying that pi is not
>precisely defined?
No, not at all. But I am saying that numbers are not precisely
defined. We know exactly which number pi is, but we cannot agree on
what is a number. Is a number something that exists in plato's
heaven and has been discovered by mathematicians? Is a number the
class of all sets with the same cardinality? Is a number a set? Or
is it that we merely have names for numbers, but the numbers
themselves do not exist? There is no agreement about what numbers
actually are.
> At the present precision of the definition, we have no grounds for
>saying that a toaster has less consciousness than does a human or would an
>advanced robot. Does that satisfy you?
I think it quite obvious that a toaster does not have the same
consciousness as a toaster. Even the operational definition used by
a medical practitioner is enough to determine that. If your
definition of `consciousness' cannot distinguish, then you should
stop pretending that you have a definition.
> You seem to think that precision is an unattainable goal, while
>anything less is just as good as anything else less.
No, you are stating that with too much specificity. I think that
precision in defining `consciousness' is an unattainable goal. The
reason I think this, is that I think `consiousness' is an inherently
fuzzy concept which therefore cannot be made precise. I think that
even if we come up with a way of constructing conscious robots, we
will still consider `consciousness' to be a fuzzy concept.
> That's why debating
>with you seems pointless; you seem to have no sense of what progress would
>consist of.
The debate may be pointless, but not for the reason you give. It is
my sense that we make progress by finding concepts that we can deal
with, rather than by foolishly attempting to take inherently fuzzy
concepts and somehow make them precise.
>: > Since my notions of what is needed for computations are not so
>: >different from Chalmers, just what notions of mine were you calling
>: >restrictive?
>: Your notions are not sufficiently clear to me. Some people identify
>: computation with that which can be exactly modelled on a Turing
>: machine. I think such a view of computation is too restrictive, and
>: leads to the type of confusion shown by Searle, Penrose and others.
>: I don't expect the theory of Turing machines, or of recursive
>: functions, to be very important in the question finding a
>: computational basis for human cognition.
> I don't see why noncomputable math would play any role.
Well, I never suggested that non-computable mathematics would play
any role. The concept of `non-computable' derives from Turing
machines. I rejected the centrality of Turing machines, so it is
strange that you would accuse me of suggesting a role for
non-computable mathematics.
> Let's
>talk artificial intelligence. Could a digital computer be conscious?
If, by a digital computer, you mean the type of appliance sitting on
my desk, then I don't believe one could be conscious in any useful
sense of the word. It is disconnected from reality, so there is
nothing for it to be conscious of. If, on the other hand, you are
asking whether a conscious robot could be constructed using digital
technology, then I see no principled reasons why not, although there
might be many practical difficulties.
> What kind of computation do you have in mind?
I listen to music from a CD player. In my book, it operates using
computation. I cannot adequately describe it with a Turing machine
model. So it uses computation in a way that is not captured by that
model. The Turing machine model only considers what is left on the
tape after a HALT state is entered. What is important for the CD
player, is the stream of information being exchanged before a HALT
state is entered. More generally, the Turing machine model does not
adequately account for real time computing.
>: Should we say that chemistry, and in particular the theory of
>: combustion is false, simply because we cannot give a satisfactory
>: definition of "phlogiston." Surely we have the option of simply
>: saying the "phlogiston" was an unsuitable concept. In much the same
>: way, further research might persuade us that "consciousness" is an
>: unsuitable concept.
> You said you were a computationalist. If you want to say that
>computation is an unsuitable concept, you clearly are not.
I didn't say that computation is an unsuitable concept. I said that
consciousness is an unsuitable concept. By that, I intended that it
is unsuitable for the type of problems investigated in AI.
Anders N Weinstein
In article <617497...@luptonpj.demon.co.uk>,
Peter Lupton <lup...@luptonpj.demon.co.uk> wrote:
>In article: <4vfju7$p...@usenet.srv.cis.pitt.edu> ande...@pitt.edu (Anders N Weinstein) writes:
>>
>> Don't see it. The question is whether your explanation is hanging
>> informational content on the brain states and, if so, how (and, I guess,
>> why). Is this informational content individualistically determined, so
>> that it is the same on Earth and Twin Earth and for a brain in a vat?
>
>As far as I'm concerned, the externalist moves are fine and dandy.
>They incorporate how things are into what meaning is. There's nothing
>wrong with that, but one must distinguish such an account from what
>I, qua individual, can distinguish using my own individualistic resources.
You qua individual are not trapped behind an interface as your brain is.
>> If not, then you should not pretend that you are theorizing from the
>> brain's eye view.
>
>Perhaps not. But then I don't know which step was not from that view
>and since you haven't specified one, I guess you don't either.
> To put it
>another way, even though parsimony is a local notion of fit, the
>representational structures created are representational of non-local
>structures. And we can see *that* they're representational of non-local
>structure if for no other reason than that they represent the bleeding
>edge of the brain's sensorimotor boundarary. But not now *as* a
>bleeding edge, but as a part of a greater whole.
I'm just trying to understand how objective reference comes into
the explanation, given that purely internal parsimony-guidedness is a
reference-free notion. At some point you think it does, and I'd like
to understand how. The proximal regularities are the same on Earth
and Twin-Earth.
>So my question is: what is it about positing an external object, say,
>that makes such an activity impossible for a parsimony-guided
>system to achieve? Now I've been thinking about this for some time,
>and I haven't yet thought of anything. But if you can pick out
>something firm (that is, something not question-begging) that a
>parsimony-guided system can't do, I'd be very interested.
Recall Searle on the difference between what a computer can do and
what it can do *in virtue of being* a computer (qua computer, you might
say). A prime-factorization program in execution can bring about changes
in the room temperature near the hardware, but it is not in virtue of
being a prime-factorization program that it does so.
In this sense there may be nothing that a parsimony-guided system can't
do -- it might make an omelette, for example. But it is not in virtue of
being a purely local parsimony guided system that it does this.
I am suggesting that a (local) parsimony-guided system *as such* never
does posit an distal object, it just looks for proximal patterns.
Unless like Quine one thinks these are the same thing, which is a kind
of phenomenalism.
Ed Faith
Neil Rickert <ric...@cs.niu.edu> wrote in article
<4vf82b$j...@ux.cs.niu.edu>...
> In <4vdu6k$j...@usenet.srv.cis.pitt.edu> ande...@pitt.edu (Anders N
Weinstein) writes:
> >I think of it this way: we use consciousness of the macro- or
> >commonsense world as a base from which to hypothesize theories about
> >the unseen micro- (or distant or very large) objects. The scientific
> >knowledge is theoretical, justified via inference from what we can
> >observe.
New devices allow us to deal with individual atoms in a way that
begins to approach the way we deal with ordinary objects.
Researchers can spell things like "IBM" with a figurative
handful of atoms. Here (conscious) inference isn't going
on, i.e., in the sense of "the spectrum is like so, therefore
the atom must be like so". If atoms nevertheless remain
theoretical entities despite this "direct" interaction (achieved
through mechanical prostheses), then I don't see how you
can deny that tables are theoretical entities.
> I don't think this distinction between theoretical and actual objects
> will get you very far. If our scientific picture is right, then the
> table in front of you is mostly empty space, but with a sparse array
> of atoms occupying a small portion of that space. If so, then the
> table you see is in some sense a theoretical construct chosen to fit
> your causal relations with the table.
Or vice versa, i.e., we can deny that atoms remain merely
theoretical once we've achieved a certain fluency of
interaction with them.
Neil Rickert
In <4vg6en$r...@usenet.srv.cis.pitt.edu> ande...@pitt.edu (Anders N Weinstein) writes:
>In article <4vftr9$j...@ux.cs.niu.edu>, Neil Rickert <ric...@cs.niu.edu> wrote:
>>>I am trying to understand the concept of information at issue, in particular
>>>whether it is the same for a brain on Twin-earth, on earth, in a vat.
>>>I didn't see an answer to that reasonably clear question.
>>Let me answer with an analogy. In an earlier message I suggested
>Let me try to extrapolate to a straight answer. Your conception of
>information is individualistic. The information content must be identical
>in all three cases. Information content is not individuated by reference to
>the world at all, but only by the causal effects of the world on the
>surface of the nervous system. Accurate?
Not really. We must also consider the causal effects of the nervous
system on the world.
>In that case the brain's thoughts do not refer to, say, water in the world
>any more than to twin-water on Twin-earth or to virtual water in the
>simulated vat-world.
If twin earth water is chemically different from earth water, then
some (but not all) of the brain's thoughts about water might rule out
twin-earth water. If virtual water in a simulated vat world is
indistinguishable from real water in the real world, then it is
foolish to say that the brain's thoughts distinguish that which is
indistinguishable.
>I think this is (1) not how we individuate mental content in everyday
>practice;
Perhaps you are right. Perhaps you are a brain in a vat, and your
brain's thoughts are exclusively about virtual water. You cannot
prove otherwise. You can present good arguments that it is
implausible, and I will agree to that implausibility. But you
nevertheless cannot prove otherwise.
> (2) not even how cognitive science determines informational
>content (see Tyler Burge, "Individualism and Psychology" for a defense
>of this claim);
I don't see that as important. We need to determine informational
content in a way that will allow science to make progress. Cognitive
science has made very little progress, so perhaps the way it has been
determining informational content is part of the problem.
> (3) dooms you to solipsism since
>reference to the extramental world is not achieved.
I don't see that at all. If anything, it seems to be the other way
around. You want to pin informational content to that which is
unknowable, and I think that dooms you to solipsism. I want to pin
informational content to the correlations I can find in all of the
information streams I can access.
> It is also a bit
>mysterious how communication over a common world of objects is
>possible, so science as an intersubjective process also becomes
>mysterious.
I don't see it at all mysterious. It is only mysterious if one ties
information content to that which is unknowable. If I tie
information content to correlations that I find, then I should be
able to communicate with anyone who discovers a sufficient number of
the same correlations.
>You seem to be thinking of science as basically a kind of measurement
>process. I am wondering if this abets your apparent individualism --
>measurement is a process which might be effected by a module in an
>individual brain alone , functioning as an instrument.
That is grossly simplistic. Measurements are only important because
they are useful to us in our causal interactions. I'm not at all
individualistic, in any important sense. Nevertheless, a complete
environment consists of interacting individuals, and so an account of
individuals must be given. To say otherwise is to say that science
should be abandoned.
> But this is a
>quite bizarre to me. I think of scientific measurement as something
>that takes place in the world with objects, that requires an awful lot
>of symbolic mediation which only training in language and mathematics
>can provide.
You are echoing Kuhn's view, which tends to trivialize the role of
measurement. I think that is based on a serious misunderstanding of
the processes of science.
>With all this context in place, we can speak of scientific discoveries
>as akin to measurement, but a spectrometer can't do science, and the
>brain of a scientist is not just a measuring device which arrives at
>knowledge by simple detection.
A spectrometer can't measure either. Measurement requires someone
using the spectrometer in the right way, and knowing when a reading
can be considered a measurement.
>>that all of the astronomical evidence we have could be produced by a
>>cloud of angels shining flashlights (with some radio beacons thrown
>>in). Let's suppose that I were to discover that in fact that is the
>>source of light from the evening sky. Perhaps I learned this via a
>>direct pipeline to god.
>>I claim that this changes nothing. I should still say that the
>>heavens are full of stars, nebulae, galaxies, supernovae, etc.
>>Describing the heavens in that way allows me to predict future
>>observations. It allows me to use the light to align my telescope,
>>or for terrestrial navigation. However, if instead I describe the
>>heavens in terms of the angels, then I have no clear basis for
>>prediction, telescope alignment, or terrestrial navigation. Thus I
>>claim the information is of stars, galaxies, nebulae, etc, and
>>whether the light is actually produced by a cloud of angels is
>>irrelevant.
>You can't have it both ways -- that is you can't have us grant by
>hypothesis that you know your theory is false and then go on to say
>it changes nothing or disproves the correspondence theory. The case just
>shows that a false theory might be useful, as the flat earth theory is.
But I didn't say that I know my theory is false. As Hume might
argue, just because the light came from angels yesterday, that does
not provide any basis in reason to conclude that the light comes from
angels today. Surely if I announced that it did come from angels,
what I said would be denounced as false by most astronomers, and I
would be treated as a deluded nut.
We should take a Rylean view, whereby it is "knowing how" that
counts, rather than "knowing that." If so, then my description in
terms of galaxies and nebulae is a conceptual structure that I use to
account for currently received information, and to allow me to use
that information effectively. Under such a view of knowledge, my
standard of truth or falsity should be based on whether I am properly
using the conceptual structure. An unknowable "knowing that" should
play no role whatsoever.
>I reject the assumption that basic knowledge of the world is mediated by
>interpretation of cues or proximal sense data so I don't think I
>have any problem with unknowability. That only applies when one must
>make an inference from some datum to a theory, not in consciousness of
>the world which is non-inferential, detection of information itself.
I have a battery, and I wish to find its voltage. Here are two
methods I could use:
(a) I connect the battery through a resister to a coil placed in
a magnetic field, and with an attached spring. Then I
measure the position of the coil. Next I use Hooke's law, to
determine what must be the force on the coil so that it is in
its current configuration. Then I use this force to infer
what must be the electrical current flowing through the coil
(Ampere's law, perhaps). Next I apply Ohm's law, and the
known resistance, to infer what must be the voltage of the
battery.
(b) I simply use a voltmeter, and read off the voltage. That
amounts to a direct non-inferential detection of the
information itself. And I need not have heard of Hooke,
Ampere, or Ohm to use this method.
But surely this is just a confusion. Method (a) is simply a
description of the operating principles used in method (b). When I
use method (b), I am using a highly inferential process. It is
simply a matter that the inference is implicit, built into the
structure of the voltmeter, and thus non-conscious.
It seems to me that there is a lot of confusion in your discussion of
direct versus inferential. You seem to be using the evidence that
there is no explicit inference, and somehow concluding that there is
also no implicit inference.
>It is not so when I am conscious of a desk. On a Gibsonian view there
>is no more basic data from which a desk is inferred.
But surely that is because Gibson is concerned with explicit
inference, rather than implicit inference. Just as I can use the
voltmeter to make implicit inferences, even if I have never heard of
Hooke's law or Ohm's law, so also my perceptual processes can be
making implicit inferences based on inference principles that are not
available to my conscious awareness.
> The process which
>constructed the transducer which makes this non- inferential knowledge
>possible is not itself an *epistemic* link in my present justification
>for belief in the desk.
Then we should say that an infant is born with enormous knowledge,
which we try to remove in our educational processes. After all, much
of our education is designed to allow us to make direct decisions
where the inference is all implicit and by your accounts non
epistemic. I think you have fallen for what Ryle referred to as the
intellectualist legend.
Anders N Weinstein
In article <01bb8ffd$806633a0$50093ccf@epfaith>,
Ed Faith <epf...@tiac.net> wrote:
|New devices allow us to deal with individual atoms in a way that
|begins to approach the way we deal with ordinary objects.
|Researchers can spell things like "IBM" with a figurative
|handful of atoms. Here (conscious) inference isn't going
|on, i.e., in the sense of "the spectrum is like so, therefore
|the atom must be like so". If atoms nevertheless remain
|theoretical entities despite this "direct" interaction (achieved
|through mechanical prostheses), then I don't see how you
|can deny that tables are theoretical entities.
I could say instead that atoms have ceased to be theoretical and
become observable entities like tables. Perhaps theoreticity is
only provisional.
Anders N Weinstein
In article <4vi251$m...@ux.cs.niu.edu>, Neil Rickert <ric...@cs.niu.edu> wrote:
>
>>With all this context in place, we can speak of scientific discoveries
>>as akin to measurement, but a spectrometer can't do science, and the
>>brain of a scientist is not just a measuring device which arrives at
>>knowledge by simple detection.
>
>A spectrometer can't measure either. Measurement requires someone
>using the spectrometer in the right way, and knowing when a reading
>can be considered a measurement.
Which is something like what I meant.
>I have a battery, and I wish to find its voltage. Here are two
>methods I could use:
>
> (a) I connect the battery through a resister to a coil placed in
> a magnetic field, and with an attached spring. Then I
> measure the position of the coil. Next I use Hooke's law, to
> determine what must be the force on the coil so that it is in
> its current configuration. Then I use this force to infer
> what must be the electrical current flowing through the coil
> (Ampere's law, perhaps). Next I apply Ohm's law, and the
> known resistance, to infer what must be the voltage of the
> battery.
>
> (b) I simply use a voltmeter, and read off the voltage. That
> amounts to a direct non-inferential detection of the
> information itself. And I need not have heard of Hooke,
> Ampere, or Ohm to use this method.
>
>But surely this is just a confusion. Method (a) is simply a
>description of the operating principles used in method (b). When I
>use method (b), I am using a highly inferential process. It is
Only if you yourself are doing that inference, I would have thought.
>simply a matter that the inference is implicit, built into the
>structure of the voltmeter, and thus non-conscious.
I don't see how what happens inside a voltmeter could be an unconscious
inference of *mine*.
You seem to confusing the standing of an epistemic agent with what
happens inside a voltmeter. If I reason through (a) the status of my
knowledge is inferential. If I do (b) it might or might not be,
depending on how automatic the process is. But in case (b) the
inferences surely do not go on inside the meter.
>It seems to me that there is a lot of confusion in your discussion of
>direct versus inferential. You seem to be using the evidence that
>there is no explicit inference, and somehow concluding that there is
>also no implicit inference.
Yes. I don't believe in unconscious inference. "Implicit" might
have some sense, but one that refers more to the rationale which explains
why something works or is effective than to any cognitive transitions
in the epistemic subject.
For example, consider the idea that the system that enables us to see
moving bodies system exploits the environmental fact that most
translations of bodies in space are rigid. We might even say it "makes
that assumption", and find it can be lead into error in circumstances
in which it does not hold. But that assumption is not itself a premise
which gets invoked as a step in a deductive or inductive argument when
one sees a moving object, not even unconsciously. It is rather part of
the analysis of the conditions under which non-inferential motion
detection is possible.
>>It is not so when I am conscious of a desk. On a Gibsonian view there
>>is no more basic data from which a desk is inferred.
>
>But surely that is because Gibson is concerned with explicit
>inference, rather than implicit inference. Just as I can use the
>voltmeter to make implicit inferences, even if I have never heard of
>Hooke's law or Ohm's law, so also my perceptual processes can be
>making implicit inferences based on inference principles that are not
>available to my conscious awareness.
I don't know what you mean by implicit inference, I don't think there
is any such thing. Certainly your perceptual processes are not
unconsciously applying Ohm's law when you look at a voltmeter.
I find your analogy baffling. Consider someone who doesn't know Ohm's
law, or any of the rest. In that case Ohm's law and the rest have no role
of any kind in their psychology.
>> The process which
>>constructed the transducer which makes this non- inferential knowledge
>>possible is not itself an *epistemic* link in my present justification
>>for belief in the desk.
>
>Then we should say that an infant is born with enormous knowledge,
>which we try to remove in our educational processes. After all, much
>of our education is designed to allow us to make direct decisions
>where the inference is all implicit and by your accounts non
>epistemic. I think you have fallen for what Ryle referred to as the
>intellectualist legend.
I also don't understand this. Finding unconscious or implicit inferences
where none occur is precisely the intellectualist legend. You seem to be
confusing an analysis of why a non-inferential transition works or is
effective with the actual performance of an inference.
Neil Rickert
One could say that the inference is partially done by the designers
and manufacturers of the voltmeter. But I don't think you can escape
from saying that you are making an implicit inference. It differs
from the explicit inference of (a), mainly in that you have delegated
responsibility for part of the inferential process to those who
designed, manufactured, and calibrated the voltmeter. But I think it
is still properly your inference, because it was your choice to
delegate responsibility.
>I don't know what you mean by implicit inference, I don't think there
>is any such thing. Certainly your perceptual processes are not
>unconsciously applying Ohm's law when you look at a voltmeter.
No, they are not unconsciously applying Ohm's law. They are
unconsciously delegating the application of Ohm's law to the
decisions of those who built the voltmeter.
>I find your analogy baffling. Consider someone who doesn't know Ohm's
>law, or any of the rest. In that case Ohm's law and the rest have no role
>of any kind in their psychology.
Why should it matter that they have no role in psychology. If we
consider knowledge as having to do with abilities, then whether
psychology is involved in the application of those abilities seems
unimportant.
>>> The process which
>>>constructed the transducer which makes this non- inferential knowledge
>>>possible is not itself an *epistemic* link in my present justification
>>>for belief in the desk.
>>Then we should say that an infant is born with enormous knowledge,
>>which we try to remove in our educational processes. After all, much
>>of our education is designed to allow us to make direct decisions
>>where the inference is all implicit and by your accounts non
>>epistemic. I think you have fallen for what Ryle referred to as the
>>intellectualist legend.
>I also don't understand this. Finding unconscious or implicit inferences
>where none occur is precisely the intellectualist legend.
As I interpret Ryle, the intellectuallist legend is the insistence on
their being a psychological component to an ability before it can be
counted as knowledge. It seems to me that you are doing just that.
H. M. Hubey
ric...@cs.niu.edu (Neil Rickert) writes:
>I didn't say that we cannot precisely define quantities. I said that
>we cannot precisely define length or force. We can define the
>measuring conventions by means of which we quantitatively determine
>length. But we cannot precisely define what it actually is that we
>measure. We can define the quantities, but not the concepts which
>those quantities refer to.
So there is much to be said for Bridgeman's concept of
"operationalism" i.e. we can "define" (tricky word)
things by giving the operations which one must perform
to obtain the observations. Then there is something to be
said for trying to dissolve at least those arguments which
seem to rage on almost because of problems with definitions
by agreeing to use certain names for certain things which we
obtain as a result of certain prescribed operations.
IOW, if we all agree that certain tests (the instrument)
defines something we can (human) intelligence, at least then
we can stop arguing about those aspects of intelligence which
are due to people using different ideas, and discuss why
such a test is or is not appropriate by specifically picking
on test questions.
Similar views hold (to me, at least) on discussion of
consciousness).
>Prove me wrong by precisely defining the concept of length.
Not me :-) maybe somebody else.
--
Regards, Mark
http://www.csam.montclair.edu/~hubey hu...@pegasus.montclair.edu
H. M. Hubey
ande...@pitt.edu (Anders N Weinstein) writes:
>Yes. I don't believe in unconscious inference. "Implicit" might
>have some sense, but one that refers more to the rationale which explains
>why something works or is effective than to any cognitive transitions
>in the epistemic subject.
Didn't we go over this at least once?
If you step outside these days do you take your overcoat with you?
NO. Do you take your overcoat during winter?
Do you think about not taking your overcoat in the morning in the
summer like today or yesterday?
This is just a simple example. If you were ice skating you would
surely be thinking about if what is under your feet is strong
enough to hold you up. If you were walking on ice/snow during
winter you would pay consciencious attention to where you put
your foot, and watch your gait. Do you consciously or conscientiously
think about this kind of stuff in August in Pittsburgh, say, when
you are walking to your car in an asphalt parking lot?
I bet you don't. How did you, then, make the inference that what
is under your foot is safe and not slippery ice/snow?
CDJ
>One could say that the inference is partially done by the designers
>and manufacturers of the voltmeter. But I don't think you can escape
>from saying that you are making an implicit inference. It differs
>from the explicit inference of (a), mainly in that you have delegated
>responsibility for part of the inferential process to those who
>designed, manufactured, and calibrated the voltmeter. But I think it
>is still properly your inference, because it was your choice to
>delegate responsibility.
>No, they are not unconsciously applying Ohm's law. They are
>unconsciously delegating the application of Ohm's law to the
>decisions of those who built the voltmeter.
>
It doesn't strike me as implausible that there be _three_ ways in which
justification can come off. The first two, direct observation and inference,
have already been mentioned. The third would then be, _authority_, as noticed
above. This last is clearly (1) a manner in which things may be justified, (2)
not observational (in the relevant sense), and (3) not inferential (again in
the relevant sense - i.e., appeal to authority is no more inferential that
appeal to observation).
Unless anyone can provide a motivation for trying to cram authority into the
observation/inference divide?
Unfortunately, the idea isn't original to me...
CDJ
Peter_...@uk.ibm.com
In <4vi24a$3...@usenet.srv.cis.pitt.edu>, ande...@pitt.edu (Anders N Weinstein) writes:
>In article <617497...@luptonpj.demon.co.uk>,
>Peter Lupton <lup...@luptonpj.demon.co.uk> wrote:
>>>
>>> Don't see it. The question is whether your explanation is hanging
>>> informational content on the brain states and, if so, how (and, I guess,
>>> why). Is this informational content individualistically determined, so
>>> that it is the same on Earth and Twin Earth and for a brain in a vat?
>>
>>As far as I'm concerned, the externalist moves are fine and dandy.
>>They incorporate how things are into what meaning is. There's nothing
>>wrong with that, but one must distinguish such an account from what
>>I, qua individual, can distinguish using my own individualistic resources.
>
>
>>> brain's eye view.
>>
>>Perhaps not. But then I don't know which step was not from that view
>>and since you haven't specified one, I guess you don't either.
>
>>another way, even though parsimony is a local notion of fit, the
>>representational structures created are representational of non-local
>>structures. And we can see *that* they're representational of non-local
>>structure if for no other reason than that they represent the bleeding
>>edge of the brain's sensorimotor boundarary. But not now *as* a
>>bleeding edge, but as a part of a greater whole.
>
>the explanation, given that purely internal parsimony-guidedness is a
>reference-free notion. At some point you think it does, and I'd like
>and Twin-Earth.
The proximal regularities may be *isomorphic* on Earth as Twin-Earth.
But they are not the same, because one is on Earth and one is on
Twin-Earth. In order to claim that this difference is unavailable to
a system or structure involved in proximal regularities, you'd have to
further argue that self-reference is unavailable to such a system or
structure.
Now I guess you believe something along those lines, but this is where
you and I part company. I don't see why one has to accept some sort
of extensional account for such a system or structure. I don't see anything
particularly odd about actual physically existing systems and structures
thereby having access to indexicals such as 'this' or 'I'. Now I agree that
the use of these terms might not be as rich as *our* use of these terms -
but I don't see that a lack of richness amounts to a complete absence.
>>So my question is: what is it about positing an external object, say,
>>that makes such an activity impossible for a parsimony-guided
>>system to achieve? Now I've been thinking about this for some time,
>>and I haven't yet thought of anything. But if you can pick out
>>something firm (that is, something not question-begging) that a
>>parsimony-guided system can't do, I'd be very interested.
>
>Recall Searle on the difference between what a computer can do and
>what it can do *in virtue of being* a computer (qua computer, you might
>say). A prime-factorization program in execution can bring about changes
>in the room temperature near the hardware, but it is not in virtue of
>being a prime-factorization program that it does so.
>
>In this sense there may be nothing that a parsimony-guided system can't
>do -- it might make an omelette, for example. But it is not in virtue of
>being a purely local parsimony guided system that it does this.
>
>I am suggesting that a (local) parsimony-guided system *as such* never
>does posit an distal object, it just looks for proximal patterns.
>Unless like Quine one thinks these are the same thing, which is a kind
>of phenomenalism.
I just don't see that there is any *argument* here. What I would expect to
see is that the task of representing patterns involves:
1. Positing a rich structure (let's not presuppose the rich structure
is non-local)
2. Positing that the bleeding edge that delimits one's sensory boundary
is just a small part of what is represented.
3. Using the local causal structure to form the extended representational
scheme above.
If it is then possible to relate these together, one observes that what
is being represented is an external causal structure of which one is
but a part, and represented as but a part. And that the external causal
structure is realised in the proximal causal structure.
Now I don't see what else is supposed to be required. So far as I can
see the Searle-in-the-Robot-Reply-Room has access to each part of
the above and, were the Searle to *notice* it, the notion of objective
reference is right there.
Now I don't think babies perform these reasoning steps, nor do you.
But I don't see why such reasoning steps *couldn't* be performed
and why they wouldn't give such a reasoner a pretty solid grip
on the external world.
Now you might say: OK, I see that actual systems aren't the same
on Twin Earth as on Earth, despite being isomorphic, and even though the
formal structure of indexicality can be reconstructed. But yet there must
be something else, something which gives the parsimony-guided system
or structure *access to* the indexicality per se. At which point I'd just
observe that the system or structure is not in need of gaining access
to something it is literally constructed of. Nor are there any detectable
differences which *could* amount to this *access to*. So if you are
suggesting there *must* be something missing, but don't ask me what,
I'll respond with 'Bah! Humbug.' and leave it at that.
Perhaps we mean something different by 'objective reference'. What I
feel in need to account for and to incorporate is a certain cluster of
behaviours and thoughts I associate with externalism. So that, *were*
I to discover Twin Earth I *would* reckon they had XYZ, not water which
is H2O. That if I said: 'That bird is fat' and it was a leaf, I'd perhaps have
said something meaningless as a token utterance, rather than meaningful
but false, say. And that were I to give an account of what these subjunctive
conditionals amount to, my response would be along the lines Kripke and
Putnam would give: essences, indexicality and such-like. Now I don't take
'objective reference' to be anything more than such clusterings, which I
take it the systems and structures I'm talking about will tend to gravitate
towards. Now if you mean something *more* than this, I'll *pass* on this
more-ness, because I think I'll find this more-ness to be nothing other than
some sort of stipulation about what *must* be the case.
Cheers,
Pete Lupton
Anders N Weinstein
In article <4vkfg9$17...@zen.hursley.ibm.com>, <Peter_...@uk.ibm.com> wrote:
>In <4vi24a$3...@usenet.srv.cis.pitt.edu>, ande...@pitt.edu (Anders N Weinstein) writes:
>
>The proximal regularities may be *isomorphic* on Earth as Twin-Earth.
>But they are not the same, because one is on Earth and one is on
>Twin-Earth. In order to claim that this difference is unavailable to
>a system or structure involved in proximal regularities, you'd have to
>further argue that self-reference is unavailable to such a system or
>structure.
Do you think, a bit like Searle, that thought reaches the external
world via self-referential token-reflexive modes of presentation?
Something like: "whatever is causing this very experience" or even
"whatever is generally responsible for this pattern in experience".
Something like that?
So that as it were the sense or mode of presentation is the same for
a brain on Earth and Twin-Earth but the reference is different because
of indexicality. That is one approach, although I happen to think it
misrepresents our epistemic relation to the world. You also have to
find covert indexicality all over the place, say in thoughts expressible
using proper names, even though these hidden indexicals don't really
show themselves in practice.
I am also curious if it is part of your view that the idea of causation
itself is innate in the system?
Also if you think a brain in a vat has largely true beliefs if it accurately
predicts the patterns of stimuli.
>I just don't see that there is any *argument* here. What I would expect to
>see is that the task of representing patterns involves:
>
> 1. Positing a rich structure (let's not presuppose the rich structure
> is non-local)
>
> 2. Positing that the bleeding edge that delimits one's sensory boundary
> is just a small part of what is represented.
>
> 3. Using the local causal structure to form the extended representational
> scheme above.
>
>If it is then possible to relate these together, one observes that what
>is being represented is an external causal structure of which one is
>but a part, and represented as but a part. And that the external causal
>structure is realised in the proximal causal structure.
>
>Now I don't see what else is supposed to be required. So far as I can
>see the Searle-in-the-Robot-Reply-Room has access to each part of
>the above and, were the Searle to *notice* it, the notion of objective
>reference is right there.
I don't think on can make a "posit" to account for the bleeding edge
unless you are already a conscious subject that can *see* the bleeding
edge, unless it is an object before your awareness, the way a desk can
be for you. In fact it might be much more accurate to say that it is
only on the basis of initial awareness of the "external" world of
objects one can see and touch that a cognitive subject could go on to
form hypotheses about what you call the "bleeding edge" or the brain
inside his or her skull. (And how one got into this initial position is
not itself a cognitive process of the subject's.)
So of course Searle in the control room of a robot, who, let us
suppose, manipulates representations of low-level retinal input to the
robot, can form the hypothesis of a world outside if his attention
happens to turn to the matter. This is because Searle is a conscious
subject for whom these representations are objects of awareness. That
means, I take it, that Searle is *already* something that has an
understanding of an "external" world of physical objects, has a body, a
language, and the like. With all this in place of course there is no
problem about him consciously forming hypotheses about what's happens to
be beyond the bit of the "external" world he can see.
If we call the Chinese-speaking robot Robbie, the problem is that
*Robbie* cannot do any of this, since Robbie can't cognize these
low-level representations without instruments. Ye Robbie might see a
desk outside the room, even *before* Searle has started to form any
hypotheses about what's outside the room.
So what the conscious subject Searle can do in his thought is totally
irrelevant it seems to me, since Searle's phenomenology is independent
of Robbie's. Searle may or may not form the hypothesis of an external
world to account for the proximal patterns he can see directly, but
he need not ever do so for Robbie to have beliefs about the external
world.
In other words, one must not confuse Searle as person forming a
hypothesis consciously in his mind, with Searle as program
instantiation manipulating symbols. These manipulations might
constitute or make possible Robbie's forming a hypothesis but are
purely meaningless for Searle.
Your theory is that if Robbie's program is a learning one then the
*program* does something called "positing an external causal
structure", even if Searle never bothers to do any such thing. My point
is that even if it does this it never does make Searle as person aware
of the external world -- that is a completely independent issue. So
personifying the program makes clear that an inner symbol manipulator
*as such* is and remains totally blind, even if Searle himself as
person -- quite apart from any formal symbol manipulations he may carry
out -- is independently capable of thinking, seeing, forming
hypotheses about what's beyond the door.
>Perhaps we mean something different by 'objective reference'. What I
>feel in need to account for and to incorporate is a certain cluster of
>behaviours and thoughts I associate with externalism. So that, *were*
>I to discover Twin Earth I *would* reckon they had XYZ, not water which
>is H2O. That if I said: 'That bird is fat' and it was a leaf, I'd perhaps have
>said something meaningless as a token utterance, rather than meaningful
>but false, say. And that were I to give an account of what these subjunctive
>conditionals amount to, my response would be along the lines Kripke and
>Putnam would give: essences, indexicality and such-like. Now I don't take
>'objective reference' to be anything more than such clusterings, which I
>take it the systems and structures I'm talking about will tend to gravitate
>towards. Now if you mean something *more* than this, I'll *pass* on this
>more-ness, because I think I'll find this more-ness to be nothing other than
>some sort of stipulation about what *must* be the case.
No I don't mean much more than that, but all those things require
taking the relevant system to be much larger than a brain, it seems to me.
An embodied organism in a material and social environment, for example.
If you ascribe to a brain the ability to represent "that bird is fat",
I think it must be derivative from the ascription to a person who could
point at a bird. Content devolves on brain states from the top down and
the outside in, it seems to me.
Neil Rickert
In <4vkrm5$9...@news.nyu.edu> mal...@acf2.nyu.edu (Jacques Maurice Mallah) writes:
>Neil Rickert (ric...@cs.niu.edu) wrote:
>: Right. So what is your problem. In front of me is a screen covered
>: with colored pixels. One interpretation is that the screen contains
>: words which you have written. The fact that it is an interpretation
>: does not imply there is anything subjective involved.
> Yes it does. An objective fact is more than just an
>interpretation. Objectives facts in your example are that the screen is
>colored in such a way that, upon seeing it, you would react in a certain
>way that is associated with visually recieving information. On a
>fundamental level, some statement about the microphysical state of the
>universe could, in principle, be made based on the information you gave
>about the screen.
> Any fundamental theory must deal with objective facts only, and
>with a precisely defined model.
You are in danger of defining 'objective fact' so narrowly that there
are no objective facts.
>: >> If computation is undefined except as something some conscious
>: >>person can use to organize data, it clearly can play no explanatory
>: >>role. Nothing in our model of the world would give rise to consciousness
>: >>in the first place.
>: > And you wrote:
>: >"That is not at all clear."
>: > implying that something other than computation could give rise to
>: >consciousness.
>: Actually, it only implied that you wrote something that was not at
>: all clear. It does not obviously follow from "computation is
>: undefined" that "nothing in our model of the world would give rise to
>: consciousness."
> With computation undefined, it certainly could not be the thing
>in our model doing the job.
I don't follow the reasoning.
Should I take it that you are claiming that heat does not reduce to
molecular kinetic energy? I raise this question, because there was a
time when people had a reasonably good idea of what they meant by
'heat', but had no definition of molecular kinetic energy. If your
reasoning is valid, it should have been that molecular kinetic energy
could not explain heat. But the advance of science has shown
otherwise.
> Note that I said in our model.
But we don't yet have a good model that encompasses consciousness.
So what model are you talking about?
> A better statement might be "with computation undefinable ...";
>then nothing could do the job.
If consciousness is also undefinable, then there would be no obvious
contradiction in consciousness reducing to computation, and
computation being undefinable.
>: > On what grounds do you say we can't precisely define physical
>: >quantities?
>: I didn't say that we cannot precisely define quantities. I said that
>: we cannot precisely define length or force. We can define the
>: measuring conventions by means of which we quantitatively determine
>: length. But we cannot precisely define what it actually is that we
>: measure. We can define the quantities, but not the concepts which
>: those quantities refer to.
>: Prove me wrong by precisely defining the concept of length.
> How about "that which is measured by those proceedures"?
Fair enough. Now to complete the definition, you need to describe
the procedures in a way which does not make use of the concept of
length.
> There is another way. This can easily be done within a mathematical
>model of physics. That is all that's required.
No, sorry, that is not what is required. You might produce a
marvellous mathematical model of physics. Then I might look at it,
and think it a marvellous model of abstract art. You haven't defined
length with your mathematical model until you have connected the
terms in the model to their real world counterparts. And you cannot
do that without going outside your mathematical model.
> If I say that in such
>and such a model, I will call certain quantities lengths, that defines
>it.
And if you call the same quantities cookies, I suppose you will
instead have defined cookies and they will automatically taste good.
Sorry, but just naming something `length' in a mathematical model
does nothing to define length.
> Given a mathematical model of physics, all that I require is an
>appropriate formula for what computations are present, which is
>consistent with the properties associated with computation.
Then you have a nice abstract model, which might have nothing to do
with the real world.
> There is
>less freedom here because computation has the property that I intuitively
>associate certain classes of computations with consciousness; and as
>defined in the model, that property must be retained.
If you need to rely on intuition, then you are not providing a
definition.
> Knowing exactly which number pi is, which we do, would be no
>different from knowing exactly which computations are present, which we
>don't and is what I want to do. Knowing exactly which number pi is, is
>in fact, what I would call a precise definition of pi. (BTW, we don't
>know which number it is, but we know a proceedure by which we could, in
>principle, approximate it to any desired degree of accuracy; that's all
>we need.)
When two mathematicians talk about pi, them mean the same number.
But when they try to assign a computation to some natural process,
they very well may not agree on what computation is appropriate.
Hilary Putnam discusses some of the problems with deciding what
computation is occurring in an appendix to his book "Representations
and Reality." You might find it useful reading.
> What I have been saying all along is that nobody has a definition
>that can make the required distinction, and that I want one to be found,
>and if it can't be, computation is an unsuitable concept for explaining
>consciousness.
Perhaps that might show that computation is not sufficient. I don't
see why it would show that computation is unsuitable.
>: Well, I never suggested that non-computable mathematics would play
>: any role. The concept of `non-computable' derives from Turing
>: machines. I rejected the centrality of Turing machines, so it is
>: strange that you would accuse me of suggesting a role for
>: non-computable mathematics.
> By rejecting Turing machines, you put non-computable and
>computable math on the same footing.
Not at all. A Turing machine is, by definition, a formal
abstraction. By rejecting Turing machines, I am simply arguing that
abstractions are not conscious.
Jacques Maurice Mallah
: >intentionality is an intrinsic property of reality, but computation
: >is only an interpretation (in his "Rediscovery of mind"). You want
: >to disagree with Searle as to whether computation is intrinsic or an
: >interpretation. I prefer to disagree with Searle as to whether
: >intentionality is intrinsic or an interpretation."
:
: > Which clearly implies that you agree with him that computation is
: >only an interpretation.
:
: Right. So what is your problem. In front of me is a screen covered
: with colored pixels. One interpretation is that the screen contains
: words which you have written. The fact that it is an interpretation
: does not imply there is anything subjective involved.
Yes it does. An objective fact is more than just an
interpretation. Objectives facts in your example are that the screen is
colored in such a way that, upon seeing it, you would react in a certain
way that is associated with visually recieving information. On a
fundamental level, some statement about the microphysical state of the
universe could, in principle, be made based on the information you gave
about the screen.
Any fundamental theory must deal with objective facts only, and
with a precisely defined model.
: >> If computation is undefined except as something some conscious
: >>person can use to organize data, it clearly can play no explanatory
: >>role. Nothing in our model of the world would give rise to consciousness
: >>in the first place.
:
: > And you wrote:
:
: >"That is not at all clear."
:
: > implying that something other than computation could give rise to
: >consciousness.
:
: Actually, it only implied that you wrote something that was not at
: all clear. It does not obviously follow from "computation is
: undefined" that "nothing in our model of the world would give rise to
: consciousness."
With computation undefined, it certainly could not be the thing
in our model doing the job. Note that I said in our model. There could
be some mathematical structure that suffices in the real case, but unless
we find out what that could be and define computation to correspond to
that, it would not be in our model.
A better statement might be "with computation undefinable ...";
then nothing could do the job.
: It seems likely enough to me that computation can be used in ways
: that give rise to consciousness. And it seems clear to me that
: computation exists in our model of the world, even if it is
: undefined.
:
: >: > Then you would agree with him, and me, that an objective
: >: >criterion for computation is needed! That seems out of step with your
: >: >previous post. Maybe you should read your own post.
:
: >: On the contrary, I think we have a pretty good idea of what
: >: computation is, and there is no need to precisely define it. We
: >: cannot precisely define length or force, but we still manage to reach
: >: an agreeable consensus. Why should we need to precisely define
: >: computation?
:
: > On what grounds do you say we can't precisely define physical
: >quantities?
:
: I didn't say that we cannot precisely define quantities. I said that
: we cannot precisely define length or force. We can define the
: measuring conventions by means of which we quantitatively determine
: length. But we cannot precisely define what it actually is that we
: measure. We can define the quantities, but not the concepts which
: those quantities refer to.
:
: Prove me wrong by precisely defining the concept of length.
How about "that which is measured by those proceedures"?
There is another way. This can easily be done within a mathematical
model of physics. That is all that's required. If I say that in such
and such a model, I will call certain quantities lengths, that defines
it. Now, the reason I call it length, is because it has the properties I
associate with the word.
Given a mathematical model of physics, all that I require is an
appropriate formula for what computations are present, which is
consistent with the properties associated with computation. There is
less freedom here because computation has the property that I intuitively
associate certain classes of computations with consciousness; and as
defined in the model, that property must be retained.
Since this "computation" is to be associated with consciousness,
then if the model is to be a plausible candidate for the correct model,
it must be distributed in such a way that my consciousness is sufficently
typical.
: > Even if we couldn't, it would be an irrellevant example.
: >Computation is mathematical, not physical; are you saying that pi is not
: >precisely defined?
:
: No, not at all. But I am saying that numbers are not precisely
: defined. We know exactly which number pi is, but we cannot agree on
: what is a number. Is a number something that exists in plato's
: heaven and has been discovered by mathematicians? Is a number the
: class of all sets with the same cardinality? Is a number a set? Or
: is it that we merely have names for numbers, but the numbers
: themselves do not exist? There is no agreement about what numbers
: actually are.
Knowing exactly which number pi is, which we do, would be no
different from knowing exactly which computations are present, which we
don't and is what I want to do. Knowing exactly which number pi is, is
in fact, what I would call a precise definition of pi. (BTW, we don't
know which number it is, but we know a proceedure by which we could, in
principle, approximate it to any desired degree of accuracy; that's all
we need.)
So has a bit of sematics been what has been preventing you from
seeing the importance of what I call a precise definition of computation?
: > At the present precision of the definition, we have no grounds for
: >saying that a toaster has less consciousness than does a human or would an
: >advanced robot. Does that satisfy you?
:
: I think it quite obvious that a toaster does not have the same
: consciousness as a toaster. Even the operational definition used by
: a medical practitioner is enough to determine that. If your
: definition of `consciousness' cannot distinguish, then you should
: stop pretending that you have a definition.
What I have been saying all along is that nobody has a definition
that can make the required distinction, and that I want one to be found,
and if it can't be, computation is an unsuitable concept for explaining
consciousness.
An operational definition can't be used to define computation
because then, in any given mathematical model of physics, who would do
the operating? Why, computers, of course. And those were defined in
terms of the same definition... it would just be an infinite regression
that leads nowhere.
: >: > Since my notions of what is needed for computations are not so
: >: >different from Chalmers, just what notions of mine were you calling
: >: >restrictive?
:
: >: Your notions are not sufficiently clear to me. Some people identify
: >: computation with that which can be exactly modelled on a Turing
: >: machine. I think such a view of computation is too restrictive, and
: >: leads to the type of confusion shown by Searle, Penrose and others.
: >: I don't expect the theory of Turing machines, or of recursive
: >: functions, to be very important in the question finding a
: >: computational basis for human cognition.
:
: > I don't see why noncomputable math would play any role.
:
: Well, I never suggested that non-computable mathematics would play
: any role. The concept of `non-computable' derives from Turing
: machines. I rejected the centrality of Turing machines, so it is
: strange that you would accuse me of suggesting a role for
: non-computable mathematics.
By rejecting Turing machines, you put non-computable and
computable math on the same footing. That in fact seems to be the only
likely motive I see for rejecting Turing machines, which are partial to
computable math. With no role for noncomputable math, you are in fact in
the realm of Turing machines.
: >talk artificial intelligence. Could a digital computer be conscious?
:
: If, by a digital computer, you mean the type of appliance sitting on
: my desk, then I don't believe one could be conscious in any useful
: sense of the word. It is disconnected from reality, so there is
: nothing for it to be conscious of. If, on the other hand, you are
: asking whether a conscious robot could be constructed using digital
: technology, then I see no principled reasons why not, although there
: might be many practical difficulties.
I'll take that as a yes, though you gratuitously threw in an
improbable assumption. Conscious is internal; and in any case, external
reality could be replaced by another program.
: > What kind of computation do you have in mind?
:
: I listen to music from a CD player. In my book, it operates using
: computation. I cannot adequately describe it with a Turing machine
: model. So it uses computation in a way that is not captured by that
: model. The Turing machine model only considers what is left on the
: tape after a HALT state is entered. What is important for the CD
: player, is the stream of information being exchanged before a HALT
: state is entered. More generally, the Turing machine model does not
: adequately account for real time computing.
No, computationalists always say that it is the operation of the
program as a function of some time-style parameter that is significant, not
just the initial and final states, and I never said anything different.
Turing machines, FSAs, CSAs, etc. operate over time.
: I didn't say that computation is an unsuitable concept. I said that
: consciousness is an unsuitable concept. By that, I intended that it
: is unsuitable for the type of problems investigated in AI.
The fundamental problem is the problem of consciousness.
It is true that those who actually work with AI have no way to experience
whatever the computer may be experiencing, which is why their work,
grounded in what is measured, does not address the fundamental problem,
but only provides an area in which any fundamental results could be
applied to guide our relationships with such machines, should a conscious
machine ever be built.
Anders N Weinstein
In article <4vkrm5$9...@news.nyu.edu>,
|: >only an interpretation.
|:
|: Right. So what is your problem. In front of me is a screen covered
|: with colored pixels. One interpretation is that the screen contains
|: words which you have written. The fact that it is an interpretation
|: does not imply there is anything subjective involved.
|
| Yes it does. An objective fact is more than just an
|interpretation. Objectives facts in your example are that the screen is
|colored in such a way that, upon seeing it, you would react in a certain
|way that is associated with visually recieving information. On a
|fundamental level, some statement about the microphysical state of the
|universe could, in principle, be made based on the information you gave
|about the screen.
|
| Any fundamental theory must deal with objective facts only, and
|with a precisely defined model.
I don't get either side here. It seems to be a perfectly objective fact
whether the screen contains words which someone has written, albeit a
fact which could not be determined solely by the present state of the
screen. If the pixels are arrayed at random, then it is an objective
that it does not contain words; if there are practices of using CRTs as
media for the transmission and presentation of words, it might or might
not.
It makes sense to say one can interpret the screen as containing words
but would be mistaken to do so. So it is not a matter of interpretation
in the sense that I am free to view it in whatever way I like, or even
in whatever way I happen to find useful.
On the other hand I don't see any need for "precisely defined models" or
"fundamental theories" in this domain. Objectivity has nothing
to do with that, merely with being constrained, not free to say whatever
one likes and get away with it.
I think a lot of careful attention would be needed to make clear what
one means when one says something is "only an interpretation".
> What I have been saying all along is that nobody has a definition
>that can make the required distinction, and that I want one to be found,
>and if it can't be, computation is an unsuitable concept for explaining
>consciousness.
I think I am with Rickert in thinking this is a fallacy. Biologists
make do without a precise criterion of "life", after all.
We need to distinguish the question of objectivity from the question of
precise definability. I think it is objective that this thing on my
desk is running a certain program and not others, but one that depends
on the human institutions that manufactured it and wrote the software, not
solely on its local physical properties. And these determinants cannot
be precisely defined.
Neil Rickert
In <hubey.8...@pegasus.montclair.edu> hu...@pegasus.montclair.edu (H. M. Hubey) writes:
>ric...@cs.niu.edu (Neil Rickert) writes:
>>I didn't say that we cannot precisely define quantities. I said that
>>we cannot precisely define length or force. We can define the
>>measuring conventions by means of which we quantitatively determine
>>length. But we cannot precisely define what it actually is that we
>>measure. We can define the quantities, but not the concepts which
>>those quantities refer to.
>So there is much to be said for Bridgeman's concept of
>"operationalism" i.e. we can "define" (tricky word)
>things by giving the operations which one must perform
>to obtain the observations. Then there is something to be
>said for trying to dissolve at least those arguments which
>seem to rage on almost because of problems with definitions
>by agreeing to use certain names for certain things which we
>obtain as a result of certain prescribed operations.
I think there is a lot to be said in favor of operationalism. What
is wrong, is the assumption that operationalism by itself could be a
complete theory.
Peter Lupton
In article: <4vkmhv$d...@usenet.srv.cis.pitt.edu> ande...@pitt.edu (Anders N Weinstein) writes:
>
> In article <4vkfg9$17...@zen.hursley.ibm.com>, <Peter_...@uk.ibm.com> wrote:
> >In <4vi24a$3...@usenet.srv.cis.pitt.edu>, ande...@pitt.edu (Anders N Weinstein) writes:
> >
> >The proximal regularities may be *isomorphic* on Earth as Twin-Earth.
> >But they are not the same, because one is on Earth and one is on
> >Twin-Earth. In order to claim that this difference is unavailable to
> >a system or structure involved in proximal regularities, you'd have to
> >further argue that self-reference is unavailable to such a system or
> >structure.
>
> Do you think, a bit like Searle, that thought reaches the external
> world via self-referential token-reflexive modes of presentation?
> Something like: "whatever is causing this very experience" or even
> "whatever is generally responsible for this pattern in experience".
> Something like that?
>
> So that as it were the sense or mode of presentation is the same for
> a brain on Earth and Twin-Earth but the reference is different because
> of indexicality. That is one approach, although I happen to think it
> misrepresents our epistemic relation to the world. You also have to
> find covert indexicality all over the place, say in thoughts expressible
> using proper names, even though these hidden indexicals don't really
> show themselves in practice.
The problem is that the whole structure of externalism only *shows* itself
in practice when push comes to shove and some actual event occurs, internally,
which reveals the difference.
I do think that what you seem to call 'objective reference' is brought about
by the fact that we are actual mechanisms and that much of our language is
'from *here* on out', as it were. This strikes me as the most natural form
of reference - forms which aren't 'here on out' are, in comparison, quite
sophisticated. Do we know how to identify a place on Earth without
exploiting some indexical? That is, that we can identify our own Galaxy/
Sun/Planet? And what if there isn't one unique universe as some think? Our
language is learned, as it were, from the local to the increasingly wide-ranging.
I don't think my use of indexicality will match Searle's because Searle is,
after all, producing a specific theory for perception as a whole. I'm not,
and I'll leave many questions unanswered because parsimony-guided systems
won't tend to fit into neat, finitely presented formulations. So I'll not
try to force it into such a mould.
That indexicals don't show themselves in practice is not really true, is it?
Were I to talk about London, then I have no idea whether there is, anywhere
in our universe, another town called London. But I'm positive I don't mean
such other towns, whether or not there are any. So my use of the word London
is within a locality, picked out in part by my being here. So it seems that my
ability as a pattern-matcher (available through parsiony-guidedness) always
comes back to me as the source of that pattern-matching. Even when I use
words deferentially, what I defer to is picked out in a similar way.
> I am also curious if it is part of your view that the idea of causation
> itself is innate in the system?
The problem of how causation as a concept is formed is certainly interesting.
My guess is that we can learn such a concept.
> Also if you think a brain in a vat has largely true beliefs if it accurately
> predicts the patterns of stimuli.
>
> >I just don't see that there is any *argument* here. What I would expect to
> >see is that the task of representing patterns involves:
> >
> > 1. Positing a rich structure (let's not presuppose the rich structure
> > is non-local)
> >
> > 2. Positing that the bleeding edge that delimits one's sensory boundary
> > is just a small part of what is represented.
> >
> > 3. Using the local causal structure to form the extended representational
> > scheme above.
> >
> >If it is then possible to relate these together, one observes that what
> >is being represented is an external causal structure of which one is
> >but a part, and represented as but a part. And that the external causal
> >structure is realised in the proximal causal structure.
> >
> >Now I don't see what else is supposed to be required. So far as I can
> >see the Searle-in-the-Robot-Reply-Room has access to each part of
> >the above and, were the Searle to *notice* it, the notion of objective
> >reference is right there.
>
> I don't think on can make a "posit" to account for the bleeding edge
> unless you are already a conscious subject that can *see* the bleeding
> edge, unless it is an object before your awareness, the way a desk can
> be for you.
Well, the term 'bleeding edge' was introduced by me to dramatise the
situation that a parsimony-guided program can't be content with the
sensory boundary and must form representational structures which go
beyond that edge. Probably the dramatisation is going too far.
What I will insist on, however, is that the representational structures
created in parsimony will, necessarily, (for our sort of world) include
far more than what falls within the boundary of the organism per se.
Now if you are suggesting that the organism can't distinguish those
parts of its own representational states which are self-representational
from those which aren't, I'd have to know why. And if the answer is
no more than the *assertion* that the subject must be able to *see* it,
then I'll be extraordinarily unimpressed.
One of the ways, for example, that parsimony would represent things
is as a 3-d manifold. But not just a 3-d manifold, because visual
sensory data are special - they are plainly (and parsimony will
produce this) associated with perspective mappings. Which means that
the 3-d representation is incomplete: it needs a distinguished point
and orientation to define that very perspective point. And that
perspective point and orientation moves with motor responses etc.
Now this is just one of the many roles for 'I' that comes in the
representational process. When I say that the bleeding edge is represented,
what I mean is just that there are a whole range of such cases, all of which
converge on the same represented sub-structure - each of them versions
of 'I'.
> ....In fact it might be much more accurate to say that it is
> only on the basis of initial awareness of the "external" world of
> objects one can see and touch that a cognitive subject could go on to
> form hypotheses about what you call the "bleeding edge" or the brain
> inside his or her skull. (And how one got into this initial position is
> not itself a cognitive process of the subject's.)
I don't think that it could be the case that the organism must *actually*
see and touch, since I don't see that I could tell the difference between
what's happening to me now and what would happen to me if the universe ended
at my skin but the sensory neurons fired just the same.
Now you could say at this point that the two explanations would just be totally
different. But I don't really think that'd wash. It would turn out to be
the *same* explanation in both cases except that you'd insist in one case
in using the term 'aware' and in the other case you'd insist on using the
term 'seems to be aware' and such-like. That apart, the explanations would be
identical.
What's necessary is the system of representation I've been stressing,
together with the observation that such representational structures
must also represent onself as part (and not the whole) of that
representational structure.
But I did think that, even though we agree that a baby does not go through
these cognitive processes as full-blown verbal reasoning processes,
I'm quite happy for there to be far lower-level processes than those
which are full-blown reasoning processes which have the effects I've
indicated above.
> So of course Searle in the control room of a robot, who, let us
> suppose, manipulates representations of low-level retinal input to the
> robot, can form the hypothesis of a world outside if his attention
> happens to turn to the matter. This is because Searle is a conscious
> subject for whom these representations are objects of awareness. That
> means, I take it, that Searle is *already* something that has an
> understanding of an "external" world of physical objects, has a body, a
> language, and the like. With all this in place of course there is no
> problem about him consciously forming hypotheses about what's happens to
> be beyond the bit of the "external" world he can see.
Well, I was dramatising with Searle, also (actually, I thought I was
using Searle's dramatisation.)
> If we call the Chinese-speaking robot Robbie, the problem is that
> *Robbie* cannot do any of this, since Robbie can't cognize these
> low-level representations without instruments. Ye Robbie might see a
> desk outside the room, even *before* Searle has started to form any
> hypotheses about what's outside the room.
Don't see that the whole system couldn't do this *at* *all*. What *is*
true is that whatever processes are involved in Robbie doing this, they
aren't done as full-blown inferential processes. But then I'm of the
opinion that inferential processes (as provided to a parsimony-guided
system) are perfectly doable in a less than full-blown inferential manner.
That's because I don't see a black and white line between inference
done explictly and verbally and inference done implicitly through the
unfolding of parsimony-guidedness.
> So what the conscious subject Searle can do in his thought is totally
> irrelevant it seems to me, since Searle's phenomenology is independent
> of Robbie's. Searle may or may not form the hypothesis of an external
> world to account for the proximal patterns he can see directly, but
> he need not ever do so for Robbie to have beliefs about the external
> world.
Fair enough. Searle is obviously too dramatic for your tastes. But I
thought the problem you raised was how sub-personal structures could
sub-personally lift themselves to the wide world. My use of Searle in
a box was to dramatise the inferential processes as explicit inferential
processes which needn't be done as explicit inferential processes.
> In other words, one must not confuse Searle as person forming a
> hypothesis consciously in his mind, with Searle as program
> instantiation manipulating symbols. These manipulations might
> constitute or make possible Robbie's forming a hypothesis but are
> purely meaningless for Searle.
>
> Your theory is that if Robbie's program is a learning one then the
> *program* does something called "positing an external causal
> structure", even if Searle never bothers to do any such thing. My point
> is that even if it does this it never does make Searle as person aware
> of the external world -- that is a completely independent issue.
True enough. But now I don't know what problem you want solved. I
thought I was addressing myself to a problem you wanted solved, but
obviously not. That's fine, I've just wasted a few hours, that's all.
> So
> personifying the program makes clear that an inner symbol manipulator
> *as such* is and remains totally blind, even if Searle himself as
> person -- quite apart from any formal symbol manipulations he may carry
> out -- is independently capable of thinking, seeing, forming
> hypotheses about what's beyond the door.
I think it shows that my dramatisation didn't impress you.
> >Perhaps we mean something different by 'objective reference'. What I
> >feel in need to account for and to incorporate is a certain cluster of
> >behaviours and thoughts I associate with externalism. So that, *were*
> >I to discover Twin Earth I *would* reckon they had XYZ, not water which
> >is H2O. That if I said: 'That bird is fat' and it was a leaf, I'd perhaps have
> >said something meaningless as a token utterance, rather than meaningful
> >but false, say. And that were I to give an account of what these subjunctive
> >conditionals amount to, my response would be along the lines Kripke and
> >Putnam would give: essences, indexicality and such-like. Now I don't take
> >'objective reference' to be anything more than such clusterings, which I
> >take it the systems and structures I'm talking about will tend to gravitate
> >towards. Now if you mean something *more* than this, I'll *pass* on this
> >more-ness, because I think I'll find this more-ness to be nothing other than
> >some sort of stipulation about what *must* be the case.
>
> No I don't mean much more than that, but all those things require
> taking the relevant system to be much larger than a brain, it seems to me.
> An embodied organism in a material and social environment, for example.
> If you ascribe to a brain the ability to represent "that bird is fat",
> I think it must be derivative from the ascription to a person who could
> point at a bird. Content devolves on brain states from the top down and
> the outside in, it seems to me.
I think all you are saying is that we can be uncertain whether we mean.
Fair enough, but I don't see how that changes the questions one jot or tittle.
If you stipulate that representation is outside in, and there happens to be no
outside, then that just means we might be uncertain whether we represent or
whether we merely seem to represent. And I don't see that our account is really
going to change too much if all it amounts to is that difficulties to do with
whether what we mean is true are replaced by similar difficulties to do with
whether it is true that we mean. I don't feel this is going to turn out to be
of a great deal of significance in accounting for how we come to do what we do.
--
Cheers,
Pete Lupton
CDJ
In article <4vl3f4$e...@usenet.srv.cis.pitt.edu>,
ande...@pitt.edu (Anders N Weinstein) wrote:
>I think I am with Rickert in thinking this is a fallacy. Biologists
>make do without a precise criterion of "life", after all.
A poor choice of example, in light of the (relatively) recent surfacing of
_Archae_? Or did you have that in mind, so that the example is in fact a very
good one?
CDJ
Anders N Weinstein
In article <643507...@luptonpj.demon.co.uk>,
Peter Lupton <lup...@luptonpj.demon.co.uk> wrote:
>In article: <4vkmhv$d...@usenet.srv.cis.pitt.edu> ande...@pitt.edu (Anders N Weinstein) writes:
>>
>I do think that what you seem to call 'objective reference' is brought about
>by the fact that we are actual mechanisms and that much of our language is
>'from *here* on out', as it were. This strikes me as the most natural form
>of reference - forms which aren't 'here on out' are, in comparison, quite
>sophisticated. Do we know how to identify a place on Earth without
>exploiting some indexical? That is, that we can identify our own Galaxy/
>Sun/Planet? And what if there isn't one unique universe as some think? Our
>language is learned, as it were, from the local to the increasingly wide-ranging.
>That indexicals don't show themselves in practice is not really true, is it?
I agree that indexicality is basic, pervasive and ineliminable. I take this
to support externalism and the idea that a cognitive subject must be an
embodied being-in-the-world, not a detached brain, since only such a thing
can *express* an indexical thought.
But Searle uses peculiar *self-referential* indexicals to build an
internalist theory. His idea, for example, is that perceptual states are
*self-referential* and refer to the world only as "the distal cause of
this very experience" -- in that way the sense or mode of presentation
is world-independent although the reference is not. It is only *this use*
of indexicality -- peculiar self-referential ones to support an internalist
theory of intentionality -- that I was opposed to.
>What I will insist on, however, is that the representational structures
>created in parsimony will, necessarily, (for our sort of world) include
>far more than what falls within the boundary of the organism per se.
>Now if you are suggesting that the organism can't distinguish those
>parts of its own representational states which are self-representational
>from those which aren't, I'd have to know why. And if the answer is
>no more than the *assertion* that the subject must be able to *see* it,
>then I'll be extraordinarily unimpressed.
Well I don't know what you mean if you think *I* am "distinguishing"
features of representational states buried in my lateral geniculate nuclei,
say. These are not objects for me.
I can of course distinguish among thoughts, and the like, various
self-referential elements.
>One of the ways, for example, that parsimony would represent things
>is as a 3-d manifold. But not just a 3-d manifold, because visual
>sensory data are special - they are plainly (and parsimony will
>produce this) associated with perspective mappings. Which means that
Not so obvious. In one sense there clearly is a point of view in
visual consciousness. On the other hand if we speak Gibson-wise, we need
not appeal to any representational states with less than three-dimensional
content (G's "layout" of environmental surfaces). In particular, we
need not view the job of the visual system as inverting a perspective
mapping, rather, detecting information about layout and other properties,
including the position of the subject.
>the 3-d representation is incomplete: it needs a distinguished point
>and orientation to define that very perspective point. And that
>perspective point and orientation moves with motor responses etc.
>Now this is just one of the many roles for 'I' that comes in the
>representational process. When I say that the bleeding edge is represented,
>what I mean is just that there are a whole range of such cases, all of which
>converge on the same represented sub-structure - each of them versions
>of 'I'.
I don't deny that the I is represented in all of this. As Gibson emphasized,
patterns in the optic array bear information about the subject at the
same time as about the environment.
>I don't think that it could be the case that the organism must *actually*
>see and touch, since I don't see that I could tell the difference between
>what's happening to me now and what would happen to me if the universe ended
>at my skin but the sensory neurons fired just the same.
>
>Now you could say at this point that the two explanations would just be totally
>different. But I don't really think that'd wash. It would turn out to be
*What* two explanations? In the second case there is nothing to explain
in the representational manner.
>the *same* explanation in both cases except that you'd insist in one case
>in using the term 'aware' and in the other case you'd insist on using the
>term 'seems to be aware' and such-like. That apart, the explanations would be
>identical.
Well this is a standard internalist intuition. But I don't think a
mere brain would even *seem* to be aware. I think *personal* awareness
stands to a brain as economic value to a coin -- in different contexts,
like pieces of metal have different value, and without the right
surroundings, it's just a worthless hunk of metal. Similarly, a brain
on Twin Earth can be part of a person who has one sort of thought, a
homologous brain on Earth can be part of a person with a different
psychology (different thoughts), and a brain that was always in a vat
would be just a brain, and no person would be constituted.
You have to think of the intentional level as globally, not locally
supervenient on the natural scientific description of the body and brain.
You must also remember that even subjective seemings are in the
higher-level world, the world of intentionality, so the content of
seemings switches too as you move from the Earth case to the Twin Earth
case. Things *seem* differently to both people.
>What's necessary is the system of representation I've been stressing,
>together with the observation that such representational structures
>must also represent onself as part (and not the whole) of that
>representational structure.
I agree that this is necessary for full-blown conceptual
representation. (By the way, I get the point from P.F. Strawson's
reading of Kant's Transcendental Deduction in _The Bounds of Sense_)
>Fair enough. Searle is obviously too dramatic for your tastes. But I
>thought the problem you raised was how sub-personal structures could
>sub-personally lift themselves to the wide world. My use of Searle in
>a box was to dramatise the inferential processes as explicit inferential
>processes which needn't be done as explicit inferential processes.
I don't think the problem is just stale drama. I thought, perhaps wrongly,
that you were trading on a confusion of what Searle qua person in his
own right can do (posit an external world outside the room) and what
Searle qua program can do (manipulate meaningless symbols). Your
picture must be that some of the things Searle *qua program* does
constitute "positing an external causal structure" to account for
patterns in sense data, sense data which Robbie does not represent
(nothing in his belief box refers to them). I continue to believe that
Searle qua program is always blind and never does come to cognize or even
posit an external world, while Robbie is a being in the world and does
not have to "posit" it.
>> No I don't mean much more than that, but all those things require
>> taking the relevant system to be much larger than a brain, it seems to me.
>> An embodied organism in a material and social environment, for example.
>> If you ascribe to a brain the ability to represent "that bird is fat",
>> I think it must be derivative from the ascription to a person who could
>> point at a bird. Content devolves on brain states from the top down and
>> the outside in, it seems to me.
>
>I think all you are saying is that we can be uncertain whether we mean.
Not at all. Rather that what we mean is not determined solely by what's in
us. In fact in many cases we can be certain what we mean.
I think the problem with such arguments is this. Note first, I am a
kind of property dualist (actually a pluralist). I am this kind of
dualist about the relation between economic properties and material
ones, and in particular I don't think the former supervene locally on
the latter. That should be uncontroversial. Externalists think a
similar relation holds between intentional contents of thoughts
(including subjective phenomenology) and structural states of brains.
Now the important point is that this non-local supervenience *includes*
both the content of "how things seem to the subject" as well as the content
of second-order states which refer to first order states. For example,
Oscar and Twin-Oscar might both be said to know what they mean.
The real issue is not the causal role of brains in making the higher level
possible, but the powerful illusion of autonomy, the illusion that content
must be determinable entirely by resources solely within myself. In this
way the conscious subject would not be dependent on anything and could,
like Descartes, hope to justify all its knowledge from a vantage point of
suspended belief in anything else.
I take it this is a chimera, and a person qua epistemic subject is
consitituted in dependence on the world and other people.
>Fair enough, but I don't see how that changes the questions one jot or tittle.
>If you stipulate that representation is outside in, and there happens to be no
>outside, then that just means we might be uncertain whether we represent or
>whether we merely seem to represent. And I don't see that our account is really
If there is no outside there is no representation and no entity that
can fall under predicates like "certain" or "uncertain." You might have
just a brain. (Most of us are not in the habit of ascribing any
psychological predicates to brains.) I think it is an illusion to
suppose we have any conception of a thinker without a world, you must
get both together as a system.
Neil Rickert
In <4vr38c$4...@usenet.srv.cis.pitt.edu> ande...@pitt.edu (Anders N Weinstein) writes:
>In article <643507...@luptonpj.demon.co.uk>,
>>I don't think that it could be the case that the organism must *actually*
>>see and touch, since I don't see that I could tell the difference between
>>what's happening to me now and what would happen to me if the universe ended
>>at my skin but the sensory neurons fired just the same.
>>Now you could say at this point that the two explanations would just be totally
>>different.
>*What* two explanations? In the second case there is nothing to explain
>in the representational manner.
I have to agree with Peter on this. I think you are quite mistaken
in saying that there is nothing to explain in the second case. The
sensory neurons are still being fired, presumably for reasons which
have no internal basis. Surely that still requires explanation.
>> But I don't really think that'd wash. It would turn out to be
>>the *same* explanation in both cases except that you'd insist in one case
>>in using the term 'aware' and in the other case you'd insist on using the
>>term 'seems to be aware' and such-like. That apart, the explanations would be
>>identical.
Whether it turns out to be the same explanation depends, I suppose,
on what you mean by 'explanation'. From a god's eye view the
explanation might be different. From the point of view of the person
whose sensory neurons are under discussion, the explanation would be
the same.
>Well this is a standard internalist intuition. But I don't think a
>mere brain would even *seem* to be aware.
Well, of course not. But Peter is not talking about a *mere* brain.
He is talking about a brain that is being stimulated at the sensory
level in a way indistinguishable from that of say your brain. Here
one presumes that indistinguisable is intended to mean
indistinguishable by the electro-chemical activities of the brain.
Such a stimulated brain would surely be aware of the stimulation.
>mere brain would even *seem* to be aware. I think *personal* awareness
>stands to a brain as economic value to a coin -- in different contexts,
>like pieces of metal have different value, and without the right
>surroundings, it's just a worthless hunk of metal. Similarly, a brain
> Similarly, a brain
>on Twin Earth can be part of a person who has one sort of thought, a
>homologous brain on Earth can be part of a person with a different
>psychology (different thoughts), and a brain that was always in a vat
>would be just a brain, and no person would be constituted.
Let's assume that TwinEarth is completely indistinguishable from
Earth. You have an exact double there. Now, while you are asleep
tonight, you are silently exchanged with your double from TwinEarth.
Thus when you wake up, you are on TwinEarth, and your double is on
Earth. Do you die a miserable death because you have suddenly lost
your ability to refer? Or do you refer with apparent continuity to
objects on TwinEarth, never realizing that any change has occurred?
And if the latter, when did this change of reference (if there was a
change in reference) occur?
Jacques Maurice Mallah
Neil Rickert (ric...@cs.niu.edu) wrote:
: mal...@acf2.nyu.edu (Jacques Maurice Mallah) writes:
: > With computation undefined, it certainly could not be the thing
:
: I don't follow the reasoning.
How unusual.
: Should I take it that you are claiming that heat does not reduce to
: molecular kinetic energy? I raise this question, because there was a
: time when people had a reasonably good idea of what they meant by
: 'heat', but had no definition of molecular kinetic energy. If your
: reasoning is valid, it should have been that molecular kinetic energy
: could not explain heat. But the advance of science has shown
: otherwise.
By definition, molecular kinetic energy was not part of those
early models, so obviously it could not be the thing doing the
explanation in the model.
: > Note that I said in our model.
:
: But we don't yet have a good model that encompasses consciousness.
: So what model are you talking about?
:
: > A better statement might be "with computation undefinable ...";
: >then nothing could do the job.
:
: If consciousness is also undefinable, then there would be no obvious
: contradiction in consciousness reducing to computation, and
: computation being undefinable.
You can't make that kind of statement about undefined concepts.
: >: > On what grounds do you say we can't precisely define physical
: >: >quantities?
:
: >: I didn't say that we cannot precisely define quantities. I said that
: >: we cannot precisely define length or force. We can define the
: >: measuring conventions by means of which we quantitatively determine
: >: length. But we cannot precisely define what it actually is that we
: >: measure. We can define the quantities, but not the concepts which
: >: those quantities refer to.
:
: >: Prove me wrong by precisely defining the concept of length.
:
: > How about "that which is measured by those proceedures"?
:
: Fair enough. Now to complete the definition, you need to describe
: the procedures in a way which does not make use of the concept of
: length.
I see no obstacle to doing so.
: > There is another way. This can easily be done within a mathematical
: >model of physics. That is all that's required.
:
: No, sorry, that is not what is required. You might produce a
: marvellous mathematical model of physics. Then I might look at it,
: and think it a marvellous model of abstract art. You haven't defined
: length with your mathematical model until you have connected the
: terms in the model to their real world counterparts. And you cannot
: do that without going outside your mathematical model.
No, sorry, it is what's required. Obviously the way to make a
connection between the model and experiment is to see how things behave.
The model must also apply to measuring instruments, of course.
(Including those that are part of the human body.) You tell me a model,
and if it's a good one, I can make the identification by finding out what
such identification would make what I observe consistent with the model.
If it's a bad one, such an identification can't be made. Often more than
one model can fit the observations; that's when simplicity of the model
is usually taken as the guide to which one to pick.
: > If I say that in such
: >and such a model, I will call certain quantities lengths, that defines
: >it.
:
: And if you call the same quantities cookies, I suppose you will
: instead have defined cookies and they will automatically taste good.
: Sorry, but just naming something `length' in a mathematical model
: does nothing to define length.
The definition within the model should be consistent with the
properties associated with the word, but is more precise. Sometimes
something new, for example the spin of an electron, which Dirac defined
within his model, is given a name which is suggestive of some of its
properties but not all. With the new meaning of the word spin, it is
straightforward to define spin within a newer model of quantum mechanics;
because it is a different model, the meaning need not be exactly the
same, but spin is clearly the right word to use. It is not hard to
concieve that someone on the internet has a model in which spin can't be
defined. What does that mean if the meaning is altered with each new
model anyway? It means there is nothing with properties sufficiently
similar.
: > Given a mathematical model of physics, all that I require is an
: >appropriate formula for what computations are present, which is
: >consistent with the properties associated with computation.
:
: Then you have a nice abstract model, which might have nothing to do
: with the real world.
:
: > There is
: >less freedom here because computation has the property that I intuitively
: >associate certain classes of computations with consciousness; and as
: >defined in the model, that property must be retained.
:
: If you need to rely on intuition, then you are not providing a
: definition.
Intuition is the only guide as to what can be conscious, as you
well know. Using it as a guide, computation is to be defined so that,
given this new and precise definition, it does have the required
property. This is not an easy hurdle for any such definition.
The connection with consciousness provides the link to what
can be observered in that, using this definition, it should be shown that
the kind of perceptions I have are not too atypical among the conscious
perceptions present in the model.
Indeed, the whole goal is to have a situation where I intuitively
find it plausible that my perceptions are not too atypical. When that
happens, I would be fairly confident that the way I think about
consciousness, once I reach that point, is correct and I can then apply
my thinking to things like AI. The definitions of words is to help in a
way similar to the way defining variables is helpful in a mathematical
proof instead of writing out every expression all the time.
The word computation is chosen because it is associated with things
I do intuitively connect with consciousness. In particular, as a
computationalist, I associate each thought (perception) with a
computation, where I mean something to do with conditional state
transitions, but I can't say exactly what I mean because I have yet to
find a suitable definition.
Alternatively, and I prefer to say it this way, a standard
definition of a computation can be used, but implementation of the
computation (whether the computation is physically present) must have a
precisely defined formula.
I'll know the defintion is right when:
1. It is intuitively plausable that each thought is associated with
one.
2. It seems that in a seeming likely model of physics, my perceptions
are not too atypical. This requires *a formula for finding which
computations are present*, and how many of each, as well as some way of at
least connecting certain general properties of some classes of
computations to general properties of thoughts.
What if this program can't be carried out *in principle*? Then the
concepts of computation we have are of no help in understanding
consciousness; and as a matter of fact, we have no better alternatives.
There's only one thing to do in such a case: feel nauseas once; for that
is the appropriate thing upon discovering that reality does not make
sense in principle. That's all I have to say on the subject.
: When two mathematicians talk about pi, them mean the same number.
: But when they try to assign a computation to some natural process,
: they very well may not agree on what computation is appropriate.
If no objective method exists to make the assignments, then my
above described program can't be carried out even in principle.
: > By rejecting Turing machines, you put non-computable and
: >computable math on the same footing.
:
: Not at all. A Turing machine is, by definition, a formal
: abstraction. By rejecting Turing machines, I am simply arguing that
: abstractions are not conscious.
That's not at all what you're doing. Computationalists claim a
central role for objectively present physical implementations of
computations; and computations can always be put into a form that a
Turing machine could perform.
According to you, who ever claimed that abstractions are conscious,
as opposed to physical implementations?
Andrzej Pindor
whoever you grant it to can be trusted in the matters under consideration.
Of course, in other cases granting the authority may be based on hard coding
(called sometimes indoctrination, brainwashing etc.)
>Unfortunately, the idea isn't original to me...
>
The idea of authority being the third alternative or the idea of cramming it
into observation/inference classification?
>CDJ
Andrzej
--
Andrzej Pindor The foolish reject what they see and
University of Toronto not what they think; the wise reject
Information Commons what they think and not what they see.
andrzej...@utoronto.ca Huang Po
CDJ
In article <DwrLA...@gpu.utcc.utoronto.ca>,
pin...@gpu.utcc.utoronto.ca (Andrzej Pindor) wrote:
[I said]
>>It doesn't strike me as implausible that there be _three_ ways in which
>>justification can come off. The first two, direct observation and inference,
>>have already been mentioned. The third would then be, _authority_, as
noticed
>>above. This last is clearly (1) a manner in which things may be justified,
(2)
>>not observational (in the relevant sense), and (3) not inferential (again in
>>the relevant sense - i.e., appeal to authority is no more inferential that
>>appeal to observation).
>>
>>Unless anyone can provide a motivation for trying to cram authority into the
>>observation/inference divide?
>>
[He said]
>At least in some cases granting authority is based on observation that
>whoever you grant it to can be trusted in the matters under consideration.
>Of course, in other cases granting the authority may be based on hard coding
>(called sometimes indoctrination, brainwashing etc.)
Ooops. The discussion has moved to another level. _Now_ we are, apparently
discussing _not_ justification for some proposition p, which might be from
authority, but rather the justification for accepting the authority in the
first place.
Hardcoding, indoctrination, and brainwashing, in the sense used above, have
nothing to do with justification. The question is of _right_, not of _fact_.
His first sentence seems essentially true, though, so long as we are under no
illusions that we are always obligated to provide such an explanation.
[I said]
>>Unfortunately, the idea isn't original to me...
[He said]
>The idea of authority being the third alternative or the idea of cramming it
>into observation/inference classification?
In fact neither idea is original with me. The latter is presumably the elder
of the two. I know of the former from Brandom's work, but I don't assume that
he is to be credited with the idea. It's really not important enough to be
worried about copyright infringement, I don't think.
I don't know that I want to be characterized as saying that authority is _the_
third alternative, just _a_ third alternative - and that only potentially. I
really haven't thought much about it.
CDJ
Anders N Weinstein
In article <4vt4kp$4...@ux.cs.niu.edu>, Neil Rickert <ric...@cs.niu.edu> wrote:
>In <4vr38c$4...@usenet.srv.cis.pitt.edu> ande...@pitt.edu (Anders N Weinstein) writes:
>>In article <643507...@luptonpj.demon.co.uk>,
>
>>>I don't think that it could be the case that the organism must *actually*
>>>see and touch, since I don't see that I could tell the difference between
>>>what's happening to me now and what would happen to me if the universe ended
>>>at my skin but the sensory neurons fired just the same.
>
>>>Now you could say at this point that the two explanations would just be totally
>>>different.
>
>>in the representational manner.
>
>in saying that there is nothing to explain in the second case. The
>sensory neurons are still being fired, presumably for reasons which
>have no internal basis. Surely that still requires explanation.
We already know how to explain neuron firings in neurophysological terms:
they are firing because they are stimulated. Which involves such things
as photosensitive cells, ion channels in the membranes, trasnmission of
an action potential etc.
It is not clear an explanation in terms of representation or
information is applicable to the always envatted brain. Arguably the
neurons in that case are not functioning to represent anything and
their firing carries zero information about an environment in which an
organism has its life.
It is not even clear you are entitled to call them *sensory* neurons,
since that is a functional classification that depends on their having
a role in an organism, which is itself an ecological notion (there can
be no living organism without a normal environment in which its living
takes place).
I should remind you of what Tyler Burge argued in the case of Marr's
theory of vision. Marr took certain computed features of the proximal
arrays to be indicative of distal edges -- computing these features was
supposed to effect edge detection. Burge pointed out that this description
depends on the actual environment in which the computation normally
operated, an environment in which the proximal features were correlated
with edges. On a different planet, in a different environment, the
formally identical computation might constitute a shadow-detector, say.
In Gibson's terms one would say that the information obtained is
different depending on the ecology of the creature. This is
unsurprising -- information is information *about* the environment of
the organism, after all. Functional explanations in biology are not
individualistic, so why should psychological explanations be, if
psychology is -- as it must be -- rooted in biology.
However one puts it, I think even cognitive science -- any science that
uses a notion of representational content or of information in its
explanations -- is not individualistic in its determination of content.
>Whether it turns out to be the same explanation depends, I suppose,
>on what you mean by 'explanation'. From a god's eye view the
>explanation might be different. From the point of view of the person
>whose sensory neurons are under discussion, the explanation would be
>the same.
I disagree. First, in the case of the mere brain, there might be no
person constituted and so no personal point of view to consider.
Secondly in one sense things even *seem* differently to the person on
Earth and on Twin Earth, because the intentional contents of the
seemings they enjoy -- itself a species of propositional attitude --
can be different in the two cases.
I am not interested in a god's eye view. I am talking about the
person's eye view. But that itself is something that is not locally
supervenient, not determined solely by what's in the brain and what
reaches its surfaces. Intentional content, like information content, is
a world-dependent notion.
>Well, of course not. But Peter is not talking about a *mere* brain.
>He is talking about a brain that is being stimulated at the sensory
>level in a way indistinguishable from that of say your brain. Here
>one presumes that indistinguisable is intended to mean
>indistinguishable by the electro-chemical activities of the brain.
>Such a stimulated brain would surely be aware of the stimulation.
Not obvious to me. I decline to apply psychological predicates like
"aware" to brains. You mean the stimulation would cause further
firings down the line, which is true. But neural firings as such are
unconscious, not instances of personal awareness. And the conditions
for talking about a subject of awareness may lie outside the brain,
in form of life, say.
>Let's assume that TwinEarth is completely indistinguishable from
>Earth. You have an exact double there. Now, while you are asleep
>tonight, you are silently exchanged with your double from TwinEarth.
>Thus when you wake up, you are on TwinEarth, and your double is on
>Earth. Do you die a miserable death because you have suddenly lost
>your ability to refer? Or do you refer with apparent continuity to
No, but your history anchors your thought to stuff on Earth.
>objects on TwinEarth, never realizing that any change has occurred?
>And if the latter, when did this change of reference (if there was a
>change in reference) occur?
The usual idea among philosophers is that at first your word for
"water" (and so your water-thoughts), say, continues to refer to the
stuff on Earth. With time and further causal interactions on
Twin-Earth, it can come to shift its reference to the stuff on Twin
Earth. There can be cases where no determinate reference can be identified.
Neil Rickert
In <4vvdva$f...@usenet.srv.cis.pitt.edu> ande...@pitt.edu (Anders N Weinstein) writes:
>In article <4vt4kp$4...@ux.cs.niu.edu>, Neil Rickert <ric...@cs.niu.edu> wrote:
>>I have to agree with Peter on this. I think you are quite mistaken
>>in saying that there is nothing to explain in the second case. The
>>sensory neurons are still being fired, presumably for reasons which
>>have no internal basis. Surely that still requires explanation.
>We already know how to explain neuron firings in neurophysological terms:
>they are firing because they are stimulated. Which involves such things
>as photosensitive cells, ion channels in the membranes, trasnmission of
>an action potential etc.
That is like saying that we know how to explain the action of a
piston, because the spark plug is igniting the gasoline. But such an
explanation says very little about how an automobile works. You have
to describe the way the pistons are interconnected with the
crankshaft, how the cam shaft rotation is synchronized with the
piston motion, etc, etc.
My point is that describing only the action of individual neurons
trivializes the situation.
>It is not clear an explanation in terms of representation or
>information is applicable to the always envatted brain.
Right. And we can't say that the automobile on a hoist represents
the road resistance. But if we place it on a dynamometer, capable of
applying suitable loads, then various parts of the operation are
representing the load, even if it is a simulated load coming from the
dymamometer.
> Arguably the
>neurons in that case are not functioning to represent anything and
>their firing carries zero information about an environment in which an
>organism has its life.
Saying it is arguably so does not make it actually so.
>It is not even clear you are entitled to call them *sensory* neurons,
>since that is a functional classification that depends on their having
>a role in an organism, which is itself an ecological notion (there can
>be no living organism without a normal environment in which its living
>takes place).
You are quite right. You can argue everything away if you want to be
pedantic enough.
>I should remind you of what Tyler Burge argued in the case of Marr's
>theory of vision. Marr took certain computed features of the proximal
>arrays to be indicative of distal edges -- computing these features was
>supposed to effect edge detection. Burge pointed out that this description
>depends on the actual environment in which the computation normally
>operated, an environment in which the proximal features were correlated
>with edges. On a different planet, in a different environment, the
>formally identical computation might constitute a shadow-detector, say.
Is that supposed to prove something?
>In Gibson's terms one would say that the information obtained is
>different depending on the ecology of the creature. This is
>unsurprising -- information is information *about* the environment of
>the organism, after all. Functional explanations in biology are not
>individualistic, so why should psychological explanations be, if
>psychology is -- as it must be -- rooted in biology.
Yesterday you isolated me in a room, and gave me a quadratic equation
to solve. It happens that the quadratic equation came from a problem
in geometry, but you didn't tell me that. I solved the problem.
Today you isolated me in a room, and gave me a quadratic equation to
solve. It happens that this quadratic equation came from a problem
in mechanics, but you again didn't tell me that. I again solved the
problem. The quadratic equation happens to have been identical on
both occasions.
You are arguing that you that you gave me different information on
the two occasions. You are arguing that yesterday you gave me
information about geometry, and today you gave me information about
mechanics. But that is simply false. Yesterday you gave me
information about a quadratic equation, and today you gave me
information about a quadratic equation. You never gave me
information about geometry or mechanics. You might have yourself had
the additional information that yesterday the equation derived from
geometry, and today's equation derived from mechanics. But that
additional information was never part of the information given to
me. I had the same information on both occasions.
>However one puts it, I think even cognitive science -- any science that
>uses a notion of representational content or of information in its
>explanations -- is not individualistic in its determination of content.
In essence, you are insisting on a God's eye view, where we take God
to be omniscient. In my quadratic equation illustration, you play
the part of God. You are insisting that the information is different
on two different occasions if God knows of a difference. In essence,
cognitive science becomes an empty subject, and we must instead study
cognitive theology.
>>Whether it turns out to be the same explanation depends, I suppose,
>>on what you mean by 'explanation'. From a god's eye view the
>>explanation might be different. From the point of view of the person
>>whose sensory neurons are under discussion, the explanation would be
>>the same.
>I disagree. First, in the case of the mere brain, there might be no
>person constituted and so no personal point of view to consider.
In the case of a mere body there might be no person constituted
either. I don't see that this amount to more than arguing about
angels on heads of pins, or about the possibility of Chalmer's
zombies.
>Secondly in one sense things even *seem* differently to the person on
>Earth and on Twin Earth, because the intentional contents of the
>seemings they enjoy -- itself a species of propositional attitude --
>can be different in the two cases.
Right. They are different in God's eyes. But the difference is
theological.
>I am not interested in a god's eye view. I am talking about the
>person's eye view.
So you say. But the hypothesis was there Earth and TwinEarth were
indistinguishable, so there could be no difference in the person's
eye view unless the person has a pipeline to God.
> But that itself is something that is not locally
>supervenient, not determined solely by what's in the brain and what
>reaches its surfaces. Intentional content, like information content, is
>a world-dependent notion.
Of course. But it is only world-dependent because you are insisting
on a God's eye view of intentionality, rather than the view of the
intentional person. In other words, I see the problem you raise as a
purely artificial problem which properly belongs in theology, not in
science.
>>Well, of course not. But Peter is not talking about a *mere* brain.
>>He is talking about a brain that is being stimulated at the sensory
>>level in a way indistinguishable from that of say your brain. Here
>>one presumes that indistinguisable is intended to mean
>>indistinguishable by the electro-chemical activities of the brain.
>>Such a stimulated brain would surely be aware of the stimulation.
>Not obvious to me. I decline to apply psychological predicates like
>"aware" to brains. You mean the stimulation would cause further
>firings down the line, which is true. But neural firings as such are
>unconscious, not instances of personal awareness. And the conditions
>for talking about a subject of awareness may lie outside the brain,
>in form of life, say.
This is just an argument about angels on heads of pins. It happens
that we don't have an agreed language to talk of awareness. Either
we agree to allow a broad interpretation, or we decide to evade the
question on the basis of unimportant technicalities.
>>Let's assume that TwinEarth is completely indistinguishable from
>>Earth. You have an exact double there. Now, while you are asleep
>>tonight, you are silently exchanged with your double from TwinEarth.
>>Thus when you wake up, you are on TwinEarth, and your double is on
>>Earth. Do you die a miserable death because you have suddenly lost
>>your ability to refer? Or do you refer with apparent continuity to
>No, but your history anchors your thought to stuff on Earth.
Only in God's view. There is nothing available to me which can do
the anchoring. It is exactly comparable to the quadratic analogy I
raised above.
>The usual idea among philosophers is that at first your word for
>"water" (and so your water-thoughts), say, continues to refer to the
>stuff on Earth. With time and further causal interactions on
>Twin-Earth, it can come to shift its reference to the stuff on Twin
>Earth. There can be cases where no determinate reference can be identified.
Again, I see this as an essentially theological view. I don't see
how it fits with science. It seems to me that if we look at things
in this way, then science becomes impossible. The only plausible
description of science would be that we are born with complete
knowledge of everything available to us in our Fodorian language of
thought, and all scientists are doing is finding a common public
language to allow them to discuss what they already know innately.
There could be no such thing as scientific discovery.
Anders N Weinstein
In article <4vvq2q$6...@ux.cs.niu.edu> you write:
>In <4vvdva$f...@usenet.srv.cis.pitt.edu> ande...@pitt.edu (Anders N Weinstein) writes:
>
>>We already know how to explain neuron firings in neurophysological terms:
>>they are firing because they are stimulated. Which involves such things
>>as photosensitive cells, ion channels in the membranes, trasnmission of
>>an action potential etc.
>
>That is like saying that we know how to explain the action of a
>piston, because the spark plug is igniting the gasoline. But such an
>explanation says very little about how an automobile works. You have
>to describe the way the pistons are interconnected with the
>crankshaft, how the cam shaft rotation is synchronized with the
>piston motion, etc, etc.
No, its like saying that only in a certain context does piece of
metal constitute a "piston", or a "crankshaft". Given that context,
we can talk about a piston as something that has a proper function,
even if its in the shop. But if an artifact was created by quantum
accident in the form of an internal combustion engine, it would contain
no pistons or crankshafts, just bits of metal with no evident function,
no distinction between proper and improper operation.
In that case, it would also not be an information processing system either.
>My point is that describing only the action of individual neurons
>trivializes the situation.
But you are describing a situation in which the context in which one
can speak of function is missing.
>Right. And we can't say that the automobile on a hoist represents
>the road resistance. But if we place it on a dynamometer, capable of
>applying suitable loads, then various parts of the operation are
>representing the load, even if it is a simulated load coming from the
>dymamometer.
OK, but you should think about the context in which we can even call it
an automobile.
>> Arguably the
>>neurons in that case are not functioning to represent anything and
>>their firing carries zero information about an environment in which an
>>organism has its life.
>
>Saying it is arguably so does not make it actually so.
I mean it is controversial. My own view is that there is no way to
determine functional norms for a brain in a vat, hence no distinction
between correct and incorrect operation, hence no information or
representational content.
One is suckered by the illusion that we can identify such things by
analogy with normal human brains. But the context that would determine
the relevant norms is lacking. It is the same for functional artifacts,
only the context is diffferent.
>>It is not even clear you are entitled to call them *sensory* neurons,
>>since that is a functional classification that depends on their having
>>a role in an organism, which is itself an ecological notion (there can
>>be no living organism without a normal environment in which its living
>>takes place).
>
>You are quite right. You can argue everything away if you want to be
>pedantic enough.
It is not pedantry to recognize that norms of proper function depend
on context, not structure alone. This is true in biology too -- whether
something is a wing, etc. cannot be determined by local structure alone.
>>I should remind you of what Tyler Burge argued in the case of Marr's
>>theory of vision. Marr took certain computed features of the proximal
>>arrays to be indicative of distal edges -- computing these features was
>>supposed to effect edge detection. Burge pointed out that this description
>>depends on the actual environment in which the computation normally
>>operated, an environment in which the proximal features were correlated
>>with edges. On a different planet, in a different environment, the
>>formally identical computation might constitute a shadow-detector, say.
>
>Is that supposed to prove something?
No. But it is an argument to consider. "Information" as used in
cognitive science is not individuated individualistically. There is no
good reason, just prejudice, for thinking information must be individuated
by the local brain structure alone.
>Yesterday you isolated me in a room, and gave me a quadratic equation
>to solve. It happens that the quadratic equation came from a problem
>in geometry, but you didn't tell me that. I solved the problem.
>
>Today you isolated me in a room, and gave me a quadratic equation to
>solve. It happens that this quadratic equation came from a problem
>in mechanics, but you again didn't tell me that. I again solved the
>problem. The quadratic equation happens to have been identical on
>both occasions.
>
>You are arguing that you that you gave me different information on
>the two occasions. You are arguing that yesterday you gave me
>information about geometry, and today you gave me information about
>mechanics.
Not at all.
> But that is simply false. Yesterday you gave me
>information about a quadratic equation, and today you gave me
>information about a quadratic equation. You never gave me
>information about geometry or mechanics.
Agreed.
I don't see that the example is relevant. I am thinking of information
mainly as the intentional content of your perceptual state, for example,
when you are immediately conscious of the world around you. These are
intentional contents that depend for their existence on that world.
I think externalism *also* holds for the sub-personal information studied
in cognitive science, either Marr's way *or* Gibson's. But that is because
sub-personal information is individuated by reference to functional norms,
ultimately rooted in the form of life of the organism.
So either way, information is externally individuated. But you are talking
about a case where I don't give you much information at all.
You are not really in the position with respect to the visible environment of
someone who has been given an equation or a representation of it and can
only guess at its distal source. Your brain if we personify it is in something
like that position, but that doesn't matter, since the brain isn't really
conscious and doesn't really understand the information that floats around
in it. So we can individuate the information at that level externally,
since *really* information is as nothing to a brain.
>In essence, you are insisting on a God's eye view, where we take God
>to be omniscient. In my quadratic equation illustration, you play
>the part of God. You are insisting that the information is different
>on two different occasions if God knows of a difference. In essence,
>cognitive science becomes an empty subject, and we must instead study
>cognitive theology.
This is ridiculous. Burge pointed out, rightly in my view, that
externalism is already true of existing cognitive science as practiced
by Marr and others. No god's eye view is needed. You might as well say
that biology requires a god's eye view because it talks about how
organisms live in their environments.
It is true that if I am wrong about the environment, then I will be
wrong about the information inside my brain if I theorize about it.
There's no way to avoid this. If I change my views about the
environment, I can go on to change my cognitive science. These are two
things that go on together.
>In the case of a mere body there might be no person constituted
>either.
True.
>I don't see that this amount to more than arguing about
>angels on heads of pins, or about the possibility of Chalmer's
>zombies.
Perhaps not, but it's tough if that's the way content is individuated. It
certainly seems true on any version of biological functionalism that you
can't determine a proper function by local structure alone, ditto for
functional explanation of automobile engines or computers. I'm not looking
to start scholastic debates, just pointing out that there's nothing really
wrong or surprising about externalism -- it's just like externalism about
economic value or biological function.
You need to think about the rationale for using the notion of representational
content at all, given that neurons don't respond as such to content properties.
Information content is *always* extrinsic to the neurons, *always* causally
irrelevant to what happens when one neuron hits another, on the cognitive
scientific view. Moreover, information already depends on a kind of holism,
since the information in a signal depends in part on what the receiver
can extract from it. So: what could be wrong with individuating it by
reference to a holistic system that is larger than the organism? It doesn't
affect the causal properties one jot, since information content doesn't
make any difference to the operation of the system.
The point is that the ascription of information content to a brain state
has no predictive value if you are explaining in the manner of
physical science. Rather it works by enabling us to view the operation of
that state in the light of a norm of proper functioning. Which norm might
as well be externally determined.
Perhaps your view is something like Fodor's in _The Elm and the Expert_.
Fodor suggests local structural properties try to reliably track
content properties, just as the government tries to keep manifest
properties of dollar bills in line with economic properties.
>>Secondly in one sense things even *seem* differently to the person on
>>Earth and on Twin Earth, because the intentional contents of the
>>seemings they enjoy -- itself a species of propositional attitude --
>>can be different in the two cases.
>
>Right. They are different in God's eyes. But the difference is
>theological.
They are different in our eyes too. If Oscar expresses a desire for
water, it is a desire for a different stuff than the desire Twin Oscar
expresses in like-sounding vocalizations.
Why is this any more problematic than a similar feature about economic
value?
>>I am not interested in a god's eye view. I am talking about the
>>person's eye view.
>
>So you say. But the hypothesis was there Earth and TwinEarth were
>indistinguishable, so there could be no difference in the person's
>eye view unless the person has a pipeline to God.
Why? This is what I am rejecting. I reject the idea that the person's eye
view is locally supervenient. What's the problem with that?
>Of course. But it is only world-dependent because you are insisting
>on a God's eye view of intentionality, rather than the view of the
>intentional person. In other words, I see the problem you raise as a
>purely artificial problem which properly belongs in theology, not in
>science.
I am not raising a problem. But I repeat that your charge of god's eye
view-ism is circular -- you are presupposing a certain view of what's
knowable from the subject's own point of view, which an externalist
can reject.
Why should psychology be world-independent, when no other science
is? Why should it adopt a vantage point of suspended belief in anything
outside the mind, when no other science does?
>>No, but your history anchors your thought to stuff on Earth.
>
>Only in God's view. There is nothing available to me which can do
>the anchoring. It is exactly comparable to the quadratic analogy I
>raised above.
Not at all. You don't have any representation which you must interpret.
>Again, I see this as an essentially theological view. I don't see
>how it fits with science. It seems to me that if we look at things
>in this way, then science becomes impossible. The only plausible
>description of science would be that we are born with complete
>knowledge of everything available to us in our Fodorian language of
>thought, and all scientists are doing is finding a common public
>language to allow them to discuss what they already know innately.
>There could be no such thing as scientific discovery.
I don't see how any of this follows. I think you need to recognize
that science is only possible on the basis of a prior being-in-the-world.
Given an organism in an environment, we can go on to talk about that
organism, if it develops science, as forming hypotheses about unseen
aspects of the world on the basis of its knowledge of the environment.
But it seems to me there is no prospect of ascribing scientific
theorizing to an antecedently worldless subject, an entity that does not
already have some understanding of its environment.
The process by which this initial understanding develops is not itself an
intellectual one.
I thought you agreed that scientific language depends for its intelligibility
upon a background of skills, which I would say include a prior common-sense
knowledge of a real environment.
Neil Rickert
In <5008a0$i...@usenet.srv.cis.pitt.edu> ande...@pitt.edu (Anders N Weinstein) writes:
>In article <4vvq2q$6...@ux.cs.niu.edu> you write:
>I mean it is controversial. My own view is that there is no way to
>determine functional norms for a brain in a vat, hence no distinction
>between correct and incorrect operation, hence no information or
>representational content.
>One is suckered by the illusion that we can identify such things by
>analogy with normal human brains. But the context that would determine
>the relevant norms is lacking. It is the same for functional artifacts,
>only the context is diffferent.
I think you are looking at this wrongly. For sure, the brains need
an environment. But if the brain is an information processing
system, then it needs an information rich environment. From an
information processing perspective, the important part of the
environment is the rich supply of information. The exact details of
how that information arose need not be of central concern.
>It is not pedantry to recognize that norms of proper function depend
>on context, not structure alone. This is true in biology too -- whether
>something is a wing, etc. cannot be determined by local structure alone.
Yes, but for an information processing system, the context is
information.
> "Information" as used in
>cognitive science is not individuated individualistically. There is no
>good reason, just prejudice, for thinking information must be individuated
>by the local brain structure alone.
I am not particularly concerned with how "information" is used in
cognitive philosophy. My concern is with a notion of "information"
which allows me to solve problems. I don't see that as possible with
the way you are using the term.
>I don't see that the example is relevant. I am thinking of information
>mainly as the intentional content of your perceptual state, for example,
>when you are immediately conscious of the world around you. These are
>intentional contents that depend for their existence on that world.
But I don't see that at all. I think you have it backwards. The
information comes first, before there could be intentional contents.
>I think externalism *also* holds for the sub-personal information studied
>in cognitive science, either Marr's way *or* Gibson's. But that is because
>sub-personal information is individuated by reference to functional norms,
>ultimately rooted in the form of life of the organism.
Externalism is a particular approach you are pushing. I'm rejecting
it, because it seems clear that it cannot solve the problems that
cognitive scientists should be investigating.
>This is ridiculous. Burge pointed out, rightly in my view, that
>externalism is already true of existing cognitive science as practiced
>by Marr and others.
So much the worse for existing cognitive science.
> You might as well say
>that biology requires a god's eye view because it talks about how
>organisms live in their environments.
Biology deals with the physical organism, so needs to consider the
physical environment. AI deals with the information processing, so
must consider the informational environment.
>It is true that if I am wrong about the environment, then I will be
>wrong about the information inside my brain if I theorize about it.
>There's no way to avoid this. If I change my views about the
>environment, I can go on to change my cognitive science. These are two
>things that go on together.
In effect, you seem to be implying that cognitive science cannot
begin until we have a completed science. That is, cognitive science
cannot begin until it has been solved.
>>I don't see that this amount to more than arguing about
>>angels on heads of pins, or about the possibility of Chalmer's
>>zombies.
>Perhaps not, but it's tough if that's the way content is
individuated.
Sorry, but there is no such thing as "the way content is
individuated." You have a particular theory, and "content" and
"individuation" are theoretical terms in that theory. There may be
such a thing as how content is individuated in your theory. But it
is a mistake to treat "concept" and "individuation" as human
independent concepts whose relation can be factually determined. The
more important question is whether your particular theory is
sufficiently extensible to allow it to settle the questions of
cognitive science. I am doubtful as to that, so have chosen a
different route.
> It
>certainly seems true on any version of biological functionalism that you
>can't determine a proper function by local structure alone, ditto for
>functional explanation of automobile engines or computers.
Leave the computers out of this. The typical functional description
of the computer is in terms of formal automata. It is a local
description based on the information processing abilities.
>You need to think about the rationale for using the notion of representational
>content at all, given that neurons don't respond as such to content properties.
Neurons are the wrong entities. Processes are the right entities.
It may take many neurons to carry out a process. Incidently, it is
you who keeps insisting on representational content. I largely avoid
the term `content', since I do not find it helpful.
> Moreover, information already depends on a kind of holism,
>since the information in a signal depends in part on what the receiver
>can extract from it.
Now you are getting closer to a suitable notion of information. But
if it depends on what the receiver can extract from it, it becomes a
subjective notion.
> So: what could be wrong with individuating it by
>reference to a holistic system that is larger than the organism?
Because to do so denies the role of information.
It is as if you are identifying two channels, 'I' and 'C' (although I
am sure you would not describe it in this manner). Channel 'I'
contains the information, and channel 'C' is the content. I am
saying that channel I doesn't contain the information, if it requires
an additional channel. Either we must say that 'C' carries the
information, or that 'I' and 'C' jointly carry the information. We
cannot say that 'I' carries the information, since 'I' by itself does
not inform.
Roughly speaking, you might be using the word `information' where I
would prefer to say `data'. You then want content to be normative
for your information. From my perspective, information needs no
norm. Data might require a norm, to distinguish that data which is
information from that data which is misinformation or noise. But
once we are talking about information, then we have already
eliminated the misinformation and the noise, so we are beyond the
point where a norm is required.
> It doesn't
>affect the causal properties one jot, since information content doesn't
>make any difference to the operation of the system.
If information content does not make any difference, then it is an
irrelevancy and a distraction. That is why I reject its importance.
Perhaps our disagreement is mainly one of different interests. My
concern is with solving the problems for which content does not
matter. You apparently have quite different concerns.
>The point is that the ascription of information content to a brain state
>has no predictive value if you are explaining in the manner of
>physical science.
I am not particularly concerned with ascribing information content to
brain states. I don't expect much progress to be made by attempts to
do so.
>Perhaps your view is something like Fodor's in _The Elm and the Expert_.
>Fodor suggests local structural properties try to reliably track
>content properties, just as the government tries to keep manifest
>properties of dollar bills in line with economic properties.
Perhaps so. I'll have to take another look at that. I am inclined
to doubt that I would mean the same thing as Fodor would mean.
You might say that I am looking at us as being in an informationally
rich environment. There is a great deal of structure in the
information, presumably derived from the structure of the world. Our
job as cognitive agents is to recover the structure within the
information, and use that as our best guide to the structure of the
world.
You seem to have a much thinner notion of information. It is
certainly possible to have information which is very thin in the
sense of having little recognizable structure, and must be supported
by conventions adopted by the sender and receiver of the
information. Many of the data communication channels built by
technology are of this form. There is very little structure that can
be recognized in the information on such a channel, except for
perhaps for syntactic information such as the transmission bit rate.
However, we are not in that situation at all. We have billions of
sensory neurons collecting information. There is a great deal of
semantic structure, in the form of correlations between the different
signals that we access.
>>>Secondly in one sense things even *seem* differently to the person on
>>>Earth and on Twin Earth, because the intentional contents of the
>>>seemings they enjoy -- itself a species of propositional attitude --
>>>can be different in the two cases.
>>Right. They are different in God's eyes. But the difference is
>>theological.
>They are different in our eyes too. If Oscar expresses a desire for
>water, it is a desire for a different stuff than the desire Twin Oscar
>expresses in like-sounding vocalizations.
You are changing the story. I did insist that TwinEarth be
indistinguishable from Earth, which implies that the water on
TwinEarth should be indistinguishable from the water on Earth and
thus be equally satisfying to Oscar and to TwinOscar.
>I am not raising a problem. But I repeat that your charge of god's eye
>view-ism is circular -- you are presupposing a certain view of what's
>knowable from the subject's own point of view, which an externalist
>can reject.
More correctly, I am rejecting your strictly externalist position as
one which cannot solve the problems of cognitive science. On a
strictly externalist position, we are all zombies, and there could be
no such thing as cognitive science.
>Why should psychology be world-independent, when no other science
>is? Why should it adopt a vantage point of suspended belief in anything
>outside the mind, when no other science does?
I haven't said that psychology is world-independent.
The position I am taking, is that psychology is information
dependent, and information is world dependent. But we can study
psychology in terms of information, without having to directly
concern ourselves as to the dependencies between the world and the
information.
> I think you need to recognize
>that science is only possible on the basis of a prior being-in-the-world.
>Given an organism in an environment, we can go on to talk about that
>organism, if it develops science, as forming hypotheses about unseen
>aspects of the world on the basis of its knowledge of the environment.
>But it seems to me there is no prospect of ascribing scientific
>theorizing to an antecedently worldless subject, an entity that does not
>already have some understanding of its environment.
No, I don't agree at all. Take the case of Priestley and Lavoisier.
They sought information about phlogiston, and conducted experiments
to acquire that information. On your way of looking at things, we
would have to say that the content was about phlogiston, and that the
content is normative for what is the information. That was
Priestley's position, and he continued to hold to the phlogiston
theory. Lavoisier, by contrast, looked at it differently. Lavoisier
made the nature of the information normative for determining the
content. As a consequence, he determined that the content was really
about oxygen, rather than phlogiston.
It seems to me that you are insisting on a way of doing science which
allows the scientist to find out all kinds of things about
phlogiston, but does not allow the scientist to discover oxygen.
Andrzej Pindor
In article <4vtngd$9...@usenet.srv.cis.pitt.edu>, CDJ <cd...@pitt.edu> wrote:
>In article <DwrLA...@gpu.utcc.utoronto.ca>,
> pin...@gpu.utcc.utoronto.ca (Andrzej Pindor) wrote:
>
>[I said]
>>>It doesn't strike me as implausible that there be _three_ ways in which
>>>justification can come off. The first two, direct observation and inference,
>>>have already been mentioned. The third would then be, _authority_, as
>noticed
>>>above. This last is clearly (1) a manner in which things may be justified,
>(2)
>>>not observational (in the relevant sense), and (3) not inferential (again in
>>>the relevant sense - i.e., appeal to authority is no more inferential that
>>>appeal to observation).
>>>
>>>Unless anyone can provide a motivation for trying to cram authority into the
>>>observation/inference divide?
>>>
>
>[He said]
>>At least in some cases granting authority is based on observation that
>>whoever you grant it to can be trusted in the matters under consideration.
>>Of course, in other cases granting the authority may be based on hard coding
>>(called sometimes indoctrination, brainwashing etc.)
>
>Ooops. The discussion has moved to another level. _Now_ we are, apparently
>discussing _not_ justification for some proposition p, which might be from
>authority, but rather the justification for accepting the authority in the
>first place.
To evaluate your suggestion one has to have a clear notion what 'from
authority' means. As I have attempted to point out depending what 'authority'
is based on it may either mean guiding oneself by a set of previous
observations or accepting something as given. In the former case, it is not
much different from accepting "what goes up must come down" as true on the
basis of numerous observations. This was basically my point. In the latter
case, it is an alternative. I would not say "the third alternative" since one
might argue that using inference is basically refering to the authority of
the rules of logic. Personally I believe that the authority granted to the
rules of logic comes from observations (that their use leads to 'true'
propositions) although they may also be hardwired in our brains (by evolution
probably).
>
>Hardcoding, indoctrination, and brainwashing, in the sense used above, have
>nothing to do with justification. The question is of _right_, not of _fact_.
>
This may depend on what you mean by 'justification'. I took 'justification'
as "a reason why we believe someting to be true". Hardcoding etc. may be
such a reason, might it not? In other words you seem to stop at "the prophet
said so" as the reason, I am trying to look behind what makes us to accept
what the prophet said as "true" - is it his record or "just because"?
>His first sentence seems essentially true, though, so long as we are under no
>illusions that we are always obligated to provide such an explanation.
>
I am not sure what 'obligated' means in this case. It seems to me a very
different meaning of "from authority" if it is a summary (even if
unconscious) of observations or if it is "just because" (i.e. hardwired one
way or another).
.....................
Anders N Weinstein
In article <501v36$8...@ux.cs.niu.edu>, Neil Rickert <ric...@cs.niu.edu> wrote:
>In <5008a0$i...@usenet.srv.cis.pitt.edu> ande...@pitt.edu (Anders N Weinstein) writes:
>I think you are looking at this wrongly. For sure, the brains need
>an environment. But if the brain is an information processing
>system, then it needs an information rich environment. From an
>information processing perspective, the important part of the
>environment is the rich supply of information. The exact details of
>how that information arose need not be of central concern.
In one sense this is right. But I would recommend a definition of
"information environment" in which (like "social environment" or "historical
environment") the information environments can be different on Earth and
Twin Earth. As I understand it, information (=content) is information
*about* the environment and so depends for its nature on the nature of
the environment.
>I am not particularly concerned with how "information" is used in
>cognitive philosophy. My concern is with a notion of "information"
>which allows me to solve problems. I don't see that as possible with
>the way you are using the term.
Evidently you don't think the Marr approach "solves problems". I wonder
why not.
It is still a bit mysterious to me how the one vs the other notion of
information helps solve a scientific problem.
>>I think externalism *also* holds for the sub-personal information studied
>>in cognitive science, either Marr's way *or* Gibson's. But that is because
>>sub-personal information is individuated by reference to functional norms,
>>ultimately rooted in the form of life of the organism.
>
>Externalism is a particular approach you are pushing. I'm rejecting
>it, because it seems clear that it cannot solve the problems that
>cognitive scientists should be investigating.
Far from clear, if it is already unreflectively presupposed by most of
them. Perhaps you could say that scientists should be investigating
things in terms that do not use the notion of information at all -- that
I would agree with, the concept of information does not add much
of scientific value as I see it. It is a way of seeing something in the
light of a norm.
>>This is ridiculous. Burge pointed out, rightly in my view, that
>>externalism is already true of existing cognitive science as practiced
>>by Marr and others.
>
>So much the worse for existing cognitive science.
You keep talking as if there's some terrible problem with externalism, as
if it were mystical or theological or something like that. But I see no
no problem. If there were a problem, how come Marr's theory doesn't bump
up against it? How come Gibson's doesn't? How come everyday assessments
of intentional contents (also not individualistic) don't?
None of these practice merit the ridiculous charges you make about "god's
eye views" and the like.
>> You might as well say
>>that biology requires a god's eye view because it talks about how
>>organisms live in their environments.
>
>Biology deals with the physical organism, so needs to consider the
>physical environment. AI deals with the information processing, so
>must consider the informational environment.
Right! But the information environment is basically a feature of the
organismic niche, and so is externally individuated. Why not? You
never cite any problem here, even though your unfounded opposition
seems very strong.
>>It is true that if I am wrong about the environment, then I will be
>>wrong about the information inside my brain if I theorize about it.
>>There's no way to avoid this. If I change my views about the
>>environment, I can go on to change my cognitive science. These are two
>>things that go on together.
>
>In effect, you seem to be implying that cognitive science cannot
>begin until we have a completed science. That is, cognitive science
>cannot begin until it has been solved.
It can begin, it just can't finish. I think this should be obvious. For
example, someone who doesn't understand quantum mechanics themselves
simply can't do the epistemology of quantum physicists arguing about
the interpretation of an experimental result. Nor could this be done by
someone in ancient times, say.
If you lack the concepts, you can't cognize the relevant aspects of the
environment. If you can't cognize the relevant aspects of the
environment, you can't see what it is your subjects are responding to
-- its like a blind man trying to theorize about color information
which they themselves are incapable of receiving. From their point of
view, the subjects are reacting to something invisible. So of course
you couldn't say what information they are picking up if you are,
as it were, not tuned to their frequency.
Similarly, someone who believes in witches *must* say that the environment
contains information about witches. Once you reject the witch theory,
you must reject that characterization of the information environment. This
seems obviously mandatory to me.
The alternative would be to take cognitive science to play the role
hoped for in traditional Cartesian epistemology -- that it provides a
neutral ahistorical framework which presupposes nothing about the world
and provides the court of appeal for all possible knowledge claims. But
I think this is fantasy, there is no such discipline, since all knowledge,
including brain and cognitive science, rests on presuppositions and
prejudices which are simply inherited from the culture, which may themselves
have evolved by accident, although they can come to be incrementally
revised.
>>>I don't see that this amount to more than arguing about
>>>angels on heads of pins, or about the possibility of Chalmer's
>>>zombies.
>
>>Perhaps not, but it's tough if that's the way content is
>individuated.
>
>Sorry, but there is no such thing as "the way content is
>individuated." You have a particular theory, and "content" and
>"individuation" are theoretical terms in that theory. There may be
>such a thing as how content is individuated in your theory. But it
>is a mistake to treat "concept" and "individuation" as human
>independent concepts whose relation can be factually determined. The
>more important question is whether your particular theory is
>sufficiently extensible to allow it to settle the questions of
>cognitive science. I am doubtful as to that, so have chosen a
>different route.
Fair enough. However, I do think, first, it is the way common-sense
individuates content, and, second, that it is necessary if knowledge is
to be an intelligible possibility.
Are you denying that if someone on Twin-Earth claims that some liquid
is "water", their claim is made true only if the liquid is (roughly) XYZ
and not Earthian H20? Why do you think we use this practice when evaluating
people's overt claims? True enough, it has nothing to do with scientific
explanation of the noises that come out of people's mouths. Why
do we individuate inner thoughts in *exactly the same way* as we do overt
claims? Is there something wrong with these practices that *science* (of
all things) could show?
I don't myself think the supposed demands of science are relevant in
any way to philosophy of mind -- it's a subject like philosophy of
tables. Wanting to be scientific is always the problem in this area,
never the solution.
>Leave the computers out of this. The typical functional description
>of the computer is in terms of formal automata. It is a local
>description based on the information processing abilities.
But the only reason such a description doesn't apply to arbitrary physical
objects is because human beings manufactured a computer with a certain
purpose.
>Neurons are the wrong entities. Processes are the right entities.
>It may take many neurons to carry out a process. Incidently, it is
>you who keeps insisting on representational content. I largely avoid
>the term `content', since I do not find it helpful.
If you ever use words to express information content you are helping yourself
to the representational content of those words. Or is your notion of content
inexpressible? For me "information" and "representational content" are
the same.
>> Moreover, information already depends on a kind of holism,
>>since the information in a signal depends in part on what the receiver
>>can extract from it.
>
>Now you are getting closer to a suitable notion of information. But
>if it depends on what the receiver can extract from it, it becomes a
>subjective notion.
I have already said I think there is nothing "subjective" about it.
Perhaps subject-dependent, but that applies to many objective things,
arguably everything.
>It is as if you are identifying two channels, 'I' and 'C' (although I
>am sure you would not describe it in this manner). Channel 'I'
>contains the information, and channel 'C' is the content. I am
>saying that channel I doesn't contain the information, if it requires
>an additional channel. Either we must say that 'C' carries the
>information, or that 'I' and 'C' jointly carry the information. We
>cannot say that 'I' carries the information, since 'I' by itself does
>not inform.
But 'I' and 'C' are one and the same. There is no distinction to be drawn,
information *is* representational content as I use it. I guess in your
terms I really only believe in channel C, but use the term "information"
interchangeably to denote it.
>Roughly speaking, you might be using the word `information' where I
>would prefer to say `data'. You then want content to be normative
OK, in your terms I should speak of "data" or "content". But my point
is then that we can simply dispense with any idea of "information" that is
independent of these in the case of human conceptual consciousness.
I guess the idea that beyond these, there is something epistemically
relevant called "information" would be an instance of the "myth of the
given", ie. the idea that something non-conceptual could stand in a
rational epistemic relation to a belief or judgment.
I should say I do acknowledge a role for a sub-personal notion of
information, but not in epistemology, not in explaining how justified
knowledge of the world arises, in science or elsewhere. Rather part
of the non-rational causal prerequisites of such a process. But even
this is a normative notion, since it is in the context of a teleological
explanation of a functional system.
>for your information. From my perspective, information needs no
>norm. Data might require a norm, to distinguish that data which is
>information from that data which is misinformation or noise. But
>once we are talking about information, then we have already
>eliminated the misinformation and the noise, so we are beyond the
>point where a norm is required.
Don't understand this exactly. The concept is still norm-involving if
it depends on distinguishing mis-information and noise.
>If information content does not make any difference, then it is an
>irrelevancy and a distraction. That is why I reject its importance.
>Perhaps our disagreement is mainly one of different interests. My
>concern is with solving the problems for which content does not
>matter. You apparently have quite different concerns.
I bet this is deceptive on your part. But can you explain to me
what some information is, if you do not use the idea of information
content? Evidently it cannot be information that one object has gone
behind another, or that something in the world affords mailing (Gibsonian
examples) or information that something is an edge (Marr example).
What's an example? Can it be expressed in words? How do you talk about
information without talking about its content?
Anyway, if you are not using any notion of representational content, then
what could your inquiry have to do with knowledge or epistemology or
philosophy of mind?
>I am not particularly concerned with ascribing information content to
>brain states. I don't expect much progress to be made by attempts to
>do so.
OK, what is the role of 'information' in your theory?
>You might say that I am looking at us as being in an informationally
>rich environment.
Me too.
>information, presumably derived from the structure of the world. Our
>job as cognitive agents is to recover the structure within the
>information, and use that as our best guide to the structure of the
>world.
Why not: recover the information itself? (Not just its structure.)
>You seem to have a much thinner notion of information. It is
I think it is thicker, since it is not formal, but reaches all the
way out to the world.
>certainly possible to have information which is very thin in the
>sense of having little recognizable structure, and must be supported
I would have thought information has no structure at all; it
is content, not structure.
>by conventions adopted by the sender and receiver of the
>information. Many of the data communication channels built by
>technology are of this form. There is very little structure that can
>be recognized in the information on such a channel, except for
>perhaps for syntactic information such as the transmission bit rate.
I don't have a syntactic concept of information, but a semantic one.
(Admittedly, I feign no a technical concepts at all but rely on a
vague and intuitive one. But still it's not syntactic or formal or structural)
>You are changing the story. I did insist that TwinEarth be
>indistinguishable from Earth, which implies that the water on
>TwinEarth should be indistinguishable from the water on Earth and
>thus be equally satisfying to Oscar and to TwinOscar.
It might be subjectively satisfying. I meant that water is not in the
"conditions of satisfaction" of his intentional state in Searle's
sense. If TwinOscar asks for water and you give him H20, he may be
subjectively satisfied but he did not get what he wanted. (This is not
mysterious, it happens all the time -- if I ask for hamburger and you
give me a soy imitation, I might be none the wiser and wind up content
nonetheless. But the content of my original request was certainly not
satisfied.)
>More correctly, I am rejecting your strictly externalist position as
>one which cannot solve the problems of cognitive science. On a
Which is what exactly, on your view?
>strictly externalist position, we are all zombies, and there could be
>no such thing as cognitive science.
How does this follow? I am an ordinary language philosopher -- I don't
think we can tell that people are not zombies by ordinary criteria, and I
am interested in the concepts we actually employ when talking about mental
states -- something as mundane and everyday as possible, not something
outrageous like a god's eye view, or the thought that there might be zombies.
It is true I don't really care if there is such a thing as cognitive science.
>I haven't said that psychology is world-independent.
>
>The position I am taking, is that psychology is information
>dependent, and information is world dependent. But we can study
I can agree with this.
>psychology in terms of information, without having to directly
>concern ourselves as to the dependencies between the world and the
>information.
How, if the information is world-dependent? Again this looks to be
the Fodor view -- the processes deal with structures that bear information
but they themselves do not deal with information itself, only the structure
that is correlated with it. But if this is your view, then you should have
no objection at all to individuating information externally, since the
information content is not a causal factor in the explanatory model.
>No, I don't agree at all. Take the case of Priestley and Lavoisier.
>They sought information about phlogiston, and conducted experiments
>to acquire that information. On your way of looking at things, we
>would have to say that the content was about phlogiston, and that the
>content is normative for what is the information. That was
We certainly have to say that the content of Priestly's judgments were
putatively concerned with phlogiston. But there is a tricky detail
here. If the phlogiston theory is false and there isn't really any
phlogiston, then there isn't any *information* about phlogiston to
obtain from the environment either.
On the view to which I incline, this also means that there was no real
content about phlogiston either -- because phlogiston is a natural kind
term, the non-existence of phlogiston in the world means that
Priestly's thoughts weren't even *possibly* true -- because we can't
be clear on what circumstances would satisfy it.
So Priestly *thought* there was information about phlogiston in the world,
thought that he could obtain it, thought that he was having thoughts that
were possibly true. But given what we now believe, we must say he was
wrong to think all these things. I.e discovering that phlogiston does
not exist changes our assessment of the nature of Priestly's thoughts
and of the information available in the environment.
You are right that such a view makes cognitive science impossible if it
is supposed to be a world-independent story which quantifies over every
possible world. But since the individuation of thoughts is not something
that is independent of what we believe about the world I don't see that
there is any vantage point from which to pull off such a project.
>Priestley's position, and he continued to hold to the phlogiston
>theory. Lavoisier, by contrast, looked at it differently. Lavoisier
>made the nature of the information normative for determining the
>content. As a consequence, he determined that the content was really
>about oxygen, rather than phlogiston.
We really need to be careful about what it is that is bearing a content here.
I am not sure exactly what it means to say that Lavoisier and Priestly
are assigning different contents to the same thing -- *what* exactly?.
I might say that they had thoughts with different contents, and further
that Priestly's phlogiston-thoughts turned out in fact to be
pseudo-contents and did not constitute his ever acquiring any information
about phlogiston, since there is none in the world to acquire.
There are many extremely difficult issues raised here by Kuhn's idea of
meaning change, e.g. whether oxygen could be taken to be denoted by
"dephlogisticated air."
>It seems to me that you are insisting on a way of doing science which
>allows the scientist to find out all kinds of things about
>phlogiston, but does not allow the scientist to discover oxygen.
Nonsense. Where do I commit myself to anything that denies this?
In view of the difficulty with respect to meaning change, I am
wondering what is your description of the process and the information
processed within?
H. M. Hubey
ande...@pitt.edu (Anders N Weinstein) writes:
>So this description is not individualistic, since it identifies it as
>having an information processing function -- it is supposed to compute
>the logical AND of two information-bearing signals by reference to
>its systemic function.
How about getting rid of some redundancy. Doesn't every signal
bear information whether it is perceivable or not by humans?
Isn't this essentially the problem with the different defns of
information?
We can only get information via our senses (directly); the
standard 5 senses. If information is objective it can't be
dependent on the receiver, so every signal bears information.
We can call that defn 1. On the other hand we can try some
other defn of information, say, something that gives more
knowledge to the receiver. This defn is not like the first since
it is receiver dependent. The same signal may or may not
carry information. In that case maybe we should define some
of these concepts like in communications books:
1) signal
2) data
3) information
in this case information is something that can be assigned
a meaning by means of some convention. Data is a representation
of facts, concepts, instructions in a formalized manner
suitable for communication, interpretation, or processing
by humans (normally, since we still can't separate knowledge
and information from the knower). The signal is what is sent
accross space (and can be acoustic or electromagnetic).
>But of course we still identify the information processing function
>*independently* of the actual information contents being AND'ed. Perhaps
>it will be deployed in one context so as to compute the AND of signals
>from sensors in the state of a nuclear reactor; in another context to
>compute the AND of signals from sensors monitoring the state of an
>automobile engine.
Then there are at least 3 kinds of processing. First is when the
signal is processed to yield data. 2nd is when data is processed to
yield information. Third is when this information(?) is processed
by different individuals to yield different levels/magnitudes of
knowledge/information. I know I am causing problems with the 3rd
one. But without that we can't give the probabilistic interpretation
of information; that is, a message that is expected (i.e. has
high probability) doesn't have much information content. For
example "NYC is in NY state" is not info to most people.
"Martians have invaded" has lots of information content.
This doesn't even take up on the folk-psychological-linguistic-logic
variants/defns of information and how it is processed. It seems
they skip all the steps and roll them into one.
>Again, however, I see no reason why we can't individuate the
>information content in the final state externally in this case as
>well. As part of "syntactic psychology" we can say that the formally
How do you propose measuring this information content. Are we
suppose to wave our hands, and "abra-cadabra", there it is.
Ken Ewell
Jacques Maurice Mallah <mal...@acf2.nyu.edu> wrote in article
<4vkrm5$9...@news.nyu.edu>...
> : > Which clearly implies that you agree with him that computation is
> : >only an interpretation.
I fail to see how one can regard the term _computation_ as appropriate to
refer to anything more than a fundamental ability at counting, adding, or
subtracting.
Kenneth Colby
A good argument against strong reductionism is that no one
has ever produced one in the sense of deducing the generalizations
of one theory from the generalizations of a more fundamental
theory WITHOUT REMAINDER. Ungrounded ideological hopes can
last a long time.
KMC
Anders N Weinstein
In article <502k3d$r...@usenet.srv.cis.pitt.edu>,
I <ande...@pitt.edu> wrote in reply to Neil Rickert:
> Again this looks to be
>the Fodor view -- the processes deal with structures that bear information
>but they themselves do not deal with information itself, only the structure
>that is correlated with it. But if this is your view, then you should have
>no objection at all to individuating information externally, since the
>information content is not a causal factor in the explanatory model.
Let me elaborate a bit on this idea. One might say that it is the job
of computational models to function in the following manner:
Suppose I am manufacturing components for digital logic circuits -- an
AND gate, say. Now this characterization identifies the task by
reference to functional norms, by reference to the role it will play in
a computational device. That means it is not, for example, a device
for heating the air near the circuit, nor an analog computer in which
continuous variations in voltage are significant. This even though the
physical object might be perfectly suited for use as either, and
type-indistinguishable physical objects might be produced across the
street in the analog computer factory and even in a heater factory which
would not have different functional norms.
So this description is not individualistic, since it identifies it as
having an information processing function -- it is supposed to compute
the logical AND of two information-bearing signals by reference to
its systemic function.
But of course we still identify the information processing function
*independently* of the actual information contents being AND'ed. Perhaps
it will be deployed in one context so as to compute the AND of signals
from sensors in the state of a nuclear reactor; in another context to
compute the AND of signals from sensors monitoring the state of an
automobile engine.
In this sense we are talking about an information processing component
in a way that does not depend on the information being processed. We
are doing so because we are talking about it purely formally (or
perhaps, because "AND" is a purely formal concept.) This gives rise to
the idea that computational pyschology should stick to "narrow
contents" which are solely a matter of formal "syntactic"
characterizations of the functioning. (This was roughly the idea of
Stich, _From Folk Psych. to Cog Sci._)
In one sense, this view of the task of computational psychology is
perfectly compatible with the idea that the actual information being
processed in any deployed device may be externally determined -- info
concerning reactor states in one case, concerning engine state in
another. It's just that computational modelling concerns the formal
properties of an information processing device, not the material ones
(the actual content that gets processed).
I am wondering if something like this may be part of Rickert's outlook.
But of course there is another philosophical issue which is concerned
with how and why the sensor signals come to have *their* particular
material content. Here causal embedding would seem to be a crucial part
of the story.
Really Rickert is more concerned with something like the following:
consider manufacturing a complex circuit that will operate in a
learning system, one that is *not* simply hard-wired to represent, say,
a reactor's state. Perhaps it will be part of a robot that must explore
unfamiliar environments. In any case, it is being designed to *evolve*
transducers for useful properties based on patterns in its low-level
sensor input.
Again, however, I see no reason why we can't individuate the
information content in the final state externally in this case as
well. As part of "syntactic psychology" we can say that the formally
identical learning algorithm might be deployed in one context as in
another, just as the same AND-gate might be deployed in different
contexts. We thus have an uninterpreted description of the structure
that is common to the two cases, even though this description has one
eye on the information-processing function and so does not eliminate
the idea that information is being processed in the system.
Still there seems to be on reason we can't say that in one context, on
Earth say, what this thing is doing is evolving an edge-detector; while
in another, on Twin Earth, it is evolving a shadow-detector. This in
the same way we might say that one organism is growing a wing while
another is growing its cooling system, even if the organisms are
structually identical (even down to the DNA!) in virtue of different
evolutionary histories.
The computational theory is not concerned with the semantic aspects,
but not because it is using a different narrow notion of information,
but because it is simply not concerned with material information
content at all, but remains at a formal, uninterpreted level. But that
is no reason to suppose the broad-content individuation of information
is problematic, just not relevant to this level of description.
Neil Rickert
>>So this description is not individualistic, since it identifies it as
>>having an information processing function -- it is supposed to compute
>>the logical AND of two information-bearing signals by reference to
>>its systemic function.
>How about getting rid of some redundancy. Doesn't every signal
>bear information whether it is perceivable or not by humans?
>Isn't this essentially the problem with the different defns of
>information?
Quite right. This is a problem I avoid, by using a subjective notion
of information (that which informs me).
Neil Rickert
In <505533$8...@usenet.srv.cis.pitt.edu> ande...@pitt.edu (Anders N Weinstein) writes:
>In article <504j5o$a...@ux.cs.niu.edu>, Neil Rickert <ric...@cs.niu.edu> wrote:
>You keep responding in these general terms. But I really am just trying
>to understand the concept of information that you are working with.
>Some examples would help.
Since my concept of information is subjective, you can't exactly
expect me to provide an objective definition. I'll try to look for
some good examples in future posts. I'm really considering
information at a primitive level, before it is conceptualized, and
that makes it hard to describe in terms of our standard concepts.
>>And that is why Kuhn's work must seem paradoxical to you.
>No, I am in a way a devotee of Kuhn. I am also however an empirical
>realist. To put it provocatively, I believe that progress happens
>through conceptual change, but I also believe that in developing better
>and better concepts we are coming to carve nature at its natural
>joints, getting closer to Nature's own language.
Ok, I pretty much agree with that, except for the dubious assumption
that there is such a thing as "Nature's own language."
>The paradox is that Kuhn in effect accepts the Kantian distinction
>between the world as it is in itself, a noumenal world which would be
>unknowable, and the world as it is conceptualized or represented by us,
>which we can know but which is merely phenomenal or a kind of
>construction.
>If one makes this distinction, you are faced with the
>problem that we have no access the transcendental process by which the
>phenomenal world is constructed.
Why must we assume it is constructed by a transcendental process?
What is wrong with information processing.
> In particular, you cannot say it is
>constructed by our brains or that cognitive science is studying this
>process. Brains themselves are mere phenomenal objects, only in the
>world as represented, and cognitive science as an empirical discipline
>also deals only with phenomenal objects.
I don't see that as a problem. If brains are mere phenomenal
objects, then they are objects we construct to model the information
processing. We can still examine the information processing, whether
or not we consider the brains to be real.
>If I were really worried that the whole world that appears to me might
>be merely the content of a constrained hallucination of mine,
I'm not a solipsist. I don't think the world is a hallucination.
I'll agree that if we allow that possibility, then everything becomes
hopeless. But if it happens that there are details of the world that
are not required in a description of the processes, the we should not
insist on incorporating those details.
>I think the main way out is to reject the idea that knowledge of the
>world is mediated by representations. Rather by concepts, which are not
>representations but capacities.
It is very likely that I am not using the term `representation' in
the same way that you are. I am opposed to the type of
representationism that Fodor supports, for example. But there is a
more primitive sense that I am using. If you receive visual
information by means of light entering your eyes, then the light wave
represent the information. If you are able to use that information
because of neural signals, then those neural signals represent the
information. I don't consider "f = ma" to be an information
representation. But I do consider "The monitor on my desk has a mass
of 8.3 kilograms" as a representation of information (or perhaps
misinformation).
>>Many of the problems being discussed in epistemology and in
>>philosophy of mind are not that much different from the problems that
>>were being discussed by Plato. There is very little evidence of
>>progress. In science, progress has come with conceptual change. The
>>old concepts were often stumbling blocks, which hindered progress.
>>I'm trying to avoid the stumbling blocks by changing to a more
>>suitable conceptual basis.
>Me too. I am trying to reject the idea that knowledge of the world
>is mediated by representations. I am also trying to reject the idea
>that cognitive science studies the mind.
I'm not much concerned with what cognitive science should be said to
study, nor am I much concerned with whether what I am doing is
properly called cognitive science. I'm interested in being able to
understand how we learn, perhaps for purposes such as designing
better curricula.
>>What problems am I supposed to cite? I thought the problems were
>>well known, and the failure to traditional approaches to solve them
>>are also well known.
>By "tradition" you mean the modern tradition from Descartes --
>Aristotle didn't have the problems. And the source of the problems with
>modern epistemology are individualism, the representational theory of
>the mind, and the corrupting influence of natural scientific paradigms
>on thinking about rational capacities. Junk these, and the
>pseudo-problems are exposed for what they are -- they vanish without a
>trace.
I agree with much of this. But just junking those ideas does not
answer the questions that concern me.
>Now suppose as many epistemologists who follow Kant believe, one doesn't
>have data or content without concepts.
I don't suppose that. Perhaps one does not have content without
concepts, particularly if one defines content in terms of concepts.
But I think the role of concepts is quite different from what is
usually assumed. In the case of the scientist, I would argue that
the scientist makes his basic discovery without any necessary
reliance on concepts. He or she must then construct concepts
suitable for making the discovery available to other scientists.
Thus I see concepts as necessary for the social aspects of science,
rather than for the basic discoveries.
> Then we cannot say about the
>brain what we can say about the scientist -- for the brain doesn't really
>deal with any data. Until the conceptual capacities have been developed,
>the organism just isn't an epistemic or cognitive system at all.
That is simply a matter of defining the problems away. But to do so
is to make everything mysterious.
> So the
>process by which a thing which is not a concept-using epistemic agent
>develops into one *cannot* itself be modelled as an intellectual one, one
>involving theory and evidence (or data).
I agree that the process is not an intellectual one, but I do so for
rather different reasons. I would rather say that knowledge has a
substantial non-conceptual basis, and therefore the process could not
be an intellectual one.
>I think we learn from Wittgenstein (and, in the scientific case, by
>Kuhn), that this emergence is brought about through training which
>constitutes intitiation into a pre-existing (though open-textured)
>conceptual repertoire.
While I agree that training is important, I think you overrate its
role.
>I don't know what the alternative is, if the idea that progress
>requires conceptual change is genuine. If the physicists themselves
>are struggling to develop new concepts, what language can the cognitive
>scientist use to describe the process they are undergoing? Surely the
>cognitive scientist does not have a better understanding of the
>confused concepts then the physicists themselves.
I don't see that the cognitive scientist has to deal with the same
concepts as the physicist.
>If new concepts are like random mutations of old species, than of
>course the process by which they emerge is not necessarily a lawful
>one, one governed by rational laws, although it can be described.
Progress is too rapid to be explained by random mutations.
>> If so, why can't the neurons follow a similar
>>methodology? If the neurons can follow such a methodology, then your
>Well mainly because neurons aren't the kind of thing that has concepts.
I don't suggest that neurons have concepts. Again, I don't give
concepts the same role you assign to them.
>We were not born knowing the latest sartorial fashions. Do you think
>there must be a systematic methodology by which cultural evolution
>develops new styles of dress?
I don't much care, since I don't take fashions to have the same
progressivity as science. I have no problem with saying that there
is a substantial random component to fashions, although the available
technology also plays a role.
>I don't know where you think I have committed myself to this. We have
>lots of intellectual history characterizing the emergence of new concepts.
>But why should one think this is a formalizable or algorithmic process
>("follows a methodology")?
Again, science has developed far too rapidly, to be explained on the
basis of the filtering of random guesses.
>>What information they are picking up is a secondary concern. They
>>must start by picking it up. Only then can they try to identify the
>>information. The information must be there, and recoverable, before
>>one can begin to talk of content. If you make consideration of
>>content a prerequisite, then divine interventionism becomes the
>>possibility.
>Again, note the importance for me of the distinction between theoretical
>and observational knowledge.
I think that is often a misleading distinction.
> In science there is data which can be cognized
>before one forms a theory. Perhaps one can say that neural circuits are
>somewhat analogous to scientists -- they 'formulate concepts" to account
>for "data" (patterns in proximal stimulation). But really they have no
>data, because the have no concepts.
The neurons can have data, but no concepts.
>I think it is the intellectualist legend to try to explain the
>emergence of conceptual capacities on the model of a scientist forming
>theories about antecedently available data. (You must think that at
>least the data language of the brain is innate, where I don't think a
>brain can have a language)
I don't see any need for the brain to have a language. It is
excessive intellectualizing to assume that a language is required.
>>Sure. And those presuppositions were in turn passed down in the DNA
>>from the earliest organisms created by God. Your argument, if
>>correct, implies creationism of some form.
>I said they came from the culture, not the DNA. DNA is irrelevant to
>this level of description.
But to say so just brushes the problems aside. You still have to
explain how they came to be in the culture, and how the person
acquires them from the culture. I don't see that this move settles
any of the problems.
>Where do I commit myself to innatism? More extreme culturalists think
>of the human being prior to socialization as an almost completely
>malleable piece of putty, one that can be shaped in almost any way
>imaginable by upbringing.
I think there is something to that, although it may be exaggerated.
But one cannot credit the shaping exclusively to the culture. Even
if one could, one would have to explain the mechanism that makes such
shaping possible.
>I guess it is important that I don't think of concepts as brain states,
>so looking inside the brain is looking in the wrong place at the wrong
>level of description.
I don't think of concepts as brain states either. I'm not even
convinced that one can usefully talk of brain states.
>>Not so if one follows Ryle by insisting that knowledge is "knowing
>>how."
>But there is *also* knowing-that to account for.
I expect that much of the `knowing that' amounts to inference from
the `knowing how'.
>>>I bet this is deceptive on your part. But can you explain to me
>>>what some information is, if you do not use the idea of information
>>>content? Evidently it cannot be information that one object has gone
>>>behind another, or that something in the world affords mailing (Gibsonian
>>>examples) or information that something is an edge (Marr example).
>>Why can that not be information?
>Because to describe it that way relates it to the world.
But that might be an artifact of the way you are describing it, and
thus not by itself decisive as to whether it is information.
>Because to describe it that way relates it to the world. It characterizes
>it in a way such that it is true if and only if a particular worldly
>object really does afford mailing or not, say.
Again, I think `truth' is the wrong standard. This is why Kuhn found
that science could not be explained on the basis of searching for
truth. Very often a scientific theory itself sets a standard for
truth. Such a theory cannot both set the standard for truth, and be
itself evaluated on the basis of truth.
> On Twin Earth, with similarly
>shaped objects but no postal system, you would not use the same words to
>describe the information.
Again, the words don't matter. What matters is that the information
provide you with an opportunity for effective action.
>>> Can it be expressed in words? How do you talk about
>>>information without talking about its content?
>>Gibson's idea of affordance is good enough. That is, a signal is
>>information if I can use it to make effective decisions. I need not
>>be concerned with what the signal affords (in the sense in which you
>>insist on content). If I can use the signal to make decisions in a
>>way that correlates with my well being, that is enough to demonstrate
>>that the signal is informative. I can postpone concerns about the
>>content until later.
>But Gibson's affordances were the affordances of distal objects, not of
>signals, and were objective properties of them -- such as whether a
>particular object really does afford mailing.
Think of the affordance as an opportunity for action. I don't fully
accept Gibson's view about distal objects, for there are times that
we have information but have not yet identified the source of that
information.
> (In general, I think the
>signals themselves do not have any affordances for the organism.)
The information in the signal has the affordance. It doesn't matter
whether one says that the signal has an affordance.
>Whereas your idea does not ascribe any content to the signal, and so any
>information. You have not explained how to talk about information without
>putting a content into words.
Not at all. I do not need to ascribe any content to the signal. That
may mean that I do not ascribe any AW-information. That has nothing
to do with whether there is NR-information. Admittedly it will often
be the case that AW-information and NR-information are both present.
But there will be also be cases when there is only NR-information.
The scientist will use that NR-information as a basis for
experiments. As a result of those experiments, the scientist will
announce that he has discovered X. Future historians, in a
restrospective analysis, will say that there was AW-information after
all, for the content was of X. But I cannot explain the action of
the scientist on the basis that there will be future ascriptions of
content. So I have to explain it on the basis of it being
NR-information. The scientist recognized the presence of
NR-information because of patterns of correlations observed in the
signals. Quite possibly the signals were not beforehand known to
carry any information. The scientist did not have to know anything
about the X that was to be discovered. It suffices that the
scientist noticed the NR-information. That is enough for the
scientist to design exploratory experiments. The scientist might say
that his experiments are to help him find the cause of the
correlational pattern he has noticed. At this stage there need by no
hypothesis and no new concept, so no hypothesis testing is going on.
The scientist is simply interested in the NR-information he has
observed. The exploratory experiments may allow him to gather more
information, and to begin to hypothesize, and to construct the
concept X.
>>That is why your position is nativist, and ultimately creationist.
>Again, this is an absurd charge. Imagine a more extreme me as someone
>who thinks a knower at birth is a tabula rasa, and that nurture is all,
>if you don't see that. In being conditioned to use symbols in accordance
>with the patterns exhibited by my fellows, one comes to be a concept user.
>Where is the innatism?
You have not explained how you can be conditioned to use symbols. I
don't see that problem as any easier than the problem of acquiring
knowledge about the world. If the second is impossible without
prerequisites, then so is the first.
>>You can't begin until you have the concepts needed to represent
>>content. Thus you require a creator to provide you with those
>>concepts.
>Why not the culture? Are you not reading what I write?
I'm reading it. But talk of it coming from the culture is easy
talk. Chomsky certainly has made strong arguments that we cannot be
conditioned suitably with the impoverished evidence available.
>> Simple signal processing methods can find such
>>structure without any conceptual prerequisites.
>Simple signal processing cannot develop concepts, however.
As I have said, concepts only become necessary when we need to
communicate our discoveries to others. The scientist may have to do
a lot more work before he is ready to form concepts. But he is
guided by the outcome of the simple signal processing.
>>>> But we can study
>>>>psychology in terms of information, without having to directly
>>>>concern ourselves as to the dependencies between the world and the
>>>>information.
>>>How, if the information is world-dependent?
>>I don't see the problem you are suggesting. To take an analogy,
>>mathematics is used for dealing with information that is world
>>dependent. But the mathematical procedures are described in a manner
>>which ignores the world dependent aspects of the information.
>OK, then you are just ignoring the content, but should have no problem
>if the content is externally determined.
It doesn't trouble me that on many occasions the content is
externally determined. But we cannot explain scientific discovery if
we have to know what it is we are going to discover before we
discover it. If you make the content a prerequisite, I think you
have precluded the possibility of explaining scientific discovery.
>>I would not agree with that. The processes deal with the
>>information. One might perhaps say that the processes are
>>implemented by physical structures, and the action of the processes
>>on the information is a consequence of the process implementing
>>structures acting on the information representing structures.
>Which is basically the formalist idea I think -- take care of the
>syntax (structure) and the semantics will take care of itself.
However I am not following the formalist approach, for my concern is
not with syntax.
> This has always
>been Fodor's view, his recent book indicates his idea that there are
>problems with it.
I think it is unworkable. The formalist approach only works because
when we construct concepts, we do so in a way which gives them
suitable syntactic properties. The formalist approach can work after
we have chosen our concepts with care. It cannot guide us in forming
those concepts.
>But if there is an algorithmic method by which the phlogiston theory
>develops under constraint by systematic rational norms into the
>oxygen theory, then Kuhn is *refuted*. For his point entails precisely
>that there is no algorithm for scientific method.
Your comment about "systematic rational norms" may be correct, but I
am not claiming that there are such rational norms. I disagree with
your comment about the possibility of a method.
I have a CD player. I place a CD into it, press the button, and
music comes out. In the music I can recognize notes, chords, pauses,
tempos, etc. The CD player operates according to a methodology. But
the concepts of notes, chords, pauses, tempos, etc, are not concepts
used in that methodology. Since our usual ideas of rationality would
require the rational processes to use those concepts, the CD player
fails the "systematic rational norm" test. Nevertheless you might
say that the CD can be described as working algorithmically with the
fragmentary digital samples used to record the CD.
Kuhn's work only reveals problems with finding a methodology
describable in the usual scientific concepts. It does not preclude
the possibilility of a methodology at an entirely different level of
analysis where the methodology is describable in terms of concepts
which are orthogonal to the concepts in the scientific theory.
>Again, if you believe all scientific debates can be resolved by appeal
>to shared observation languages, you may be right, but you are
>disagreeing with Kuhn.
No, I am not suggesting that at all. Scientific debates can be very
difficult to resolve. But my concern has been with what happens
before there can even be a scientific debate.
>Again, I can concede this is possible where both had the requisite
>concepts. In Quine's terms this information is in observation
>sentences.
I find Quine's observation sentences quite unsatisfactory for
explaining scientific discovery.
H. M. Hubey
ande...@pitt.edu (Anders N Weinstein) writes:
>you astray. When I am conscious of my desk there is no representation,
>no signal, that I have applied an interpretation to. If I have interpreted
>anything, it can only be the desk itself. This is why I say that in
>being conscious of a desk as a desk I am taking it as what it is, not
>applying an interpretation to some proximal signal or representation.
Methinks you are contradicting yourself.
A desk is a desk because of the function it serves. And its form
derives from its function. So it can't be a desk unless and until
you interpret the signals impinging on your retina (assuming you are
using your eyes to be conscious of the desk). Otherwise you
would be saying something like "There's an object comprised of
wood/metal/plastic with a large flat top, and which is supported
on some vertical beams, ... etc." By calling it a desk you have
already interpreted it as to its function, and the word that you use
to describe it is an acknowledgement of such.
Anders N Weinstein
In article <504j5o$a...@ux.cs.niu.edu>, Neil Rickert <ric...@cs.niu.edu> wrote:
>In <502k3d$r...@usenet.srv.cis.pitt.edu> ande...@pitt.edu (Anders N Weinstein) writes:
>
>>It is still a bit mysterious to me how the one vs the other notion of
>>information helps solve a scientific problem.
>
>problems that Ptolemaic notions did not solve. Perhaps it is
>mysterious as to how relativity solves problems that Newtonian
>mechanics did not solve. Perhaps it is mysterious that science seems
>to be a sequence of paradigm shifts. I am trying to provide a basis
>which removes the mystery. I need a notion of information suitable
>to what I am trying to accomplish.
These things are not mysterious to me. The problem has to do specifically
with the role of the concept of information. Many philosophers run into
this problem. It stems from the idea that information content is an
extrinsic property of a system, devoid of causal effects.
You keep responding in these general terms. But I really am just trying
to understand the concept of information that you are working with.
Some examples would help.
>> -- that
>>I would agree with, the concept of information does not add much
>>of scientific value as I see it. It is a way of seeing something in the
>>light of a norm.
>
>And that is why Kuhn's work must seem paradoxical to you.
No, I am in a way a devotee of Kuhn. I am also however an empirical
realist. To put it provocatively, I believe that progress happens
through conceptual change, but I also believe that in developing better
and better concepts we are coming to carve nature at its natural
joints, getting closer to Nature's own language.
The paradox is that Kuhn in effect accepts the Kantian distinction
between the world as it is in itself, a noumenal world which would be
unknowable, and the world as it is conceptualized or represented by us,
which we can know but which is merely phenomenal or a kind of
construction.
If one makes this distinction, you are faced with the
problem that we have no access the transcendental process by which the
phenomenal world is constructed. In particular, you cannot say it is
constructed by our brains or that cognitive science is studying this
process. Brains themselves are mere phenomenal objects, only in the
world as represented, and cognitive science as an empirical discipline
also deals only with phenomenal objects.
If I were really worried that the whole world that appears to me might
be merely the content of a constrained hallucination of mine, I would
have to be equally worried that my belief that I have a brain or that
brains are causally involved in my mental states is similarly only true
inside the hallucinated world, and might not be true in the world
outside of that. I would have to worry not only that there might be no
distant stars, but also that there are really no telescopes, no
spectrometers, no voltmeters, no test tubes, no laboratories, no other
people to collaborate or discuss ideas with. Really no *data* and no
*science* to do the philosophy of. I don't know why you think one
should even consider this possibility.
I think the main way out is to reject the idea that knowledge of the
world is mediated by representations. Rather by concepts, which are not
representations but capacities.
>Many of the problems being discussed in epistemology and in
>philosophy of mind are not that much different from the problems that
>were being discussed by Plato. There is very little evidence of
>progress. In science, progress has come with conceptual change. The
>old concepts were often stumbling blocks, which hindered progress.
>I'm trying to avoid the stumbling blocks by changing to a more
>suitable conceptual basis.
Me too. I am trying to reject the idea that knowledge of the world
is mediated by representations. I am also trying to reject the idea
that cognitive science studies the mind.
>What problems am I supposed to cite? I thought the problems were
>well known, and the failure to traditional approaches to solve them
>are also well known.
By "tradition" you mean the modern tradition from Descartes --
Aristotle didn't have the problems. And the source of the problems with
modern epistemology are individualism, the representational theory of
the mind, and the corrupting influence of natural scientific paradigms
on thinking about rational capacities. Junk these, and the
pseudo-problems are exposed for what they are -- they vanish without a
trace.
>But then what you are arguing, is that we must describe science in a
>way which precludes the possibility of there being a scientific
>method within the description. You are saying that the first quantum
>physicists could not be described as working with information, since
>there were no known quantum particles to provide the content that
>would allow them to have information. It is only in historical
>perspective that we can see that they were working with information
>after all.
Loosely speaking this is correct. It is only retrospectively that we
can judge these things.
But notice that in the case of *scientific knowledge* we are talking
about antecedent data and theory. Up to a point we can describe
experimental results -- atomic spectra, scattering experiments, etc --
in a way that doesn't presuppose the later concepts. In that sense they
had reasonably clear contents concerning observable results, but as
they were struggling to develop the new theoretical framework, did not
have well-defined concepts, and so well-defined theoretical contents.
In retrospect we can see that, say, certain confused ideas were latent
in their activity, struggling to become better defined.
Now suppose as many epistemologists who follow Kant believe, one doesn't
have data or content without concepts. Then we cannot say about the
brain what we can say about the scientist -- for the brain doesn't really
deal with any data. Until the conceptual capacities have been developed,
the organism just isn't an epistemic or cognitive system at all. So the
process by which a thing which is not a concept-using epistemic agent
develops into one *cannot* itself be modelled as an intellectual one, one
involving theory and evidence (or data). It is rather a "transition from
quantity into quality", the emergence of a subject of a new range of
predicates.
I think we learn from Wittgenstein (and, in the scientific case, by
Kuhn), that this emergence is brought about through training which
constitutes intitiation into a pre-existing (though open-textured)
conceptual repertoire.
I don't know what the alternative is, if the idea that progress
requires conceptual change is genuine. If the physicists themselves
are struggling to develop new concepts, what language can the cognitive
scientist use to describe the process they are undergoing? Surely the
cognitive scientist does not have a better understanding of the
confused concepts then the physicists themselves.
Obviously historians of science like Kuhn can describe the process
using the ordinary language they do.
>This leads directly to Fodorian nativism. All of the basic concepts
>must be innate, for otherwise we could not get started. But if they
Nonsense. Your initial concepts were inherited from the culture when
you learned its language. The culture itself emerged out of simpler
forms of activity, presumably including symbolic activity like animal
communication.
If new concepts are like random mutations of old species, than of
course the process by which they emerge is not necessarily a lawful
one, one governed by rational laws, although it can be described.
>are innate, whence did they come. Is there a highly efficient
>methodology available to evolution to provide us with these innate
>concepts? If so, why can't the neurons follow a similar
>methodology? If the neurons can follow such a methodology, then your
Well mainly because neurons aren't the kind of thing that has concepts.
I am not sure evolution "follows a methodology" in the development of
new species, either, so why should there be a methodology for the
development of new concepts?
We were not born knowing the latest sartorial fashions. Do you think
there must be a systematic methodology by which cultural evolution
develops new styles of dress?
>argument about concepts must be wrong. If the neurons cannot follow
>such a methodology, then neither can evolution. You are left with
>them being implanted by a creator. Perhaps the earliest bacterium
>(or whatever was the most primitive form of life) already came fully
>equipped with the concepts we need.
I don't know where you think I have committed myself to this. We have
lots of intellectual history characterizing the emergence of new concepts.
But why should one think this is a formalizable or algorithmic process
("follows a methodology")?
>What information they are picking up is a secondary concern. They
>must start by picking it up. Only then can they try to identify the
>information. The information must be there, and recoverable, before
>one can begin to talk of content. If you make consideration of
>content a prerequisite, then divine interventionism becomes the
>possibility.
Again, note the importance for me of the distinction between theoretical
and observational knowledge. In science there is data which can be cognized
before one forms a theory. Perhaps one can say that neural circuits are
somewhat analogous to scientists -- they 'formulate concepts" to account
for "data" (patterns in proximal stimulation). But really they have no
data, because the have no concepts.
I think it is the intellectualist legend to try to explain the
emergence of conceptual capacities on the model of a scientist forming
theories about antecedently available data. (You must think that at
least the data language of the brain is innate, where I don't think a
brain can have a language)
>>Similarly, someone who believes in witches *must* say that the environment
>>contains information about witches.
>
>Right. And then we can have another round of Salem witch trials.
>All the more reason to be skeptical about the purported value of
>beliefs.
Don't understand this. You've got to go by what you believe -- if it's
false, then tough, but there's no getting around relying on your beliefs
is there?
>But that is only because you tie information to characterization.
>I'm concerned about uncharacterized information.
No such thing, in my view.
>it later if we wish. We need to start collecting it first. We
>should always reserve some skepticism as to any characterization.
How could "uncharacterized information" play a role in reasoning or
scientific practice? You mean experimental results in advance of
theory, but this is not uncharacterized.
>Sure. And those presuppositions were in turn passed down in the DNA
>from the earliest organisms created by God. Your argument, if
>correct, implies creationism of some form.
I said they came from the culture, not the DNA. DNA is irrelevant to
this level of description.
Where do I commit myself to innatism? More extreme culturalists think
of the human being prior to socialization as an almost completely
malleable piece of putty, one that can be shaped in almost any way
imaginable by upbringing. This is the complete opposite of any kind
of innatism -- it says that what is innate is irrelevant, *everything*
that matters in epistemology is acquired. The only thing that's innate
is something like a propensity to imitate the members of one's community
and come to behave like them.
I don't go so far, but I am wondering why you keep ascribing innatism
to me. I have constantly said I think concepts as rules for the use of
linguistic expressions have an ontological status a bit like fashions
or the ability to use signposts. These capacities are acquired in the
individual by internalization from patterns of behavior that are abroad
in the culture. Unlike Skinner, I don't think this is a matter of stimulus
and response; but to have a concept is still primarily to have a skill,
a way of behaving, chiefly with signs.
I guess it is important that I don't think of concepts as brain states,
so looking inside the brain is looking in the wrong place at the wrong
level of description.
>Not so if one follows Ryle by insisting that knowledge is "knowing
>how."
But there is *also* knowing-that to account for.
>>I bet this is deceptive on your part. But can you explain to me
>>what some information is, if you do not use the idea of information
>>content? Evidently it cannot be information that one object has gone
>>behind another, or that something in the world affords mailing (Gibsonian
>>examples) or information that something is an edge (Marr example).
>
>Why can that not be information?
Because to describe it that way relates it to the world. It characterizes
it in a way such that it is true if and only if a particular worldly
object really does afford mailing or not, say. On Twin Earth, with similarly
shaped objects but no postal system, you would not use the same words to
describe the information.
>> Can it be expressed in words? How do you talk about
>>information without talking about its content?
>
>Gibson's idea of affordance is good enough. That is, a signal is
>information if I can use it to make effective decisions. I need not
>be concerned with what the signal affords (in the sense in which you
>insist on content). If I can use the signal to make decisions in a
>way that correlates with my well being, that is enough to demonstrate
>that the signal is informative. I can postpone concerns about the
>content until later.
But Gibson's affordances were the affordances of distal objects, not of
signals, and were objective properties of them -- such as whether a
particular object really does afford mailing. The signal contained
*information* about the affordances of distal objects, which is a
matter of information content in the signal. (In general, I think the
signals themselves do not have any affordances for the organism.)
Whereas your idea does not ascribe any content to the signal, and so any
information. You have not explained how to talk about information without
putting a content into words.
>That is why your position is nativist, and ultimately creationist.
Again, this is an absurd charge. Imagine a more extreme me as someone
who thinks a knower at birth is a tabula rasa, and that nurture is all,
if you don't see that. In being conditioned to use symbols in accordance
with the patterns exhibited by my fellows, one comes to be a concept user.
Where is the innatism?
>You can't begin until you have the concepts needed to represent
>content. Thus you require a creator to provide you with those
>concepts.
Why not the culture? Are you not reading what I write?
>no such problem. Simple signal processing methods can find such
>structure without any conceptual prerequisites.
Simple signal processing cannot develop concepts, however.
>>How does this follow?
>
>On a strictly externalist position, we are collections of quantum
>particles acting under the laws of quantum mechanics.
I'm afraid I must repeat my question: how does this follow? Obviously
there are many externalist descriptive levels. For example, externalism
is true about our social properties, which have nothing to do with
descriptions in terms of quantum mechanics. Why couldn't our
intentional properties be like that? Objectively real, externally
determined, and, in a certain unproblematic sense, immaterial and
non-physical?
>>> But we can study
>>>psychology in terms of information, without having to directly
>>>concern ourselves as to the dependencies between the world and the
>>>information.
>
>>How, if the information is world-dependent?
>
>I don't see the problem you are suggesting. To take an analogy,
>mathematics is used for dealing with information that is world
>dependent. But the mathematical procedures are described in a manner
>which ignores the world dependent aspects of the information.
OK, then you are just ignoring the content, but should have no problem
if the content is externally determined.
>I would not agree with that. The processes deal with the
>information. One might perhaps say that the processes are
>implemented by physical structures, and the action of the processes
>on the information is a consequence of the process implementing
>structures acting on the information representing structures.
Which is basically the formalist idea I think -- take care of the
syntax (structure) and the semantics will take care of itself. (From
John Haugeland's Artificial Intelligence The Very Idea). This has always
been Fodor's view, his recent book indicates his idea that there are
problems with it.
This leaves it open that semantics (= content = information) can be
externally determined, but raises the fear that structure cannot
track the semantics we intuitively want to ascribe. Narrow content
theorists want to invent some new notion of content, but Fodor at least
has better sense than that.
>But then you have to describe the scientists as taking random actions
>with no methodological basis. Only in a historical retrospective,
>when oxygen is available to fill the role of content, can they be
>described as following a method.
They can be described as non-random and in a way methodical. They
just can't be described as having information about phlogiston.
>I am saying that the reason philosophers of science have so much
>trouble with finding a documentable scientific method, one which
>meets the type of historical test used by Kuhn, is that they are
>insisting that science be described in a manner which precludes the
>possibility of their being a scientific method which could fit with
>their description.
But if there is an algorithmic method by which the phlogiston theory
develops under constraint by systematic rational norms into the
oxygen theory, then Kuhn is *refuted*. For his point entails precisely
that there is no algorithm for scientific method.
>Nevertheless, they were dealing with the same information. Both were
>heating mercury calx, and finding the weights of the recovered
>mercury and the properties of the resulting gas. They both found the
Agreed. These descriptions are of information that both had, because the
apply to observations in a theory-neutral way.
Again, if you believe all scientific debates can be resolved by appeal
to shared observation languages, you may be right, but you are
disagreeing with Kuhn.
>same pattern of correlations. As I am using the term, they dealt
>with the same information.
Again, I can concede this is possible where both had the requisite
concepts. In Quine's terms this information is in observation
sentences. I don't deny that scientists in different paradigms share
many observation sentences. My comments concerned the contents of
theoretical sentences.
Anders N Weinstein
In article <504mbt$a...@ux.cs.niu.edu>, Neil Rickert <ric...@cs.niu.edu> wrote:
>In <502mud$s...@usenet.srv.cis.pitt.edu> ande...@pitt.edu (Anders N Weinstein) writes:
>
>>So this description is not individualistic, since it identifies it as
>>having an information processing function -- it is supposed to compute
>>the logical AND of two information-bearing signals by reference to
>>its systemic function.
>
>the `individual' is the culture as a whole, rather than a lone
>person.
This is just wordplay, given the way "individualistic" is used.
> Somebody from a completely different culture might well take
>the computer as a heating device.
They might, but they'd be mistaken about an objective fact. It's not a
heating device.
If people start using computers as heating devices then one could construct
indeterminate cases, but so what, it doesn't affect the clear cases.
>In a computer, some AND gates are being used to AND the information
>in two inputs. Other AND gates are being used as a switches, with
>one input controlling whether the information in the other input is
>switched through to the output. If you were to ask computer
>engineers which was which in a particular computer, there would
>probably be high agreement among them. But the AND gates used for
>one purpose are manufactured in the identical way to the AND gates
>used for the other purpose.
Indeed. But I don't think it affects the general point.
>No, not at all. But something like that is part of Weinstein's
>outlook. Specifically, you insist that the brain be described in the
Rather, my understanding of the computational view in cognitive
science. It is not something I endorse myself, I ascribe it to others.
>The reason that the computer cannot have any psychology, is that the
>human programmer makes the decision as to what is the information the
>computer shall process. Because of that, the computer is at best
OK, but what if something evolved into the same state? Would it then
have a psycholgy? Say yes, and you are one kind of externalist, one
that cites evolutionary history not present state. This is explicit in
the "teleofunctionalist" Fred Dretske, who denies that intelligence
could be artificial (though it might be evolved).
>In effect, you want to describe the ordinary human as processing
>information in a way determined by a committee of expert
>epistemologists who are able to determine what is `content'. In that
I am not sure I want to describe the ordinary human as processing
"information" at all, at least not as you use the term. I do want
to describe the person as being in certain intentional states. Some of
them refer to the visible environment. If the person is a scientist,
then others constitute scientific hypotheses.
And we need no experts, the content is determined by the world itself.
>case the ordinary human has no psychology, and is at best described
>as acting out part of the psychology of the epistemologists. Thus
>the human is a zombie, has no free will, and is merely a slave to its
>circumstances.
This is absurd on my view. I think your representationalism is leading
you astray. When I am conscious of my desk there is no representation,
no signal, that I have applied an interpretation to. If I have interpreted
anything, it can only be the desk itself. This is why I say that in
being conscious of a desk as a desk I am taking it as what it is, not
applying an interpretation to some proximal signal or representation.
>I see no validity to this description of a human. As an autonomous
>cognitive agent, I can not only make my own decisions on the
>information available to me -- I can also make my own decisions as to
>what components of the signals I collect can be considered
>information. Thus, no matter how good your science you cannot
What do you mean you can make decisions about the signals you collect?
if you know what these signals are then you must conceptualize them,
and we are talking about the relation between some objects of consciousness
and others -- which is fine with me, since its all conceptualized.
But if you are talking about signals reaching your brain, we must deny
that you can decide what to do about them since you don't represent
them. At best you can decide where in the world to look and what to do
with your instruments, which of course I agree with.
>predict my next decision, for my next decision might be based on
>information I am collecting but which you are not collecting and thus
>on information which is not included in your predictive model.
I think I can agree with this. This is why I can't do cognitive science
about you if I don't myself share your concepts.
>The environment is informationally rich. We might say that there is
>an infinite amount of information available in the environment. But
Yes.
>we are finite creatures, and can only access a finite part of that
>information. No matter how good is your model, it is only able to
>externally individuate a finite quantity of information. I might be
>collecting and using information which is not in your model.
Indeed. But this would seem to be everybody's problem, not just mine.
How does *your* theory enable you to do cognitive science for a subject
that is responding to information which you the theorist do not
detect?
>> As part of "syntactic psychology" we can say that the formally
>>identical learning algorithm might be deployed in one context as in
>>another, just as the same AND-gate might be deployed in different
>>contexts.
>
>I think yours is really the syntactic psychology, although I expect
>that you will disagree. You are committed to a finitely
I believe that computational modelling might well be taken to be
committed to such a thing. You are the one who is constantly denying
a role for semantics = content.
>Likewise, I doubt that there could be non-zombies in your externalist
>approach.
We are all non-zombies on my externalist approach. Only if you think
we are really locked up in a room "interpreting signals", so that
consciousness of the world is mediated by representations, would you think
something otherwise. But in my view there are no representations
mediating between consciousness and its objects, and so nothing to
interpret but worldly objects themselves.
Neil Rickert
In <502mud$s...@usenet.srv.cis.pitt.edu> ande...@pitt.edu (Anders N Weinstein) writes:
>In article <502k3d$r...@usenet.srv.cis.pitt.edu>,
>I <ande...@pitt.edu> wrote in reply to Neil Rickert:
>Suppose I am manufacturing components for digital logic circuits -- an
>AND gate, say. Now this characterization identifies the task by
>reference to functional norms, by reference to the role it will play in
>a computational device. That means it is not, for example, a device
>for heating the air near the circuit, nor an analog computer in which
>continuous variations in voltage are significant.
This is not obvious to me. People have been known to use old
computers as door stops, and to use milk bottles as flower vases.
>So this description is not individualistic, since it identifies it as
>having an information processing function -- it is supposed to compute
>the logical AND of two information-bearing signals by reference to
>its systemic function.
On the contrary, I think it is highly individualistic. In this case
the `individual' is the culture as a whole, rather than a lone
person. Somebody from a completely different culture might well take
the computer as a heating device.
>But of course we still identify the information processing function
>*independently* of the actual information contents being AND'ed. Perhaps
>it will be deployed in one context so as to compute the AND of signals
>from sensors in the state of a nuclear reactor; in another context to
>compute the AND of signals from sensors monitoring the state of an
>automobile engine.
In a computer, some AND gates are being used to AND the information
in two inputs. Other AND gates are being used as a switches, with
one input controlling whether the information in the other input is
switched through to the output. If you were to ask computer
engineers which was which in a particular computer, there would
probably be high agreement among them. But the AND gates used for
one purpose are manufactured in the identical way to the AND gates
used for the other purpose.
>In this sense we are talking about an information processing component
>in a way that does not depend on the information being processed. We
>are doing so because we are talking about it purely formally (or
>perhaps, because "AND" is a purely formal concept.)
Fair enough. The distinctions you are making are reasonable ones for
talking about our ordinary computers.
> This gives rise to
>the idea that computational pyschology should stick to "narrow
>contents" which are solely a matter of formal "syntactic"
>characterizations of the functioning. (This was roughly the idea of
>Stich, _From Folk Psych. to Cog Sci._)
The fallacy is the implicit assumption that the operation of the
brain can be usefully described in the same way we describe ordinary
computers. I am not making any such assumption.
>In one sense, this view of the task of computational psychology is
>perfectly compatible with the idea that the actual information being
>processed in any deployed device may be externally determined -- info
>concerning reactor states in one case, concerning engine state in
>another. It's just that computational modelling concerns the formal
>properties of an information processing device, not the material ones
>(the actual content that gets processed).
>I am wondering if something like this may be part of Rickert's outlook.
No, not at all. But something like that is part of Weinstein's
outlook. Specifically, you insist that the brain be described in the
same way that we describe ordinary computers. You then conclude that
brains have no psychology, and cannot be persons. I'll agree with
you on how describe conventional computers, and I will thus agree
that conventional computers have no psychology and cannot be
persons. But that type of description is not adequate for the
brain.
The reason that the computer cannot have any psychology, is that the
human programmer makes the decision as to what is the information the
computer shall process. Because of that, the computer is at best
acting out part of the human psychology under the direction of the
programmer. So the computer has no psychology, and is a zombie.
In effect, you want to describe the ordinary human as processing
information in a way determined by a committee of expert
epistemologists who are able to determine what is `content'. In that
case the ordinary human has no psychology, and is at best described
as acting out part of the psychology of the epistemologists. Thus
the human is a zombie, has no free will, and is merely a slave to its
circumstances.
I see no validity to this description of a human. As an autonomous
cognitive agent, I can not only make my own decisions on the
information available to me -- I can also make my own decisions as to
what components of the signals I collect can be considered
information. Thus, no matter how good your science you cannot
predict my next decision, for my next decision might be based on
information I am collecting but which you are not collecting and thus
on information which is not included in your predictive model.
>Really Rickert is more concerned with something like the following:
>consider manufacturing a complex circuit that will operate in a
>learning system, one that is *not* simply hard-wired to represent, say,
>a reactor's state. Perhaps it will be part of a robot that must explore
>unfamiliar environments. In any case, it is being designed to *evolve*
>transducers for useful properties based on patterns in its low-level
>sensor input.
>Again, however, I see no reason why we can't individuate the
>information content in the final state externally in this case as
>well.
The environment is informationally rich. We might say that there is
an infinite amount of information available in the environment. But
we are finite creatures, and can only access a finite part of that
information. No matter how good is your model, it is only able to
externally individuate a finite quantity of information. I might be
collecting and using information which is not in your model.
> As part of "syntactic psychology" we can say that the formally
>identical learning algorithm might be deployed in one context as in
>another, just as the same AND-gate might be deployed in different
>contexts.
I think yours is really the syntactic psychology, although I expect
that you will disagree. You are committed to a finitely
representable amount of information. If so, it can in principle be
reduced to syntax. I am committed to the availability of an infinite
amount of information, of which I choose a finite sample. I try to
choose that sample in a way that is near optimum for my purposes.
I suspect that your approach might work well for mosquito
psychology. It is likely that the information collecting abilities
of insects are rather more limited than ours, and can reasonably fit
your requirements of external individuation. Personally, I doubt
very much that there is such a thing as mosquito psychology.
Neil Rickert
In <502k3d$r...@usenet.srv.cis.pitt.edu> ande...@pitt.edu (Anders N Weinstein) writes:
>In article <501v36$8...@ux.cs.niu.edu>, Neil Rickert <ric...@cs.niu.edu> wrote:
>>I am not particularly concerned with how "information" is used in
>>cognitive philosophy. My concern is with a notion of "information"
>>which allows me to solve problems. I don't see that as possible with
>>the way you are using the term.
>Evidently you don't think the Marr approach "solves problems". I wonder
>why not.
It solves some problems. It does not solve the problems that
interest me, specifically the problems of learning and problems of
scientific discovery.
>It is still a bit mysterious to me how the one vs the other notion of
>information helps solve a scientific problem.
Then perhaps it is mysterious as to how Copernican notions solved
problems that Ptolemaic notions did not solve. Perhaps it is
mysterious as to how relativity solves problems that Newtonian
mechanics did not solve. Perhaps it is mysterious that science seems
to be a sequence of paradigm shifts. I am trying to provide a basis
which removes the mystery. I need a notion of information suitable
to what I am trying to accomplish.
>>Externalism is a particular approach you are pushing. I'm rejecting
>>it, because it seems clear that it cannot solve the problems that
>>cognitive scientists should be investigating.
>Far from clear, if it is already unreflectively presupposed by most of
>them.
The fact that it is unreflectively presupposed could only argue that
I will have a difficult time selling my approach. I assure you that
I am quite aware of that difficulty. Unreflective presuppositions
have been obstacles that others have had to overcome. There is no
reason I should not make the attempt.
> Perhaps you could say that scientists should be investigating
>things in terms that do not use the notion of information at all
I think they largely do that already, at least in the way they talk
about what they are doing. It has lead to the paradox that a Kuhnian
viewpoint can be taken to suggest that science is not progressive,
while nevertheless it is patently obvious that science has made
enormous progress. I want to describe science in a way which avoids
such a paradox.
> -- that
>I would agree with, the concept of information does not add much
>of scientific value as I see it. It is a way of seeing something in the
>light of a norm.
And that is why Kuhn's work must seem paradoxical to you.
>You keep talking as if there's some terrible problem with externalism, as
>if it were mystical or theological or something like that.
Many of the problems being discussed in epistemology and in
philosophy of mind are not that much different from the problems that
were being discussed by Plato. There is very little evidence of
progress. In science, progress has come with conceptual change. The
old concepts were often stumbling blocks, which hindered progress.
I'm trying to avoid the stumbling blocks by changing to a more
suitable conceptual basis.
>>Biology deals with the physical organism, so needs to consider the
>>physical environment. AI deals with the information processing, so
>>must consider the informational environment.
>Right! But the information environment is basically a feature of the
>organismic niche, and so is externally individuated. Why not? You
>never cite any problem here, even though your unfounded opposition
>seems very strong.
What problems am I supposed to cite? I thought the problems were
well known, and the failure to traditional approaches to solve them
are also well known.
>>In effect, you seem to be implying that cognitive science cannot
>>begin until we have a completed science. That is, cognitive science
>>cannot begin until it has been solved.
>It can begin, it just can't finish. I think this should be obvious. For
>example, someone who doesn't understand quantum mechanics themselves
>simply can't do the epistemology of quantum physicists arguing about
>the interpretation of an experimental result. Nor could this be done by
>someone in ancient times, say.
But then what you are arguing, is that we must describe science in a
way which precludes the possibility of there being a scientific
method within the description. You are saying that the first quantum
physicists could not be described as working with information, since
there were no known quantum particles to provide the content that
would allow them to have information. It is only in historical
perspective that we can see that they were working with information
after all.
>If you lack the concepts, you can't cognize the relevant aspects of the
>environment.
This leads directly to Fodorian nativism. All of the basic concepts
must be innate, for otherwise we could not get started. But if they
are innate, whence did they come. Is there a highly efficient
methodology available to evolution to provide us with these innate
concepts? If so, why can't the neurons follow a similar
methodology? If the neurons can follow such a methodology, then your
argument about concepts must be wrong. If the neurons cannot follow
such a methodology, then neither can evolution. You are left with
them being implanted by a creator. Perhaps the earliest bacterium
(or whatever was the most primitive form of life) already came fully
equipped with the concepts we need.
> So of course
>you couldn't say what information they are picking up if you are,
>as it were, not tuned to their frequency.
What information they are picking up is a secondary concern. They
must start by picking it up. Only then can they try to identify the
information. The information must be there, and recoverable, before
one can begin to talk of content. If you make consideration of
content a prerequisite, then divine interventionism becomes the
possibility.
>Similarly, someone who believes in witches *must* say that the environment
>contains information about witches.
Right. And then we can have another round of Salem witch trials.
All the more reason to be skeptical about the purported value of
beliefs.
> Once you reject the witch theory,
>you must reject that characterization of the information environment. This
>seems obviously mandatory to me.
But that is only because you tie information to characterization.
I'm concerned about uncharacterized information. We can characterize
it later if we wish. We need to start collecting it first. We
should always reserve some skepticism as to any characterization.
>The alternative would be to take cognitive science to play the role
>hoped for in traditional Cartesian epistemology -- that it provides a
>neutral ahistorical framework which presupposes nothing about the world
>and provides the court of appeal for all possible knowledge claims.
That is your alternative, based on your unreflective
presuppositions. Since I am not bound by your presuppositions, I
have no such problem.
> But
>I think this is fantasy,
I agree that it is a fantasy. But that is far more a problem for
your approach than for my approach.
> since all knowledge,
>including brain and cognitive science, rests on presuppositions and
>prejudices which are simply inherited from the culture, which may themselves
>have evolved by accident, although they can come to be incrementally
>revised.
Sure. And those presuppositions were in turn passed down in the DNA
from the earliest organisms created by God. Your argument, if
correct, implies creationism of some form.
>>Sorry, but there is no such thing as "the way content is
>>individuated." You have a particular theory, and "content" and
>>"individuation" are theoretical terms in that theory. There may be
>>such a thing as how content is individuated in your theory. But it
>>is a mistake to treat "concept" and "individuation" as human
>>independent concepts whose relation can be factually determined. The
>>more important question is whether your particular theory is
>>sufficiently extensible to allow it to settle the questions of
>>cognitive science. I am doubtful as to that, so have chosen a
>>different route.
>Fair enough. However, I do think, first, it is the way common-sense
>individuates content,
Common sense is itself dependent on presuppositions derived from the
culture. These can change over time. After all, it was once common
sense that Ptolemaic astronomy was the correct one.
> and, second, that it is necessary if knowledge is
>to be an intelligible possibility.
Not so if one follows Ryle by insisting that knowledge is "knowing
how."
>Are you denying that if someone on Twin-Earth claims that some liquid
>is "water", their claim is made true only if the liquid is (roughly) XYZ
>and not Earthian H20?
Since I have serious problems with traditional theories of truth, I
don't see this as relevant.
If we were able to construct spaceships, and fly to TwinEarth, I
believe the travellers would mostly refer to the TwinEarth liquid as
water. Occasionally, when the distinction became important for a
particular purpose, they might talk about 'TwinEarth water' and
'Earth water'. But for many purposes they would not need such a
distinction, and would not make such a distinction.
> Why do you think we use this practice when evaluating
>people's overt claims?
Normal people, in normal conversation, allow a lot of give and take
(or the principle of charity, if you like). There is usually no
problem in evaluating most claims of most people.
> True enough, it has nothing to do with scientific
>explanation of the noises that come out of people's mouths. Why
>do we individuate inner thoughts in *exactly the same way* as we do overt
>claims?
I don't think that we do individuate inner thoughts in the sense that
you are suggesting, although perhaps philosophers attempt to do so.
>I don't myself think the supposed demands of science are relevant in
>any way to philosophy of mind -- it's a subject like philosophy of
>tables. Wanting to be scientific is always the problem in this area,
>never the solution.
I don't much care whether I solve the problems of philosophy of
mind. I am more concerned with scientific problems.
>>If information content does not make any difference, then it is an
>>irrelevancy and a distraction. That is why I reject its importance.
>>Perhaps our disagreement is mainly one of different interests. My
>>concern is with solving the problems for which content does not
>>matter. You apparently have quite different concerns.
>I bet this is deceptive on your part. But can you explain to me
>what some information is, if you do not use the idea of information
>content? Evidently it cannot be information that one object has gone
>behind another, or that something in the world affords mailing (Gibsonian
>examples) or information that something is an edge (Marr example).
Why can that not be information?
You are giving a content based description of information. I'm not
precluding that. I am simply insisting that the availability of such
a description is not a prerequisite to it being information.
> Can it be expressed in words? How do you talk about
>information without talking about its content?
Gibson's idea of affordance is good enough. That is, a signal is
information if I can use it to make effective decisions. I need not
be concerned with what the signal affords (in the sense in which you
insist on content). If I can use the signal to make decisions in a
way that correlates with my well being, that is enough to demonstrate
that the signal is informative. I can postpone concerns about the
content until later.
>Anyway, if you are not using any notion of representational content, then
>what could your inquiry have to do with knowledge or epistemology or
>philosophy of mind?
If you are asking its relevance to knowledge as traditionally
discussed by epistemologists, then there is probably little. I don't
see why that should matter? If, on the other hand, you take a Rylean
view that emphasizes knowing how, then it should be evident that it
is highly relevant.
>> Our
>>job as cognitive agents is to recover the structure within the
>>information, and use that as our best guide to the structure of the
>>world.
>Why not: recover the information itself? (Not just its structure.)
To a large extent, the information is the structure.
>I would have thought information has no structure at all; it
>is content, not structure.
That is why your position is nativist, and ultimately creationist.
You can't begin until you have the concepts needed to represent
content. Thus you require a creator to provide you with those
concepts. If I am concerned with structure (as correlation), I have
no such problem. Simple signal processing methods can find such
structure without any conceptual prerequisites. Basic evolutionary
processes can already act on correlations between the composition of
the DNA and survival expectations.
>> On a
>>strictly externalist position, we are all zombies, and there could be
>>no such thing as cognitive science.
>How does this follow?
On a strictly externalist position, we are collections of quantum
particles acting under the laws of quantum mechanics. There seems to
be no room for anything but zombies in such an approach.
>> But we can study
>>psychology in terms of information, without having to directly
>>concern ourselves as to the dependencies between the world and the
>>information.
>How, if the information is world-dependent?
I don't see the problem you are suggesting. To take an analogy,
mathematics is used for dealing with information that is world
dependent. But the mathematical procedures are described in a manner
which ignores the world dependent aspects of the information.
> Again this looks to be
>the Fodor view -- the processes deal with structures that bear information
>but they themselves do not deal with information itself, only the structure
>that is correlated with it.
I would not agree with that. The processes deal with the
information. One might perhaps say that the processes are
implemented by physical structures, and the action of the processes
on the information is a consequence of the process implementing
structures acting on the information representing structures.
>>No, I don't agree at all. Take the case of Priestley and Lavoisier.
>>They sought information about phlogiston, and conducted experiments
>>to acquire that information. On your way of looking at things, we
>>would have to say that the content was about phlogiston, and that the
>>content is normative for what is the information. That was
>We certainly have to say that the content of Priestly's judgments were
>putatively concerned with phlogiston. But there is a tricky detail
>here. If the phlogiston theory is false and there isn't really any
>phlogiston, then there isn't any *information* about phlogiston to
>obtain from the environment either.
>On the view to which I incline, this also means that there was no real
>content about phlogiston either -- because phlogiston is a natural kind
>term, the non-existence of phlogiston in the world means that
>Priestly's thoughts weren't even *possibly* true -- because we can't
>be clear on what circumstances would satisfy it.
But then you have to describe the scientists as taking random actions
with no methodological basis. Only in a historical retrospective,
when oxygen is available to fill the role of content, can they be
described as following a method.
I am saying that the reason philosophers of science have so much
trouble with finding a documentable scientific method, one which
meets the type of historical test used by Kuhn, is that they are
insisting that science be described in a manner which precludes the
possibility of their being a scientific method which could fit with
their description.
>>Priestley's position, and he continued to hold to the phlogiston
>>theory. Lavoisier, by contrast, looked at it differently. Lavoisier
>>made the nature of the information normative for determining the
>>content. As a consequence, he determined that the content was really
>>about oxygen, rather than phlogiston.
>We really need to be careful about what it is that is bearing a content here.
>I am not sure exactly what it means to say that Lavoisier and Priestly
>are assigning different contents to the same thing -- *what* exactly?.
>I might say that they had thoughts with different contents, and further
>that Priestly's phlogiston-thoughts turned out in fact to be
>pseudo-contents and did not constitute his ever acquiring any information
>about phlogiston, since there is none in the world to acquire.
Nevertheless, they were dealing with the same information. Both were
heating mercury calx, and finding the weights of the recovered
mercury and the properties of the resulting gas. They both found the
same pattern of correlations. As I am using the term, they dealt
with the same information. Whether the `content' should be described
in terms of phlogiston or of oxygen is not relevant here, except to
the extent that Priestley allowed himself to be confused by such
considerations.
>There are many extremely difficult issues raised here by Kuhn's idea of
>meaning change, e.g. whether oxygen could be taken to be denoted by
>"dephlogisticated air."
Meaning change only comes later. It does not affect the pattern of
correlations observed. Meaning becomes involved when the scientist
must attempt to communicate his results to the larger society.
Marketability in the market place of ideas might actually become an
important consideration.
CDJ
In article <Dwv1...@gpu.utcc.utoronto.ca>,
pin...@gpu.utcc.utoronto.ca (Andrzej Pindor) wrote:
[I said]
>>Ooops. The discussion has moved to another level. _Now_ we are, apparently
>>discussing _not_ justification for some proposition p, which might be from
>>authority, but rather the justification for accepting the authority in the
>>first place.
[He said]
>To evaluate your suggestion one has to have a clear notion what 'from
>authority' means.
Okay, Euclid. (I'm in the middle of my bi-annual reading of Lakatos :)
>As I have attempted to point out depending what 'authority'
>is based on it may either mean guiding oneself by a set of previous
>observations or accepting something as given. In the former case, it is not
>much different from accepting "what goes up must come down" as true on the
>basis of numerous observations. This was basically my point. In the latter
>case, it is an alternative. I would not say "the third alternative" since one
>might argue that using inference is basically refering to the authority of
>the rules of logic. Personally I believe that the authority granted to the
>rules of logic comes from observations (that their use leads to 'true'
>propositions) although they may also be hardwired in our brains (by evolution
>probably).
Not too clear on what any of this means - a fact which only has implications
for me, not its author.
Let me try to be a little more explicit with the line of thought I was
attempting to express, and then maybe I'll be able, via your response(s), to
better understand your concerns.
The proposition Steve holds, which is to be justified, is p. (I don't recall
seeing any Steves in these threads.)
Then, depending on Steve, and the exact nature of his p, I don't see any
problem with countenancing the three following three justificatory responses
that Steve might make when his p is challenged:
(1) I saw (heard, felt, smelt, tasted) X; that's why p.
(2) q; that's why p.
(3) Billy said that p; that's why p.
Traditionally, (3), the authoritative response, has received little attention
in the philosophical literature (else my reading is too impoverished).
Traditionally, (1), the observational response, is thought of as being a
(potentially) acceptable justification for an observable and non-inferential
proposition, while (2), the inferential response, is a (potentially) good
justification for an inferential proposition.
I'm not here concerned to ask after the exclusiveness of these classes
(particularly, following tradition, of the first two).
Surely my (1), (2), and (3) are poorly stated. Hopefully they are clear enough
to bring a certain amount of contrast into relief.
Now, externalism is true - I have hereby laid down the law. (More relevantly,
externalism/internalism has not been an issue thus far.) Therefore, while I
above brought out what I have no problem countenancing as three
justificatory-response forms in terms of Steve's activity of justify_ing_,
there is no need in general to suppose either that Steve go through this
activity, or would even be inclined to, if challenged. We can simply say, on
the assumption that one of these response-forms is in fact appropriate, true,
or whatever, that Steve's p is justifi_ed_. Or something like that.
(3) has a certain "perspectival" nature to it that I'm completely ignoring.
That is, (I think) it isn't so much that Billy told Steve p that counts, but
rather that Steve _takes_ Billy to have told him that p. Or something like
that. Maybe not.
Back to the matter at hand. As I recall, the suggestion was that responses of
form (3) are well thought of as responses of form (2). The grounds for this I
recall being given were something like: that the inference being made is from
the reliability of the source to p.
Now in a certain way, I don't have too much of a problem with this, though the
thought would need to be stated more carefully so as to clearly not infringe
on the Law of Externalism. On the other hand, the thought goes too far: for
the same could just as easily be said if you wanted to assimilate responses of
type _(1)_ to those of type (2) as well. The inference made would be from the
reliability of the source to Steve's p. But I took it that another issue that
was not being contested was that _observational_ claims are (or at least could
be) _non-inferential_.
Maybe you do want to contest this matter; you wouldn't be the first. One does
quickly get into regress problems on this sort of view however - there just
seems to be too much inferring going on!
I guess internalism/externalism _needs_ to be an issue here. Oh well. I think
I'll try to leave that issue to Rickert and Weinstein to hack out.
If we (1) all spoke ancient Greek, and (2) were all fairly familiar with some
Chinese thought, I could have left behind my bathroom humor about Steve's p,
and switched over to Steve's chi (just for variety, not because it's any
funnier).
I think I'll also stand pat on what I said in my previous post about changing
the topic of discussion from the justification of p to the justification of
the justification of p.
>>Hardcoding, indoctrination, and brainwashing, in the sense used above, have
>>nothing to do with justification. The question is of _right_, not of _fact_.
>>
>This may depend on what you mean by 'justification'.
Okay, Euclid.
>I took 'justification'
>as "a reason why we believe someting to be true". Hardcoding etc. may be
>such a reason, might it not? In other words you seem to stop at "the prophet
>said so" as the reason, I am trying to look behind what makes us to accept
>what the prophet said as "true" - is it his record or "just because"?
There's not much more I can do on what the difference between a cause of
belief and a reason for belief is. If you're interested, I think Frege is the
best place to go. The introductory remarks to his Beggriffsschrift and various
places in the Grundlagen spring to my mind immediately (!), but there are
other places as well.
I seem to recall reading realtives of the word "normativity" in some of these
threads. Unfortunately, most of them have been spoken by Weinstein, with whom
nobody seems to agree.
This was a long post. The odds are therefore prohibitively against there being
more than a 30% rate of truth here. Beware.
CDJ
CDJ
Weinstein once wrote:
> Moreover, information already depends on a kind of holism,
>since the information in a signal depends in part on what the receiver
>can extract from it.
And Rickert subsequently once wrote:
>Now you are getting closer to a suitable notion of information. But
>if it depends on what the receiver can extract from it, it becomes a
>subjective notion.
And I believe that Weinstein essentially agrees to there being something
"subjective" in these bushes.
That seems to me to be mistaken. Weinstein's remark can, and should, I say, be
accepted without accepting Rickert's inference.
One way at the matter is to focus on Weinstein's use of "can" in his "can
extract from [a signal]". Rickert apparently takes this use to have the sense
of "is able to". I take Weinstein, from other posts, to agree with this.
If the sense of "can" is taken in this way, then Rickert's inference is
plausible. But it needn't be taken in this way, and I don't think it should.
Rather, "can" ought to be understood as having the sense of "is permissible",
so that Weinstein's remark becomes:
... the information in a signal depends in part on what the receiver
may permissibly extract from it.
And there is no reason to suppsose that the concept of _permissibility of
information extraction_ is a subjective one - at least not in the
"yahoo!!-we-can-do-whatever-we-damn-well-please" sense. Rather, it should be
said that what the receiver _may_ extract will depend on that receiver's other
coommittments - which is the whole point of relativizing to a receiver.
It's not a bad choice of language to call this kind of idea "subjective",
except that is may be easy to confuse it with the other kind. Perhaps it would
therefore be better to call my suggestion "perspectival", rather than
"subjective", to mark out the difference with a word.
Like almost all of my suggestions, the above is not original to me. The way I
stated it is, I believe, original to Brandom.
For readers of Lakatos: I have here attempted to go thru the process of
"moster-adjustment".
CDJ
Anders N Weinstein
In article <505tts$b...@usenet.srv.cis.pitt.edu>, CDJ <cd...@pitt.edu> wrote:
>Weinstein once wrote:
>
>> Moreover, information already depends on a kind of holism,
>>since the information in a signal depends in part on what the receiver
>>can extract from it.
I was a bit loose here. I am prone to stress the distinction between
the sense of "subjective" which means "dependent on whim, caprice,
emotion" or "without a distinction between seeming right and being
right" or "yahoo, we can say whatever we want here" from another sense,
which I would render "subject-dependent" or "subject-involving in
account". There are many objective facts which are in various ways
subject- involving in account, say facts about colors or the value of
dollar bills or the meanings of words or the functions of artifacts.
All are objective in that someone can be wrong about them.
(Though it doesn't make sense to say that *everyone* has always been
wrong about the value of a dollar, the striking peculiarity of the social
-- it is made by Mind, but not by any individual mind.)
The stirling virtue of JJ Gibson's concept of information is that the
information is objectively there in space whether or not it is picked
up, as potentially pick-up-able. This in just the way information
contained in TV broadcasts is out there in space whether or not any
receiver is tuned to pick them up. So, for example, the information
that someone was out at first base might be specified in the light, yet
an infant, or an aborigine, or a person who does not know the rules of
baseball can not detect it. Still it is objectively there.
Gibson did not have quite the epistemologists concern for concepts,
however, and thought of information as non-conceptual, like Rickert.
Anders N Weinstein
In article <hubey.8...@pegasus.montclair.edu>,
H. M. Hubey <hu...@pegasus.montclair.edu> wrote:
>ande...@pitt.edu (Anders N Weinstein) writes:
>
>>you astray. When I am conscious of my desk there is no representation,
>>no signal, that I have applied an interpretation to. If I have interpreted
>
True.
>derives from its function. So it can't be a desk unless and until
>you interpret the signals impinging on your retina (assuming you are
>using your eyes to be conscious of the desk).
Wrong. There can't be desks until a practice of using desks exists.
So there can't be desks unless people standardly *take* certain things as
desks.
But taking something as a desk is not an activity that involves interpreting
retinal signals. Taking something as a desk is a practical act that finds
its primary actuality on the outside, in the world, in, say, sitting
down and working at it. Once we understand that we can go on to form a
conception of merely recognizing something as a desk in thought, without
putting it to use in any way.
> Otherwise you
>would be saying something like "There's an object comprised of
>wood/metal/plastic with a large flat top, and which is supported
>on some vertical beams, ... etc." By calling it a desk you have
>already interpreted it as to its function, and the word that you use
>to describe it is an acknowledgement of such.
Taking something *as* a desk can be said to be a kind of
interpretation. But it is not an interpretation of a retinal signal
which stands as an intermediate object of interpretation. It is an
interpretation of the object in the world itself.
And if the object really is a desk, then it interprets it as what it
is, so no construction or imposition is involved.
Anders N Weinstein
In article <5078ca$c...@ux.cs.niu.edu>, Neil Rickert <ric...@cs.niu.edu> wrote:
>In <5056hs$8...@usenet.srv.cis.pitt.edu> ande...@pitt.edu (Anders N Weinstein) writes:
>
>>This is absurd on my view. I think your representationalism is leading
>>you astray. When I am conscious of my desk there is no representation,
>>no signal, that I have applied an interpretation to.
>
>signal, and you will no longer be visually conscious of the desk.
Of course I agree.
>Of course there are signals.
I didn't say there weren't signals, only that signals aren't interpreted.
>course there are signals. I don't think I actually said that you
>applied an interpretation. I have no problem with saying that your
>perception is direct, unmediated by interpretations. But it is still
>mediated by signals which represent the information.
They carry the information, they don't represent it.
>>We are all non-zombies on my externalist approach.
>
>Sure. But we are non-zombies only there because you say we are. You
>split your universe into two disjoint worlds, one containing the
>physical sciences, and the other containing consciousness. You have
>to keep them separate, because they don't fit together.
This is an exaggeration. The aspects which constitute the worlds of chess
or economics are distinguishable from the physical aspects of
of wood figurines or pieces of paper. Yet the figurines constitute the
chess pieces and the pieces of paper constitute the money. And they are
made of ordinary physical stuff, subject to the laws of physics.
Similarly, perhaps, a living organism constitutes a phenomenological subject,
but is made of physical stuff and also subject to the laws of physics.
The subject is only "in a different world" in that we can distinguish
aspects as in the economic case. And there could be no disembodied subject
of conscious mental states. So consciousness is not really in an autonomous
and independent world, rather a distinguishable aspect of the one world.
Similarly we can distinguish the informational aspect of a signal from
the physical description of the signal, and the information-detection
function from the physical response level of description. But all these
things go on together in one world.
>Well sure. Having split your universe into two disjoint worlds, you
Not exactly disjoint. Living bodies are made of physical stuff.
>have put all of physical reality into one of those worlds. Thus the
>signals and representations have all gone into that physical world.
No. Playing along with the two-worlds idea, the signals are also in both
worlds too -- they have both physical, but also functional or
informational properties.
It is true that if people are described solely in the language of
physics then they show up only as "zombies". So much the worse for
physicalism. It is similarly true that if economic transactions are
described only in the language of physics that value does not show up.
>Now in your other worlds there is only consciousness, perceptual
>experience, etc. In essence you have a physical world in which you
>are a zombie and a phenomenal world in which you are conscious but
>there is no physical reality and thus there are no signals and no
>representations. Since you can't connect your two worlds, I think
>that makes you a closet solipsist.
First, physical reality would be one of the things I can become
conscious of, so it is certainly on part of my "phenomenal world."
Second there are numerous connections between the two worlds you
are saddling me with -- for example, conscious subjects are living
organisms, and have bodies.
>In order to avoid the charge of solipsism, you adopt the ploy of
>pretending that some of the entities in your phenomenal world are
>identical to the entities in the physical world. But to allow that
>identification to carry through, you would need to show that every
>change in the physical world resulted in a corresponding change in
>the phenomenal world. Since you don't allow talk of signals or
>representations, you have no means of carrying through the
>identification.
Don't understand this. I allow talk of signals. I allow talk of
representations in a great variety of ways. I only oppose the idea that
non-conceptual mental representations intervene epistemically between
consciousness and its objects, for example as things that get
interpreted.
Neil Rickert
In <5056hs$8...@usenet.srv.cis.pitt.edu> ande...@pitt.edu (Anders N Weinstein) writes:
>In article <504mbt$a...@ux.cs.niu.edu>, Neil Rickert <ric...@cs.niu.edu> wrote:
>> But something like that is part of Weinstein's
>>outlook.
>Rather, my understanding of the computational view in cognitive
>science. It is not something I endorse myself, I ascribe it to others.
Ok. I don't subscribe to the computation view which is most commonly
held in cognitive science.
>>The reason that the computer cannot have any psychology, is that the
>>human programmer makes the decision as to what is the information the
>>computer shall process. Because of that, the computer is at best
>OK, but what if something evolved into the same state? Would it then
>have a psycholgy?
No, that wouldn't make any difference. In any case, it is
inconceivable that such a system could evolve which had no autonomous
ability to survive.
> Say yes, and you are one kind of externalist, one
>that cites evolutionary history not present state. This is explicit in
>the "teleofunctionalist" Fred Dretske, who denies that intelligence
>could be artificial (though it might be evolved).
>> In that
>>case the ordinary human has no psychology, and is at best described
>>as acting out part of the psychology of the epistemologists. Thus
>>the human is a zombie, has no free will, and is merely a slave to its
>>circumstances.
>This is absurd on my view. I think your representationalism is leading
>you astray. When I am conscious of my desk there is no representation,
>no signal, that I have applied an interpretation to.
If we place a black cloth over your eyes, we will interrupt the
signal, and you will no longer be visually conscious of the desk. Of
course there are signals. I don't think I actually said that you
applied an interpretation. I have no problem with saying that your
perception is direct, unmediated by interpretations. But it is still
mediated by signals which represent the information.
>no signal, that I have applied an interpretation to. If I have interpreted
>anything, it can only be the desk itself. This is why I say that in
>being conscious of a desk as a desk I am taking it as what it is, not
>applying an interpretation to some proximal signal or representation.
>>I see no validity to this description of a human. As an autonomous
>>cognitive agent, I can not only make my own decisions on the
>>information available to me -- I can also make my own decisions as to
>>what components of the signals I collect can be considered
>>information. Thus, no matter how good your science you cannot
>What do you mean you can make decisions about the signals you collect?
I should have been clearer that I was not talking of conscious
decisions. Rather, I was describing the effect of my learning
processes.
>>we are finite creatures, and can only access a finite part of that
>>information. No matter how good is your model, it is only able to
>>externally individuate a finite quantity of information. I might be
>>collecting and using information which is not in your model.
>Indeed. But this would seem to be everybody's problem, not just mine.
>How does *your* theory enable you to do cognitive science for a subject
>that is responding to information which you the theorist do not
>detect?
I make no claims of being able to have a predictive model for another
person. The idea that we might someday have such models is a pipe
dream. But just as biology can be a science without being able to
predict what species will evolve next, we can have cognitive science
even if it cannot predict individual behavior.
>>I think yours is really the syntactic psychology, although I expect
>>that you will disagree. You are committed to a finitely
>I believe that computational modelling might well be taken to be
>committed to such a thing. You are the one who is constantly denying
>a role for semantics = content.
No. I am denying information = content. I said no such thing about
semantics. Actually, I tend to think of `semantics' as a theoretical
term in a not-very-good theory of natural language.
>>Likewise, I doubt that there could be non-zombies in your externalist
>>approach.
>We are all non-zombies on my externalist approach.
Sure. But we are non-zombies only there because you say we are. You
split your universe into two disjoint worlds, one containing the
physical sciences, and the other containing consciousness. You have
to keep them separate, because they don't fit together.
> Only if you think
>we are really locked up in a room "interpreting signals", so that
>consciousness of the world is mediated by representations, would you think
>something otherwise. But in my view there are no representations
>mediating between consciousness and its objects, and so nothing to
>interpret but worldly objects themselves.
Well sure. Having split your universe into two disjoint worlds, you
have put all of physical reality into one of those worlds. Thus the
signals and representations have all gone into that physical world.
Now in your other worlds there is only consciousness, perceptual
experience, etc. In essence you have a physical world in which you
are a zombie and a phenomenal world in which you are conscious but
there is no physical reality and thus there are no signals and no
representations. Since you can't connect your two worlds, I think
that makes you a closet solipsist.
In order to avoid the charge of solipsism, you adopt the ploy of
Andrzej Pindor
In article <502k3d$r...@usenet.srv.cis.pitt.edu>,
Anders N Weinstein <ande...@pitt.edu> wrote:
......
>Are you denying that if someone on Twin-Earth claims that some liquid
>is "water", their claim is made true only if the liquid is (roughly) XYZ
>and not Earthian H20? Why do you think we use this practice when evaluating
This claim only makes sense from an external point of view (exactly the God's
eye view) which allows one to see that the "water" on TwinEarth is XYZ and
"water" on Earth is H2O. If no such point of view exists, then it does not
make any sense. All the time you are making such assumption and then
vehemently deny it. If only points of view are the one heled by inhabitants
of Earth and the other one held by the TwinEarthers, you simply cannot say
the above. Please tell us how (in absence of the external, or God's eye view)
would anyone in the story ever knew if Earth "water" and TwinEarth "water"
were the same or different substances? By all the criteria used by the
participants in the story they would be the same and in tha lack of other
criteria (no God's eye view) it makes no sense at all even consider that
they might be different. One can of course talk about it but such noises
have as much content as a discussion whether a difference which makes no
difference is a difference.
[Neil Rickert]
>>Now you are getting closer to a suitable notion of information. But
>>if it depends on what the receiver can extract from it, it becomes a
>>subjective notion.
>
>I have already said I think there is nothing "subjective" about it.
>Perhaps subject-dependent, but that applies to many objective things,
>arguably everything.
>
One of the problem with your statements is that you use the words in a way
which is in contradiction with their usual definitions. I have always felt
that people understand the word "subjective" as "subject-dependent". To
make sure that I am not deluded I've just checked my on-line Webster. This
is what it says:
sub-jec-tive adj. 1. a. Proceeding from or taking place within an
individual's mind such as to be unaffected by the external world. b.
Particular to a given individual; personal: subjective experience.
2. Moodily introspective. 3. Existing only in the mind; illusory. 4.
Psychol. Existing only within the experiencer's mind and incapable
of external verification.
Could you explain where did you get your version of "subjective" from?
No surprise it is hard to understand what you claim to be saying.
[Neil Rickert]
>>You are changing the story. I did insist that TwinEarth be
>>indistinguishable from Earth, which implies that the water on
>>TwinEarth should be indistinguishable from the water on Earth and
>>thus be equally satisfying to Oscar and to TwinOscar.
>
>It might be subjectively satisfying. I meant that water is not in the
>"conditions of satisfaction" of his intentional state in Searle's
>sense. If TwinOscar asks for water and you give him H20, he may be
>subjectively satisfied but he did not get what he wanted. (This is not
>mysterious, it happens all the time -- if I ask for hamburger and you
>give me a soy imitation, I might be none the wiser and wind up content
>nonetheless. But the content of my original request was certainly not
>satisfied.)
>
Another example where you assume the god's eye view without realising it.
If Oscar is unable (in principle) to tell that he was given H2O and not
XYZ, he got what he wanted. If you disagree, please tell me how do you know
that he wanted XYZ? When he said "I want water" he meant a substance with
certain properties he knows about (or can at least possibly know about) and
he got a substance with such properties, did he not? You can only say that
it was a different substance if you know that on all previous occasions he
got XYZ and now he got H2O, but such knowledge assumes an external point of
view (the God's eye view).
The same applies to the hamburger case - you could only say that your original
request was not satisfied if you assumed a point of view from which it was
possible to distinguish between the beef hamburger or the soya one.
Without such a point of view (the God's eye view) the claim is meaningless.
Anders N Weinstein
In article <Dwyqz...@gpu.utcc.utoronto.ca>,
Andrzej Pindor <pin...@gpu.utcc.utoronto.ca> wrote:
>In article <502k3d$r...@usenet.srv.cis.pitt.edu>,
>Anders N Weinstein <ande...@pitt.edu> wrote:
>......
>>Are you denying that if someone on Twin-Earth claims that some liquid
>>is "water", their claim is made true only if the liquid is (roughly) XYZ
>>and not Earthian H20? Why do you think we use this practice when evaluating
>
>This claim only makes sense from an external point of view (exactly the God's
>eye view) which allows one to see that the "water" on TwinEarth is XYZ and
>"water" on Earth is H2O. If no such point of view exists, then it does not
The hypothetical story was told from the vantage point of present day
scientists. We discovered that water is H20 from the ordinary human point
of view. The inhabitants of the story could discover as much from their
point of view too.
In another sense many counterfactuals necessarily assume something like
a god's eye view. For example, I might ask whethere there was an eclipse
on a certain date when no conscious observer was around to see it, or what
would have happened on that day if such and such occurred. Only a Berklean
idealist objects that such stories require an objectionable god's eye view.
All the time you are making such assumption and then
>vehemently deny it. If only points of view are the one heled by inhabitants
>of Earth and the other one held by the TwinEarthers, you simply cannot say
>the above. Please tell us how (in absence of the external, or God's eye view)
>would anyone in the story ever knew if Earth "water" and TwinEarth "water"
>were the same or different substances? By all the criteria used by the
By doing science, silly. (In fact, as Putnam told the first part of the
story, scientists on Earth and Twin Earth already knew the difference.)
>One of the problem with your statements is that you use the words in a way
>which is in contradiction with their usual definitions. I have always felt
>that people understand the word "subjective" as "subject-dependent". To
It depends on the way in which things are subject dependent. The predicate
"x is looked at by someone" is subject-dependent. But it is an objective
fact whether it applies -- thinking it is so doesn't make it so.
>make sure that I am not deluded I've just checked my on-line Webster. This
>is what it says:
>sub-jec-tive adj. 1. a. Proceeding from or taking place within an
> individual's mind such as to be unaffected by the external world. b.
> Particular to a given individual; personal: subjective experience.
>2. Moodily introspective. 3. Existing only in the mind; illusory. 4.
> Psychol. Existing only within the experiencer's mind and incapable
> of external verification.
But my definition of "subjective" is pretty much this. In this sense,
Gibson's information is not "subjective". Whether the information that
something affords mailing is in the light is not any of the above
things, on Gibson's view.
>Could you explain where did you get your version of "subjective" from?
I get it mainly from Frege, Grundlagen der Arithmetik. There is a famous
passage where Frege insists that judgments about the North Sea are objective,
even though a conventional determination is required to define its
boundaries. I could point you to several stimulating discussions of
Frege's concept of objectivity, if you were interested.
However, I have also explained several times that I take an objective matter
of fact to be something such that it makes sense to say: "John takes it to
be the case that p but he is wrong". In this sense the objective is
something subject to a rightness or wrongness that is not up to us, so that
we have no freedom to think whatever we want on the matter.
Put another way, a subject matter is objective if judgments concerning
its objects are rationally constrained, so you can't say whatever you
think on the matter.
Of course objectivity in this sense is much wider than physical science,
or the human-independent world. One of the important consequences for
me is that science does not exhaust the objective.
>Another example where you assume the god's eye view without realising it.
>If Oscar is unable (in principle) to tell that he was given H2O and not
>XYZ, he got what he wanted. If you disagree, please tell me how do you know
>that he wanted XYZ?
First, he might be able to learn what the difference is and so tell.
But really whether he can tell is irrelevant, he is not the authority on
the nature of the stuff he wants. He *said* he wanted XYZ and if we wanted
to find out what this is we would have to look to the environment.
Even in the case where Oscar *can* tell by tests, he might be fooled and
wind up satisfied by something he didn't ask for. The distinction between
being satisfied and actually getting what you asked for is quite independent.
>certain properties he knows about (or can at least possibly know about) and
>he got a substance with such properties, did he not? You can only say that
>it was a different substance if you know that on all previous occasions he
>got XYZ and now he got H2O, but such knowledge assumes an external point of
>view (the God's eye view).
No, just the ordinary point of view.
>The same applies to the hamburger case - you could only say that your original
>request was not satisfied if you assumed a point of view from which it was
>possible to distinguish between the beef hamburger or the soya one.
But that is possible. And if its not now possible, it might yet become
possible.
Neil Rickert
In <507aaf$h...@usenet.srv.cis.pitt.edu> ande...@pitt.edu (Anders N Weinstein) writes:
>>>If one makes this distinction, you are faced with the
>>>problem that we have no access the transcendental process by which the
>>>phenomenal world is constructed.
>>Why must we assume it is constructed by a transcendental process?
>>What is wrong with information processing.
>Just that all knowledge of information processing is a posteriori,
>empirical knowledge, studying something in the phenomenal world.
>Nothing empirically discoverable can play the role of Kant's
>transcendental philosophy, which would have to be a priori, since it is
>concerned with the formal features of any possible human knowledge.
This is something you see as a problem, only because you are
concerned about "knowing that". I have no such concern.
>>I don't see that as a problem. If brains are mere phenomenal
>>objects, then they are objects we construct to model the information
>>processing. We can still examine the information processing, whether
>>or not we consider the brains to be real.
>Then it sounds like you are taking 'information processing" to be
>something that is a purely a priori or formal concept in Kant's sense,
>something we might know from the Cartesian standpoint of suspended
>belief in everything empirical.
Not at all. I'm really not concerned with whether the information
processing is conscious.
> So that even if we were conscious
>subjects in vats, and brains were unreal, we would still know that
>there was information in the way, according to Descartes, we can know
>the contents of our own minds.
None of this really matters.
From my perspective, a scientific theory is something like a formal
model which fits the evidence reasonably well. If you want to say
that my `information' is a theoretical term, that's good enough for
me. I am not requiring any conscious awareness, which is why I don't
care for your insistence on content. I am trying to provide a basis
for explaining neural activity.
>>I don't suppose that. Perhaps one does not have content without
>>concepts, particularly if one defines content in terms of concepts.
>>But I think the role of concepts is quite different from what is
>>usually assumed. In the case of the scientist, I would argue that
>>the scientist makes his basic discovery without any necessary
>>reliance on concepts. He or she must then construct concepts
>>suitable for making the discovery available to other scientists.
>>Thus I see concepts as necessary for the social aspects of science,
>>rather than for the basic discoveries.
>So the basic discovery is solipsistic and incommunicable?
This does not follow. The most basic discoveries may be made by the
scientist's neurons, and I suppose you could say that they are
incommunicable. The scientist may then become aware of some
peculiarity in what he observes, although perhaps not a peculiarity
that he can verbally describe. At that stage the scientist may be
able to use methods such as pointing, to place other scientists in a
position where their neurons are likely to make the corresponding
discovery. I think that amounts to a form of communication, but one
that does not require the information to have been conceptualized.
> What happens
>when we give words to preconceptual information, in your view? Does that
>distort it, so that the real information is known only to the subject?o
I don't understand the question. The words might be additional
information, but they could not distort the original information.
>My fundamental idea is that one should define thought content or
>information as what is communicable.
Again, my primary concern is not with thought content. You probably
should take me as somewhat of a behaviorist. Although Longley
repeatedly accuses me of the sins of inten{t,s}ionality and
cognitivism, you need not assume he is correct.
>Couldn't there be a man who could play the science game but had no
>preconceptual information inside of him? Still he made the same public
>claims that others did?
A random number machine might write Hamlet, too.
>>I agree that the process is not an intellectual one, but I do so for
>>rather different reasons. I would rather say that knowledge has a
>>substantial non-conceptual basis, and therefore the process could not
>>be an intellectual one.
>I agree there is a non-conceptual background which is prerequisite.
You should take me as using `knowledge' mainly to refer to that
prerequisite background.
My experience as an educator is that if I teach in such a way as to
best help the students gain the non-conceptual background, the
beliefs take care of themselves. However if instead I concentrate on
trying to communicate beliefs to the students, many of them do very
poorly even in acquiring those exact beliefs. I take this as
evidence that the "justified true belief" characterization of
knowledge is little more than part of a "Just So" story.
>>I don't see any need for the brain to have a language. It is
>>excessive intellectualizing to assume that a language is required.
>OK but you say the neurons have data. What is this data? When you
>describe it that way you use language and so ascribe some innate
>representational capacity.
When a neuron transmits a signal, that signal is data. Yes, I'll
grant that the neuron has the innate ability to transmit signals. I
suppose in that sense it could be said to have an innate
representational capacity. But all the neural signal is innately
capable of representing is the neural signal itself.
> What is innate in a normal human being is the ability to acquire
>abilities, most importantly linguistic ability.
It is the ability to acquire abilities that needs to be explained.
You want to brush it aside, take it for granted, and as a consequence
leave everthing important unexplained. I don't agree that linguistic
ability is most important. Surely the ability to feed oneself is of
greater importance.
>But what needs to be explained and how is not at all obvious to me. In
>many ways we already understand this process. For example, I know what
>tables, chairs, knives, forks, dinner-times are because I was raised in
>a culture in which these things have a function, which I came to master
>(more or less) myself.
Ah, yes. And I know mathematics because I took many mathematics
classes. Learning theory is so simple when you explain it that way.
But haven't you just explained everything away? And you wonder why I
accuse you of making huge innatist assumptions.
> No doubt stuff has to happen in my brain when
>these skills rubbed off on me, but that's not the interesting level as
>far as I am concerned.
Then why don't we just eat, drink, and be merry? We can always say
that everything else is not at the interesting level.
> And if you want to know how the culture came to
>start using these things you have cultural history to explain it.
That doesn't explain anything. If you take a bunch of mosquitoes,
and put them together, they will not form a culture. If you could
take a bunch of children, and had them raised together, they would
form a culture even though there had been no cultural history for
them to rely on.
>In the case of language use there is more to be said about what
>abilities with words constitute linguistic competence. Quine and
>Sellars, for example, give us a bit here. But I am not seeing what
>is unexplained, if we accept ordinary descriptions as explanations,
>as we must in this domain.
What is the biological role of language? How does it perform that
role? How is language acquired?
>>I think there is something to that, although it may be exaggerated.
>>But one cannot credit the shaping exclusively to the culture. Even
>>if one could, one would have to explain the mechanism that makes such
>>shaping possible.
>Why? I think the mechanism merely a causal prerequisite for the epistemic
>process we *ought* to study, not an epistemic process itself.
It really should be obvious to you why we are not communicating very
well. You want to sweep aside, as not needing explanation, all of
the things I consider it important to explain. Instead you are more
interested in explaining the things I consider to be little more than
arguments about angels on heads of pins.
>>Again, I think `truth' is the wrong standard. This is why Kuhn found
>I don't really need any substantial theory of truth like the correspondance
>theory. Frege argued that truth was necessarily primitive and indefinable, yet
>held a realist theory according to which judgment aimed at objective
>truth. Davidson perhaps has a similar view today, although I find the details
>confusing.
If your theory of the world depends on truth as a standard of its
validity, and if you insist that truth is indefinable, then you have
no theory of the world.
>Anyway, all I need is the internal connection between meaning and truth. If
>you say that this signal bears information that something affords mailing,
>that relates it to a state of affairs in the world, that lays down
>conditions of satisfaction for it -- it is true if and only if the thing
>really does afford mailing. This is neutral as between different theories
>of truth. Do you deny it?
I don't deny it. But I insist that the standard you are giving is
only applicable to evaluate the suitability of my description of the
signal. It could not be a standard for the signal itself.
>I am really still trying to understand how you can talk about information
>without using the notion of content. Examples would be the best answer,
>rather than generalities -- what are some examples of information that
>is processed, according to you?
If two signals are correlated, each signal provides partial
information as to the state of the other signal. These can be neural
signals of which I am not conscious, so I don't see how content could
be assigned.
If we live in an information rich environment, then it is expected
that several of our sensory neurons will pick up some of the same
information from the environment. That would lead to such a
correlation between neural signals. Thus whenever two neurons carry
correlated signals originating from external sources, there is a good
likelihood that the correlation indicates that the common part of the
signals is information. This is information that brain processes
might be able to detect without requiring any assumptions as to the
content of that information.
>OK, again, I am struggling to understand what could be NR-information,
>and why you resist the idea that it is externally determined. Examples
>would be the best answer to my puzzlement.
Being a realist, I assume that the information has an external
source. In that sense you could say that the information is
externally determined. But that is not how you are using your
externalism. You want to say that knowledge of what is external is
what allows us to know that the signal bears information. I think
that is backward. Rather, it is the information in the signal
(whether or not we can be said to `know' that there is such
information) which allows us to have knowledge of what is external.
>But there is no data for the brain at birth, and so no "logical problem"
>of language acquisition.
There are billions of sensory receptors feeding signals into the
brain. How could it have no data at birth?
>>As I have said, concepts only become necessary when we need to
>>communicate our discoveries to others. The scientist may have to do
>>a lot more work before he is ready to form concepts. But he is
>>guided by the outcome of the simple signal processing.
>This might result in skills for dealing with the world practically, but I have
>trouble understanding how it can be knowledge if it is incommunicable.
This is the problem you have for having fallen for the
intellectualist legend.
>Do you really think Einstein had the information about special relativity
>detected in his brain *prior* to his ability to articulate a theory and put it
>forward in symbols?
No, of course not. But then I don't count the theory of special
relativity as being information either. It is analytic, so has no
informational content. (Well, perhaps it has social information as
to the form of scientific laws used by this society.) The role of
theories such as special relativity is not to carry information, but
to set standards for the information we gather in our observations.
> Information
>without articulation can play a practical role in guiding behavior, but
>I don't see how it can count as scientific knowledge.
I don't count "information without articulation" as scientific
knowledge. I don't think scientific knowledge is information at
all. Measurements made by scientists are information. But the
scientific knowledge is not the information. Rather, the knowledge
is the ability to carry out the processes which acquire information.
>Again, this is an area where we are talking about relations between one
>sort of conceptual content and another, so I have no objection.
>Indeed, but I am not sure what the point of the analogy is, given that
>CDs are mindless. Are you suggesting that Lavoisier's brain developed
>the capacity to transduce or amplify information about oxygen,
>mindlessly as it were (without conscious reasoning or conceptual
>activity), and only later did Lavoisier develop the concepts needed to
>put it into words and communicate some of this information to his fellows?
Roughly, yes. I would rather say that the brain's learning ability
is an ability to concentrate information. If it becomes sufficiently
concentrated, then conscious awareness can arise.
>This strikes me as absurd, given the constant interplay between
>articulation and discovery in the process.
History only reports that which is articulated. You must be cautious
in trying to use historical evidence to show that there is no
discovery without articulation.
Anders N Weinstein
In article <505lec$b...@ux.cs.niu.edu>, Neil Rickert <ric...@cs.niu.edu> wrote:
>In <505533$8...@usenet.srv.cis.pitt.edu> ande...@pitt.edu (Anders N Weinstein) writes:
>
> . I'll try to look for
>some good examples in future posts. I'm really considering
>information at a primitive level, before it is conceptualized, and
>that makes it hard to describe in terms of our standard concepts.
And I'm really trying to understand what this "information" is.
>>If one makes this distinction, you are faced with the
>>problem that we have no access the transcendental process by which the
>>phenomenal world is constructed.
>
>Why must we assume it is constructed by a transcendental process?
>What is wrong with information processing.
Just that all knowledge of information processing is a posteriori,
empirical knowledge, studying something in the phenomenal world.
Nothing empirically discoverable can play the role of Kant's
transcendental philosophy, which would have to be a priori, since it is
concerned with the formal features of any possible human knowledge.
>I don't see that as a problem. If brains are mere phenomenal
>objects, then they are objects we construct to model the information
>processing. We can still examine the information processing, whether
>or not we consider the brains to be real.
Then it sounds like you are taking 'information processing" to be
something that is a purely a priori or formal concept in Kant's sense,
something we might know from the Cartesian standpoint of suspended
belief in everything empirical. So that even if we were conscious
subjects in vats, and brains were unreal, we would still know that
there was information in the way, according to Descartes, we can know
the contents of our own minds.
I am dubious how this can be, however. A lot of your examples presuppose
the world.
>I don't suppose that. Perhaps one does not have content without
>concepts, particularly if one defines content in terms of concepts.
>But I think the role of concepts is quite different from what is
>usually assumed. In the case of the scientist, I would argue that
>the scientist makes his basic discovery without any necessary
>reliance on concepts. He or she must then construct concepts
>suitable for making the discovery available to other scientists.
>Thus I see concepts as necessary for the social aspects of science,
>rather than for the basic discoveries.
So the basic discovery is solipsistic and incommunicable? What happens
when we give words to preconceptual information, in your view? Does that
distort it, so that the real information is known only to the subject?o
My fundamental idea is that one should define thought content or
information as what is communicable. What is incommunicable drops out
like Wittgenstein's beetle in the box -- it has no role to play in the
social process of justifying and arguing about the validity of claims.
Couldn't there be a man who could play the science game but had no
preconceptual information inside of him? Still he made the same public
claims that others did?
>I agree that the process is not an intellectual one, but I do so for
>rather different reasons. I would rather say that knowledge has a
>substantial non-conceptual basis, and therefore the process could not
>be an intellectual one.
I agree there is a non-conceptual background which is prerequisite.
>>I think it is the intellectualist legend to try to explain the
>>emergence of conceptual capacities on the model of a scientist forming
>>theories about antecedently available data. (You must think that at
>>least the data language of the brain is innate, where I don't think a
>>brain can have a language)
>
>I don't see any need for the brain to have a language. It is
>excessive intellectualizing to assume that a language is required.
OK but you say the neurons have data. What is this data? When you
describe it that way you use language and so ascribe some innate
representational capacity.
>>>Sure. And those presuppositions were in turn passed down in the DNA
>>>from the earliest organisms created by God. Your argument, if
>>>correct, implies creationism of some form.
>
>>I said they came from the culture, not the DNA. DNA is irrelevant to
>>this level of description.
>
>But to say so just brushes the problems aside. You still have to
>explain how they came to be in the culture, and how the person
>acquires them from the culture. I don't see that this move settles
>any of the problems.
Well it does avoid the absurd innatism or theological views you ascribe
to me. What is innate in a normal human being is the ability to acquire
abilities, most importantly linguistic ability.
But what needs to be explained and how is not at all obvious to me. In
many ways we already understand this process. For example, I know what
tables, chairs, knives, forks, dinner-times are because I was raised in
a culture in which these things have a function, which I came to master
(more or less) myself. No doubt stuff has to happen in my brain when
these skills rubbed off on me, but that's not the interesting level as
far as I am concerned. And if you want to know how the culture came to
start using these things you have cultural history to explain it.
In the case of language use there is more to be said about what
abilities with words constitute linguistic competence. Quine and
Sellars, for example, give us a bit here. But I am not seeing what
is unexplained, if we accept ordinary descriptions as explanations,
as we must in this domain.
>>Where do I commit myself to innatism? More extreme culturalists think
>>of the human being prior to socialization as an almost completely
>>malleable piece of putty, one that can be shaped in almost any way
>>imaginable by upbringing.
>
>I think there is something to that, although it may be exaggerated.
>But one cannot credit the shaping exclusively to the culture. Even
>if one could, one would have to explain the mechanism that makes such
>shaping possible.
Why? I think the mechanism merely a causal prerequisite for the epistemic
process we *ought* to study, not an epistemic process itself.
>>Because to describe it that way relates it to the world.
>
>But that might be an artifact of the way you are describing it, and
>thus not by itself decisive as to whether it is information.
Show me another way to describe it, without using words that have meanings
that relate to things in the public world.
>>Because to describe it that way relates it to the world. It characterizes
>>it in a way such that it is true if and only if a particular worldly
>>object really does afford mailing or not, say.
>
>Again, I think `truth' is the wrong standard. This is why Kuhn found
I don't really need any substantial theory of truth like the correspondance
theory. Frege argued that truth was necessarily primitive and indefinable, yet
held a realist theory according to which judgment aimed at objective
truth. Davidson perhaps has a similar view today, although I find the details
confusing.
Anyway, all I need is the internal connection between meaning and truth. If
you say that this signal bears information that something affords mailing,
that relates it to a state of affairs in the world, that lays down
conditions of satisfaction for it -- it is true if and only if the thing
really does afford mailing. This is neutral as between different theories
of truth. Do you deny it?
I am really still trying to understand how you can talk about information
without using the notion of content. Examples would be the best answer,
rather than generalities -- what are some examples of information that
is processed, according to you?
>Think of the affordance as an opportunity for action. I don't fully
>accept Gibson's view about distal objects, for there are times that
>we have information but have not yet identified the source of that
>information.
But that's OK, that just involves different, somewhat weaker objective
content, as I might obtain the information by smell that "there's at least
one dog in the room" without knowing the source or seeing any dogs.
>Not at all. I do not need to ascribe any content to the signal. That
>may mean that I do not ascribe any AW-information. That has nothing
>to do with whether there is NR-information. Admittedly it will often
OK, again, I am struggling to understand what could be NR-information,
and why you resist the idea that it is externally determined. Examples
would be the best answer to my puzzlement.
>The scientist will use that NR-information as a basis for
>experiments. As a result of those experiments, the scientist will
Again, scientists have conceptualized intentional contents which refer
to observations, so I have no problem with this. We don't need non-conceptual
content to account for the relation between observation sentences and
theoretical sentences. We can do this all with AW-content, not NR-information.
And we can be externalists, since scientific observation is something that
involves the world with instruments and other people, not only what's
inside the skull.
>You have not explained how you can be conditioned to use symbols. I
>don't see that problem as any easier than the problem of acquiring
>knowledge about the world. If the second is impossible without
>prerequisites, then so is the first.
What needs to be explained here and in what terms? The language rubbed
off on me like a viral behavior pattern through my exposure to other
speakers. Human beings are innately disposed to have such patterns
impressed upon them, in a way that cats are not. No doubt brain structures
plays some role in making this possible, but I don't need to worry about
that for any explanatory purpose.
>I'm reading it. But talk of it coming from the culture is easy
>talk. Chomsky certainly has made strong arguments that we cannot be
>conditioned suitably with the impoverished evidence available.
I don't think they are strong. They depend entirely on the intellectualist
legend, the bogus idea that the brain is like a scientist presented with
data propositions and faced with the task of inducing a metatheory for
the local language. From this perspective, only innate knowledge could
take up the slack between data and theory.
But there is no data for the brain at birth, and so no "logical problem"
of language acquisition. Because this is not an intellectual process,
Chomsky's conviction that innate reasoning is required can be rejected.
>As I have said, concepts only become necessary when we need to
>communicate our discoveries to others. The scientist may have to do
>a lot more work before he is ready to form concepts. But he is
>guided by the outcome of the simple signal processing.
This might result in skills for dealing with the world practically, but I have
trouble understanding how it can be knowledge if it is incommunicable.
Do you really think Einstein had the information about special relativity
detected in his brain *prior* to his ability to articulate a theory and put it
forward in symbols? Even if something like this were true, I think it
would be irrelevant to the epistemology of the special theory. Information
without articulation can play a practical role in guiding behavior, but
I don't see how it can count as scientific knowledge.
>It doesn't trouble me that on many occasions the content is
>externally determined. But we cannot explain scientific discovery if
>we have to know what it is we are going to discover before we
>discover it. If you make the content a prerequisite, I think you
>have precluded the possibility of explaining scientific discovery.
Again, this is an area where we are talking about relations between one
sort of conceptual content and another, so I have no objection.
>I have a CD player. I place a CD into it, press the button, and
>music comes out. In the music I can recognize notes, chords, pauses,
>tempos, etc. The CD player operates according to a methodology. But
>the concepts of notes, chords, pauses, tempos, etc, are not concepts
>used in that methodology. Since our usual ideas of rationality would
>require the rational processes to use those concepts, the CD player
>fails the "systematic rational norm" test. Nevertheless you might
>say that the CD can be described as working algorithmically with the
>fragmentary digital samples used to record the CD.
Indeed, but I am not sure what the point of the analogy is, given that
CDs are mindless. Are you suggesting that Lavoisier's brain developed
the capacity to transduce or amplify information about oxygen,
mindlessly as it were (without conscious reasoning or conceptual
activity), and only later did Lavoisier develop the concepts needed to
put it into words and communicate some of this information to his fellows?
Anders N Weinstein
In article <507nec$d...@ux.cs.niu.edu>, Neil Rickert <ric...@cs.niu.edu> wrote:
>In <507de2$h...@usenet.srv.cis.pitt.edu> ande...@pitt.edu (Anders N Weinstein) writes:
>>> I don't think I actually said that you
>>>applied an interpretation. I have no problem with saying that your
>>>perception is direct, unmediated by interpretations. But it is still
>>>mediated by signals which represent the information.
>
>>They carry the information, they don't represent it.
>
>representations as you are, and you should read me as talking about
>information carrying rather than information representation.
Good. But the real issue is not the terminology, it is the supposed
analogy between what a scientist does with datal contents and
theoretical contents -- where there is in a sense something to
interpret, namely the data -- and what the brain does in initial
concept development, which does *not* involve interpreting anything
"data" like. Rather evolving detectors for that same information, which
is a non-representational process. After the detectors have evolved one
might of course speak of the information they detect as data.
I want the terminology to suggest that proximal patterns of ambient
light "carry information" whether or not anyone detects it. But the
patterns are not themselves data, rather one can only talk of data
after a detector for the information has evolved (and even after that,
only once this detector functions to implement the input side to a
concept-using, inference making system).
This is in contrast to the scientific case, where something can fairly
be said to be interpreted, since the data there is conceptualized.
Neil Rickert
In <507de2$h...@usenet.srv.cis.pitt.edu> ande...@pitt.edu (Anders N Weinstein) writes:
>In article <5078ca$c...@ux.cs.niu.edu>, Neil Rickert <ric...@cs.niu.edu> wrote:
>>>This is absurd on my view. I think your representationalism is leading
>>>you astray. When I am conscious of my desk there is no representation,
>>>no signal, that I have applied an interpretation to.
>>If we place a black cloth over your eyes, we will interrupt the
>>signal, and you will no longer be visually conscious of the desk.
>Of course I agree.
>>Of course there are signals.
>I didn't say there weren't signals, only that signals aren't interpreted.
>> I don't think I actually said that you
>>applied an interpretation. I have no problem with saying that your
>>perception is direct, unmediated by interpretations. But it is still
>>mediated by signals which represent the information.
>They carry the information, they don't represent it.
Ok. If you want to make that distinction, then I am as much against
H. M. Hubey
ande...@pitt.edu (Anders N Weinstein) writes:
About desks..
>>derives from its function. So it can't be a desk unless and until
>>you interpret the signals impinging on your retina (assuming you are
>>using your eyes to be conscious of the desk).
>Wrong. There can't be desks until a practice of using desks exists.
>So there can't be desks unless people standardly *take* certain things as
>desks.
Presumably desks were made to serve certain functions. What I say still
stands. When you use the word desk you are already interpreting
signals that impinge on your retina.
>But taking something as a desk is not an activity that involves interpreting
>retinal signals. Taking something as a desk is a practical act that finds
What? How do you interpret signals?
Let's try to clarify a few things. Off the top of my head, from
Stallings (book Communications) we have (as a simplified version)
computer human
-------- ------
i(t)=info to be sent ASCII,picture,sound etc In the brain/mind
Meaning is here,and
not necessarily
Fodor's LOT.
d(t)= data to be sent i(t) converted to Ditto-speech or
suitable form writing usually
m(t) == message sent d(t)changed into suitable actual production
form for transmission of speech or wrtng
m'(t)= message received not same as m(t) because Ditto -wrong words
of noise might be used
d'(t)=data received recovered from m'(t) by some Sounds processed
kind of processing into words, or
writing processed
into its acoustic
equivalent
i'(t)=information received more processing; if done could fail in
correctly this is i(t) humans, due to
wrong choice of
words.
Now, I ignored vision to make it simple. Obviously when the message is
in writing vision is at work. In the case above (of the desk) the
problem is less formal, so the fact that you can say you see a desk
involves similar processing except that there is no active agent
sending a message. The information is produced passively by the "desk"
purely by virtue of its material existence and its capability of
reflecting light. There are certainly passive agents that can alter
messages/data. For example copper cable can induce changes in the
electrical signals that it carries.
>its primary actuality on the outside, in the world, in, say, sitting
>down and working at it. Once we understand that we can go on to form a
>conception of merely recognizing something as a desk in thought, without
>putting it to use in any way.
We are still discussing information,data, message, and processing,
are we not? To make matters more complex, we now have an additional
problem of interpretation for humans, and we have "passive"
message sending by objects in the universe. What I outlined above is
a simple case since I only stretched computer communication to
make analogies to human communication and perception.
>Taking something *as* a desk can be said to be a kind of
>interpretation. But it is not an interpretation of a retinal signal
>which stands as an intermediate object of interpretation. It is an
>interpretation of the object in the world itself.
We always interpret what our senses tell us. So it doesn't matter
usually. Where it does matter is if something goes wrong with
our perception organs or processing organs. It is at this point that
the problems of noise matter. If there is no problem with the perceiving
organs nor the processing, then the interpretation of the signal or
the interpretation of the object are equivalent. But to be sure, it
is the signals we interpret. That is how we can come to the conclusion
(without knowledge of physics) that it is colder outside on windy
days even if the thermometer registers the same. With the attitude
I take above, the anomolies are also accounted for not only
correctly functioning perception/interpretation processing of
sentient agents.
>And if the object really is a desk, then it interprets it as what it
>is, so no construction or imposition is involved.
This is meaningless unless you consider the sentence in some statistical
sense of averaging what perceptions of many many sentient beings are
in order to come to the conclusion that "the object is really a desk".
One can make such statements in philosophy but not when you are
discussing the very act of perception and interpretation itself. How
is it possible to know something "is really x" without involving
perception and interpretation of the received signal?
Oliver Sparrow
(Neil Rickert) writes:
> This is not obvious to me. People have been known to use old
> computers as door stops, and to use milk bottles as flower vases.
but then, they are not being milk bottle or computers. They are the
properties which they emit.
Put it this way: all that anything can know by experience is by the
direct or indirect things which the entity which is being known does
to the knower. What the subject is for the object is what the subject
does to the object. The subject 'is' other things to other objects.
Us cognitive objects can create models by whch to extend our perception:
we can recall what it was doing earlier, we can contextualise, we can - to
a degree - predict. We make a lot of fuss about this because we think, rightly,
that it helps us to extend the amount of the 'is' of any given object that we can
know. Reductionists extend this, to the belief that one can, at least in theory,
know all of the possible states of being of a subject, in respect of all
the conceivable objects on which it can operate. We - they - believe this
because they believe that a common set of regularities underpin all
interactions, of whatever kind. We - not they - may see this as unrepresentative
of experience and therefore silly.
The most obvious reason why this perspective is sillyis that we observe a
complex world that has complex structures in it. One cannot map up from the
description of billiard ball atoms, retaining only the terms of reference which
handle the phenomena of interatomic subject-and-object interactions, and arrive
at road traffic regulations. Given Mrs Jones obeying these, you can show how
the billard balls play their role. If you assume all that is the circumsatnces of Mrs
Jones (like the starving economists viewing a heap of baked bean cans - 'let us
deem the existence of a can opener: now, how is the allocation structure to be
set?' ) then reductionism may seem to work: Aha!, that's how her V4 processes
the red light and look - aren't they sweet! - teeny billiard balls, bumbling about to
order. We are surrounded by systems in which simple things - like the teeny
bumblers - emit more complex behaviour (turbulence, phase behaviour,
thermodymanics) when they are allowed to interact. And yes, of *course* if
you take turbulence as read, you can derive ways in which billiards get played.
But not, perhaps, the other way around: derive thermodynamics for me,
using only one atom. The gotta conect, it gotta emerge.
_________________________________________________
Oliver Sparrow
oh...@chatham.demon.co.uk
Anders N Weinstein
In article <507n36$d...@ux.cs.niu.edu>, Neil Rickert <ric...@cs.niu.edu> wrote:
>In <507aaf$h...@usenet.srv.cis.pitt.edu> ande...@pitt.edu (Anders N Weinstein) writes:
>This is something you see as a problem, only because you are
>concerned about "knowing that". I have no such concern.
Well even Ryle acknowledged there was such a thing as knowing that -- hence
the importance of the distinction. So someone's got to deal with it. If
your theory can't account for knowing that, something's wrong.
>From my perspective, a scientific theory is something like a formal
>model which fits the evidence reasonably well. If you want to say
>that my `information' is a theoretical term, that's good enough for
>me. I am not requiring any conscious awareness, which is why I don't
>care for your insistence on content. I am trying to provide a basis
>for explaining neural activity.
Again I don't mind speaking of evidence and theory in the case of
science, but don't see how it applies to the neural activity which
constitutes the acquisition of evidence itself.
>This does not follow. The most basic discoveries may be made by the
>scientist's neurons, and I suppose you could say that they are
>incommunicable. The scientist may then become aware of some
>peculiarity in what he observes, although perhaps not a peculiarity
>that he can verbally describe. At that stage the scientist may be
>able to use methods such as pointing, to place other scientists in a
>position where their neurons are likely to make the corresponding
>discovery. I think that amounts to a form of communication, but one
>that does not require the information to have been conceptualized.
Well how could he know that he has succeeded in getting the others
neurons to make the same "discoveries"? If others agree in what they
say and do towards the relevant phenomena, wouldn't that be sufficient,
and let the neural discovery drop out of the account, like Wittgenstein's
beetle in the box?
>> What happens
>>when we give words to preconceptual information, in your view? Does that
>>distort it, so that the real information is known only to the subject?o
>
>I don't understand the question. The words might be additional
>information, but they could not distort the original information.
Then the information is expressible in words and has the same sort of
content public words do (which is, I take it, externally determined)?
>My experience as an educator is that if I teach in such a way as to
>best help the students gain the non-conceptual background, the
>beliefs take care of themselves. However if instead I concentrate on
>trying to communicate beliefs to the students, many of them do very
>poorly even in acquiring those exact beliefs. I take this as
>evidence that the "justified true belief" characterization of
>knowledge is little more than part of a "Just So" story.
The claim about knowledge (knowing-that) does not follow. Gettier aside,
the JTB story might be fine about knowing-that even if there is a
prerequisite background which it leaves out.
>>OK but you say the neurons have data. What is this data? When you
>>describe it that way you use language and so ascribe some innate
>>representational capacity.
>
>When a neuron transmits a signal, that signal is data. Yes, I'll
>grant that the neuron has the innate ability to transmit signals. I
>suppose in that sense it could be said to have an innate
>representational capacity. But all the neural signal is innately
>capable of representing is the neural signal itself.
OK, are you really saying that neural signals represent there own
occurrence? That makes them infallible all right, but does give them
a representational content. This means that the signal on the optic
nerve doesn't even represent *light*, however.
But I think you are thinking of further signals as representing the
occurrence of earlier ones.
If this is your view it does seem doomed to solipsism to me. Ultimately
you are taking the inputs to be representations which get interpreted
but which never come to refer to anything in the world.
>> What is innate in a normal human being is the ability to acquire
>>abilities, most importantly linguistic ability.
>
>It is the ability to acquire abilities that needs to be explained.
>You want to brush it aside, take it for granted, and as a consequence
>leave everthing important unexplained.
Not exactly, I just oppose explanations of a certain form, or supposing
that mechanical explanations of how these capacities are possible can
play an epistemic role, in explaining how articulate claims get
to be justified or not.
> I don't agree that linguistic
>ability is most important. Surely the ability to feed oneself is of
>greater importance.
I meant important from the point of view of understanding epistemic
states, scientific knowledge, intentionality and the like.
>>But what needs to be explained and how is not at all obvious to me. In
>>many ways we already understand this process. For example, I know what
>>tables, chairs, knives, forks, dinner-times are because I was raised in
>>a culture in which these things have a function, which I came to master
>>(more or less) myself.
>
>Ah, yes. And I know mathematics because I took many mathematics
>classes. Learning theory is so simple when you explain it that way.
I don't see anything wrong with this explanation. Perhaps you think
instead your *brain* took the classes and learned the math for you?
>But haven't you just explained everything away? And you wonder why I
>accuse you of making huge innatist assumptions.
If someone insisted I explain economic value in terms of physical
properties of coins, I would have no choice but to resist. I might
instead explain economic value as derived from human practices of using
the tokens. Now suppose someone were to say: haven't you just explained
everything away, or simply taken for granted all the hard questions.
But no, I haven't.
This phenomenon requires explanationa at *its own level*, using a
certain rich vocabulary for speaking about practices and their
maintenance. Using that level of description, we get all the
explanation we are going to get. If someone were to insist on
redescribing what people do with coins at the physical level as
movements, or even in Skinnerian terms of stimulus-response links,
their descriptions would be true but outside the domain of the
economic. The economic world disappears from view when you describe it
that way, as the aesthetic features of a painting disappear when you
put the painting under a microscope. Start with that basic level and
try to claw your way back to the social normative level, and all we can
say is: you can't get there from here.
In this sense I think the explanation of concept acquistion must be
done at its own level, the level of teaching and learning as cultural
transmission of understanding embodied in practices of using words (and
making measurements and the like in the case of science).
There is some role for explanation at the sub-personal level of
description, sure, but I don't think it can account for the sort of
normative force involved in the notion of concept use.
>> No doubt stuff has to happen in my brain when
>>these skills rubbed off on me, but that's not the interesting level as
>>far as I am concerned.
>
>Then why don't we just eat, drink, and be merry?
I try to when I get the time.
> We can always say
>that everything else is not at the interesting level.
Vindication of this attitude depends on having something to say at
the interesting level, something to say about science or knowledge or
language or mind or consciousness or culture or normativity.
>> And if you want to know how the culture came to
>>start using these things you have cultural history to explain it.
>
>That doesn't explain anything. If you take a bunch of mosquitoes,
>and put them together, they will not form a culture. If you could
>take a bunch of children, and had them raised together, they would
>form a culture even though there had been no cultural history for
>them to rely on.
Not so clear to me -- assuming that "raised" does not imply mature
members with a culture which they impart. If infants were thrown
together on a desert island and they could survive, it might be some
time before a culture emerged, and I bet it would happen very slowly.
Look, of course there are transitions in history between distinct
levels of nature -- from the merely physical, to the living, to the
animal, to the primitive forms of community and communication, to tool
use, other cultural artifacts, language-use and science. The only
point is that if these transitions do involve emergence of domains
which require new explanatory vocabulary, the process itself, while
it can be made intelligible, cannot be explained at the old level.
>>In the case of language use there is more to be said about what
>>abilities with words constitute linguistic competence. Quine and
>>Sellars, for example, give us a bit here. But I am not seeing what
>>is unexplained, if we accept ordinary descriptions as explanations,
>>as we must in this domain.
>
>What is the biological role of language? How does it perform that
>role? How is language acquired?
I am not sure it falls to science, least of all brain science to
explain these things. Perhaps it is a branch of social science --
not very precise, to be sure, but that's the domain in which these
phenomena are located.
>It really should be obvious to you why we are not communicating very
>well. You want to sweep aside, as not needing explanation, all of
>the things I consider it important to explain.
As not needing explanations of the wrong sort.
>If your theory of the world depends on truth as a standard of its
>validity, and if you insist that truth is indefinable, then you have
>no theory of the world.
Not true. Frege had an argument for his view. Anyway I think if you
are making assertions at all you are thereby aiming at truth, and if
you are not making assertions, there is no point talking or trying to
convince one of your views.
>>Anyway, all I need is the internal connection between meaning and truth. If
>>you say that this signal bears information that something affords mailing,
>>that relates it to a state of affairs in the world, that lays down
>>conditions of satisfaction for it -- it is true if and only if the thing
>>really does afford mailing. This is neutral as between different theories
>>of truth. Do you deny it?
>
>I don't deny it. But I insist that the standard you are giving is
>only applicable to evaluate the suitability of my description of the
>signal. It could not be a standard for the signal itself.
But this makes no sense, since the description sets a standard for the
signal itself. So if the description is suitable, there is a standard for
the signal. Anyway I'd be happy enough if this is an apt description of
the information in the signal -- then it would be externally determined,
which is what I am after.
>>I am really still trying to understand how you can talk about information
>>without using the notion of content. Examples would be the best answer,
>>rather than generalities -- what are some examples of information that
>>is processed, according to you?
>
>If two signals are correlated, each signal provides partial
>information as to the state of the other signal. These can be neural
>signals of which I am not conscious, so I don't see how content could
>be assigned.
OK these would be representations *of* neural signals. Now is it your
view that the only information we have available to us is really
information that is *about* neural signals?
>If we live in an information rich environment, then it is expected
>that several of our sensory neurons will pick up some of the same
>information from the environment. That would lead to such a
OK, but now the neurons are picking up information from the environment.
But what is that? It is not just a matter of representing neural signals
themselves, is it?
>correlation between neural signals. Thus whenever two neurons carry
>correlated signals originating from external sources, there is a good
>likelihood that the correlation indicates that the common part of the
>signals is information. This is information that brain processes
>might be able to detect without requiring any assumptions as to the
>content of that information.
Now we have a kind of meta-informational content to the effect that
certain first-order signals themselves are carrying information. Which
information indeed can remain indeterminate for this content.
But again, we never seem to get to the first-order information itself,
only the second-order representation of the fact that first order information
of some unspecified form is being carried.
>Being a realist, I assume that the information has an external
>source. In that sense you could say that the information is
>externally determined. But that is not how you are using your
>externalism. You want to say that knowledge of what is external is
>what allows us to know that the signal bears information. I think
Only in the roundabout way that knowledge of the world enables us to
look inside our own brains and study what we find there, ascribe
information content to it, and the like. We do this on the basis of our
prior knowledge of the environment, true.
>that is backward. Rather, it is the information in the signal
>(whether or not we can be said to `know' that there is such
>information) which allows us to have knowledge of what is external.
It may be true that this is a causal prerequisite for our
having knowledge of the world.
>>But there is no data for the brain at birth, and so no "logical problem"
>>of language acquisition.
>
>There are billions of sensory receptors feeding signals into the
>brain. How could it have no data at birth?
Data is something like observing that a piece of litmus paper has
turned blue, or that the weight of the mixture has decreased, and the
like. Such information may be in the light in Gibson's sense. But
brains can't detect it at birth.
A chaotic spray of photons on the surfaces is not data in this sense --
rather more like James' "bloomin', buzzin' confusion." Soaking in the
photon spray does is not a cognitive act, it does not issue in states
with representational content.
>>Do you really think Einstein had the information about special relativity
>>detected in his brain *prior* to his ability to articulate a theory and put it
>>forward in symbols?
>
>No, of course not. But then I don't count the theory of special
>relativity as being information either. It is analytic, so has no
>informational content. (Well, perhaps it has social information as
>to the form of scientific laws used by this society.) The role of
>theories such as special relativity is not to carry information, but
>to set standards for the information we gather in our observations.
I would like to hear more about this theory. I basically can't understand
it from this sketch, although it sounds intriguing to me.
>> Information
>>without articulation can play a practical role in guiding behavior, but
>>I don't see how it can count as scientific knowledge.
>
>I don't count "information without articulation" as scientific
>knowledge. I don't think scientific knowledge is information at
>all. Measurements made by scientists are information. But the
>scientific knowledge is not the information. Rather, the knowledge
>is the ability to carry out the processes which acquire information.
Which are highly mediated by symbolic and inferential activities,
wouldn't you say?
>>CDs are mindless. Are you suggesting that Lavoisier's brain developed
>>the capacity to transduce or amplify information about oxygen,
>>mindlessly as it were (without conscious reasoning or conceptual
>>activity), and only later did Lavoisier develop the concepts needed to
>>put it into words and communicate some of this information to his fellows?
>
>Roughly, yes. I would rather say that the brain's learning ability
>is an ability to concentrate information. If it becomes sufficiently
>concentrated, then conscious awareness can arise.
>
>>This strikes me as absurd, given the constant interplay between
>>articulation and discovery in the process.
>
>History only reports that which is articulated. You must be cautious
>in trying to use historical evidence to show that there is no
>discovery without articulation.
One need not accept the scientist's own articulations of what they did.
(contra Longley). But I do think that it is only insofar as it could
be articulated that it counts as knowledge or discovery.
Neil Rickert
In <507um0$k...@usenet.srv.cis.pitt.edu> ande...@pitt.edu (Anders N Weinstein) writes:
>In article <507nec$d...@ux.cs.niu.edu>, Neil Rickert <ric...@cs.niu.edu> wrote:
>>>They carry the information, they don't represent it.
>>Ok. If you want to make that distinction, then I am as much against
>>representations as you are, and you should read me as talking about
>>information carrying rather than information representation.
>Good. But the real issue is not the terminology, it is the supposed
>analogy between what a scientist does with datal contents and
>theoretical contents -- where there is in a sense something to
>interpret, namely the data -- and what the brain does in initial
>concept development, which does *not* involve interpreting anything
>"data" like. Rather evolving detectors for that same information, which
>is a non-representational process. After the detectors have evolved one
>might of course speak of the information they detect as data.
From my perspective you are making too much of the distinction
between the neural level and the scientist level. The neurons form a
community of cooperating individuals, and communicate with neural
signals. The scientists form a community of cooperating individuals,
and communicate with concepts.
There is a difference in the complexity of the unit of information
communicated. But otherwise there is considerable similarity. It
only looks very different because we are the community in the first
case and the individual in the second case. But if one takes a
detached viewpoint, the methodology is similar in both cases,
involving cooperation and a division of labor between cooperating
individuals.
>I want the terminology to suggest that proximal patterns of ambient
>light "carry information" whether or not anyone detects it. But the
>patterns are not themselves data, rather one can only talk of data
>after a detector for the information has evolved (and even after that,
>only once this detector functions to implement the input side to a
>concept-using, inference making system).
We are using terms 'information' and 'data' somewhat differently. I
have been using the term 'data' broadly, where even random noise
could count as data. I am using 'information' more selectively, so
that it is not information unless it informs. You seem to want
'data' to be the more selective term.
Neil Rickert
In <5082to$k...@usenet.srv.cis.pitt.edu> ande...@pitt.edu (Anders N Weinstein) writes:
>In article <507n36$d...@ux.cs.niu.edu>, Neil Rickert <ric...@cs.niu.edu> wrote:
>>This is something you see as a problem, only because you are
>>concerned about "knowing that". I have no such concern.
>Well even Ryle acknowledged there was such a thing as knowing that -- hence
>the importance of the distinction. So someone's got to deal with it. If
>your theory can't account for knowing that, something's wrong.
I haven't said that I can't account for "knowing that". Indeed, in
an earlier post, I think I said that the "knowing that" is inferred
from "knowing how" or other background capabilities. But once I
account for it in that way, it doesn't look all that important.
>>From my perspective, a scientific theory is something like a formal
>>model which fits the evidence reasonably well. If you want to say
>>that my `information' is a theoretical term, that's good enough for
>>me. I am not requiring any conscious awareness, which is why I don't
>>care for your insistence on content. I am trying to provide a basis
>>for explaining neural activity.
>Again I don't mind speaking of evidence and theory in the case of
>science, but don't see how it applies to the neural activity which
>constitutes the acquisition of evidence itself.
Neurally speaking, the evidence arrives from stimulation of sensory
neurons, so is represented (or should that be 'carried') in the
pattern of neural activation. The connectivity structure of the
neural network forms the representation system, which allows this
information or evidence to be represented (or carried). Learning is
a process of adjusting the representation system to optimize its
ability to carry the type of information required.
As an analogy, we represent information about positions of ships on a
map. Thus the map is a representation system, and the plotted points
correspond to the represented information. Optimizing the
representation system means constructing the representation system so
as to make the information as readily available as possible with as
little effort as possible. If for my map I use a globe, then
directions and distances can be read off with little effort. So we
might say that the globe is an optimum representation system.
We can infer the sphericity of the earth from the structure of the
representation system (the globe). So the knowing-that of sphericity
is inferrable from the background capability of being able to
optimally represent received information (the positions of ships,
etc).
Analogously, the neural system represents perceptual data, and
constructs the neural network to optimize the representational
capability. Thus my perceptual beliefs -- my reports of what I
currently perceive -- are representations. Anything else is
inferrable from the structure of the neural network, or
representation system. In effect this means that my perceptual
beliefs are information. Longer term beliefs, the type usually
considered to be knowledge, are inferrable from the structure of my
neural representation system. Since they are implied by the
representational structure, in some sense they are analytic, or at
least analytic for me.
When we come to the scientist, there is a public representation
system. This can only work if adequately supported by the private
representation systems of the scientists. The public representation
system depends on concepts. In effect, the community of scientists
represents current information as measurements (of mass, temperature,
velocity, pressure, etc) or as categorizations (identification of
chemical elements, members of species, etc).
Scientific laws specify the relationships between the concepts which
constitute our categorization and measurement framework. In effect
the scientific laws define the structure of our public representation
system. This makes them analytic. That is, the laws are not
themselves representations, but are instead the formal specifications
of our public representation system.
It has often been said that if scientific laws are analytic, they
would have no empirical content. But empirical content is the wrong
requirement for the structure of a representation system. What we
want is high empirical significance. Our laws have high
significance, to the extent that we have designed them to optimize
our ability to represent and use information.
Putting this in terms of cognitive science, what we need to study is
not the ability to make representations, but the ability to construct
optimal representation systems.
----------
Relate my perspective to my charges of your innatism. Your
assumption, and that of most people, seems to be that our
representation system is innate. That is an implicit assumption of
Fodor, when he `proves' that concepts are innate (in "The language of
Thought"). I think Fodor's arguments is correct, based on his
assumptions. If the the structure of representation system in
innately fixed, then then our concepts are also innately fixed, and
it is difficult to see why science should even be useful. Everything
we could ever know would be implicit in the structure of our innate
representation system, and empirical research could not extend that.
If, however, I am correct in saying that we modify our information
representation system, then the activity and effectiveness of science
becomes far easier to see.
>>This does not follow. The most basic discoveries may be made by the
>>scientist's neurons, and I suppose you could say that they are
>>incommunicable. The scientist may then become aware of some
>>peculiarity in what he observes, although perhaps not a peculiarity
>>that he can verbally describe. At that stage the scientist may be
>>able to use methods such as pointing, to place other scientists in a
>>position where their neurons are likely to make the corresponding
>>discovery. I think that amounts to a form of communication, but one
>>that does not require the information to have been conceptualized.
>Well how could he know that he has succeeded in getting the others
>neurons to make the same "discoveries"?
Perhaps he couldn't 'know' in the sense of there being certainty.
But a lot of communication is possible with non-conceptual body
language, and that would allow him to know well enough.
> If others agree in what they
>say and do towards the relevant phenomena, wouldn't that be sufficient,
>and let the neural discovery drop out of the account, like Wittgenstein's
>beetle in the box?
Sure, you can drop the neural discovery from your account. But that
amnounts to deciding that you will take the process of scientific
discovery as being forever mysterious.
>>When a neuron transmits a signal, that signal is data. Yes, I'll
>>grant that the neuron has the innate ability to transmit signals. I
>>suppose in that sense it could be said to have an innate
>>representational capacity. But all the neural signal is innately
>>capable of representing is the neural signal itself.
>OK, are you really saying that neural signals represent there own
>occurrence? That makes them infallible all right, but does give them
>a representational content. This means that the signal on the optic
>nerve doesn't even represent *light*, however.
A neural signal represents only itself. But our concern is not with
individual neural signals, but with the pattern of signals in a
network of causally connected neurons.
A number represents only that number. A number, together with a
causal chain connecting the number to a measuring procedure might
represent a position on a line. Three numbers with suitable causal
connections might represent a position in space.
>But I think you are thinking of further signals as representing the
>occurrence of earlier ones.
But in order to represent earlier signals, we also need the causal
links to be considered as part of the representing system. Otherwise
we are only talking of formal representations in the abstract.
>If this is your view it does seem doomed to solipsism to me. Ultimately
>you are taking the inputs to be representations which get interpreted
>but which never come to refer to anything in the world.
The solipsism comes from your ignoring the causal connections which I
want to consider part of the representing system. Thus you require
an interpreter to fill in the missing causal links.
>>> And if you want to know how the culture came to
>>>start using these things you have cultural history to explain it.
>>That doesn't explain anything. If you take a bunch of mosquitoes,
>>and put them together, they will not form a culture. If you could
>>take a bunch of children, and had them raised together, they would
>>form a culture even though there had been no cultural history for
>>them to rely on.
>Not so clear to me -- assuming that "raised" does not imply mature
>members with a culture which they impart. If infants were thrown
>together on a desert island and they could survive, it might be some
>time before a culture emerged, and I bet it would happen very slowly.
Have you looked at the spontaneous development of sign languages and
the associated cultures, when groups of deaf children get together?
That doesn't quite correspond to the desert island conditions, but is
close enough to be suggestive.
>Look, of course there are transitions in history between distinct
>levels of nature -- from the merely physical, to the living, to the
>animal, to the primitive forms of community and communication, to tool
>use, other cultural artifacts, language-use and science. The only
>point is that if these transitions do involve emergence of domains
>which require new explanatory vocabulary, the process itself, while
>it can be made intelligible, cannot be explained at the old level.
But what is wrong with seeking a meta-level explanation which can tie
these together?
>>If your theory of the world depends on truth as a standard of its
>>validity, and if you insist that truth is indefinable, then you have
>>no theory of the world.
> Anyway I think if you
>are making assertions at all you are thereby aiming at truth, and if
>you are not making assertions, there is no point talking or trying to
>convince one of your views.
That is a fallacy. When you buy a shirt, you don't want the shirt to
be true, you want it to be a good fit. My standard is how well a
theory fits the data, not whether the theory is `true'.
>>If two signals are correlated, each signal provides partial
>>information as to the state of the other signal. These can be neural
>>signals of which I am not conscious, so I don't see how content could
>>be assigned.
>OK these would be representations *of* neural signals. Now is it your
>view that the only information we have available to us is really
>information that is *about* neural signals?
I didn't suggest that. I was suggesting only that such internal
considerations can be evidence that a signal is informative. That is
enough for further investigation. I'm trying to provide a basis for
scientific discovery, and I can't do that if I the scientist is
required to know what he will discover before he discovers it. Thus
there needs to be a basis for considering a signal to be probably
informative, before it is known what the signal informs about.
>>If we live in an information rich environment, then it is expected
>>that several of our sensory neurons will pick up some of the same
>>information from the environment. That would lead to such a
>OK, but now the neurons are picking up information from the environment.
>But what is that? It is not just a matter of representing neural signals
>themselves, is it?
Again, "what is the information" is the question the scientist is
trying to answer. If the scientist is required to have the answer
before he raises the question, then scientific discovery would be
impossible.
>> Thus whenever two neurons carry
>>correlated signals originating from external sources, there is a good
>>likelihood that the correlation indicates that the common part of the
>>signals is information. This is information that brain processes
>>might be able to detect without requiring any assumptions as to the
>>content of that information.
>Now we have a kind of meta-informational content to the effect that
>certain first-order signals themselves are carrying information. Which
>information indeed can remain indeterminate for this content.
Right. The description has to be at a meta-level, if it is not to
presuppose what is to be discovered.
>But again, we never seem to get to the first-order information itself,
>only the second-order representation of the fact that first order information
>of some unspecified form is being carried.
If we could really get to first-order information, then scientific
theories would not be under-determined by the evidence, and when I
looked at the desk in front of me I would see that that it is mostly
empty space with a sparse array of atoms.
However, for most purposes it is far more informative for me to
consider what is in front of me to be a desk made of solid wood, than
to consider it a sparse array of atoms. Thus an optimized
representation system should present it to me as a desk, rather than
as a sparse array.
>>>Do you really think Einstein had the information about special relativity
>>>detected in his brain *prior* to his ability to articulate a theory and put it
>>>forward in symbols?
>>No, of course not. But then I don't count the theory of special
>>relativity as being information either. It is analytic, so has no
>>informational content. (Well, perhaps it has social information as
>>to the form of scientific laws used by this society.) The role of
>>theories such as special relativity is not to carry information, but
>>to set standards for the information we gather in our observations.
>I would like to hear more about this theory. I basically can't understand
>it from this sketch, although it sounds intriguing to me.
In a way, I have given you more earlier in this response.
Anders N Weinstein
>
>>But taking something as a desk is not an activity that involves interpreting
>>retinal signals. Taking something as a desk is a practical act that finds
>
>What? How do you interpret signals?
When I see a desk, I don't interpret a signal. I can't, because I can't
see any signal and so of course I can't apply an interpretation to it.
Maybe you could say that my cortex "interprets" a signal, but that's only
metaphor. It could not be literally true, although it's a useful way
of talking. It's a bit like saying the immune system interprets allergens
as invaders, a useful figure of speech.
> computer human
> -------- ------
>i(t)=info to be sent ASCII,picture,sound etc In the brain/mind
> Meaning is here,and
> not necessarily
> Fodor's LOT.
Well first, Stalling's categories are not all so well defined.
>Now, I ignored vision to make it simple. Obviously when the message is
And second, you are talking about a case like pictures or written
language where there is a representational vehicle involved -- the
image or words on the page. But I was talking about seeing a desk,
not seeing a picture of a desk or a written description of a desk.
>in writing vision is at work. In the case above (of the desk) the
>problem is less formal, so the fact that you can say you see a desk
>involves similar processing except that there is no active agent
>sending a message. The information is produced passively by the "desk"
>purely by virtue of its material existence and its capability of
>reflecting light. There are certainly passive agents that can alter
>messages/data. For example copper cable can induce changes in the
>electrical signals that it carries.
Even if we talk of signals here, we have to distinguish the signal
as EM radiation from the information it contains. Say it contains the
information that there is a desk in front of me. Then when I see that
there is a desk in front of me, I am simply receiving or detecting the
information contained in the signal, not applying an interpretation to it
as if it were a representation.
>We always interpret what our senses tell us.
If what my senses tell me is that there's a desk in front of me, I don't
have to interpret it further.
>This is meaningless unless you consider the sentence in some statistical
>sense of averaging what perceptions of many many sentient beings are
>in order to come to the conclusion that "the object is really a desk".
>One can make such statements in philosophy but not when you are
>discussing the very act of perception and interpretation itself. How
>is it possible to know something "is really x" without involving
>perception and interpretation of the received signal?
Because receipt of information *by means of* an information-bearing
signal constitutes perception of an object in the world. It is not
perception *of* the signal itself as an object which might be
interpreted.
Anders N Weinstein
In article <509cq6$d...@ux.cs.niu.edu>, Neil Rickert <ric...@cs.niu.edu> wrote:
>In <507um0$k...@usenet.srv.cis.pitt.edu> ande...@pitt.edu (Anders N Weinstein) writes:
>
>From my perspective you are making too much of the distinction
>between the neural level and the scientist level. The neurons form a
>community of cooperating individuals, and communicate with neural
>signals. The scientists form a community of cooperating individuals,
>and communicate with concepts.
No doubt there are lots of analogies. I have no problem with exploiting
the analogies and using the metaphors, even formal isomorphisms between
the two processes. But I think many cognitive scientists are carried
away by the intellectualization of sub-personal processes.
The Chomskian idea that there is a "logical problem" of language
acquisition, a problem of how acquired tacit meta-theory relates to
available evidence about the language, is a paradigm of the confusion.
There simply is no such problem. First, of course, mastering a language
need not involve internally representing a metatheory for it. But more
importantly, a cognitive subject literally has no evidence propositions
or capacity to base inferential moves upon them until it has been given
a language in which to express these propositions and conduct its
intellectual operations.
>We are using terms 'information' and 'data' somewhat differently. I
>have been using the term 'data' broadly, where even random noise
>could count as data. I am using 'information' more selectively, so
>that it is not information unless it informs. You seem to want
>'data' to be the more selective term.
Yes, I was speaking such that information that Jones is out at first
base might be objectively there in the light, but is not data for Al
who doesn't know the rules of baseball. Bob who does might detect it
and then use it as data for some further processing if he wanted.
So I guess I was using "information" to denote potential data, "data"
to denote actual data (actually obtained obserational contents). I make
no use of a concept that includes random noise -- maybe energy flux, or
something like that.
Anders N Weinstein
In article <509i2n$d...@ux.cs.niu.edu>, Neil Rickert <ric...@cs.niu.edu> wrote:
>In <5082to$k...@usenet.srv.cis.pitt.edu> ande...@pitt.edu (Anders N Weinstein) writes:
>
>When we come to the scientist, there is a public representation
>system. This can only work if adequately supported by the private
>representation systems of the scientists. The public representation
>system depends on concepts. In effect, the community of scientists
>represents current information as measurements (of mass, temperature,
>velocity, pressure, etc) or as categorizations (identification of
>chemical elements, members of species, etc).
While it seems you have a very stimulating theory here, I can't help
noticing what looks like a key point of disagreement. I take
Wittgenstein and others to have taught us that the only thing that
matters for epistemology or philosophy of science is the public
representational systems. The "private" ones could be different for
different people; in principle if not in fact, the "private" ones could
even be nonexistent and it would make no difference to the public
practice, to what is communicable and shareable in language. For that
reason we can and should "divide through" by any differences in the
"private" systems. It is very important to me that what is hidden or
incommunicable is of no significance *at this level of description*,
(although it might have a role in causal explanation).
Another point is that one person can test whether another has mastered
the rules for the use of the representational system by public
criteria, without having to make speculative hypotheses about what's
inside the other's mind or brain.
Don't you think we can theorize at a level that abstracts from what is
private, and talk instead about the function of public concept-using
practices as systems for representing the results of measurements, in
the way you want? In these terms, what fits people to take their place
in the scientific game is their practical mastery of the use of these
public representational systems. This is a behavioral skill that, like
a virus, transmits itself from generation to generation; it can also be
described in complete independence from whatever neural circuitry
supports it.
>It has often been said that if scientific laws are analytic, they
>would have no empirical content. But empirical content is the wrong
>requirement for the structure of a representation system. What we
>want is high empirical significance. Our laws have high
>significance, to the extent that we have designed them to optimize
>our ability to represent and use information.
I wonder if you could give an example here of what you mean.
>Putting this in terms of cognitive science, what we need to study is
>not the ability to make representations, but the ability to construct
>optimal representation systems.
If we stick to the public ones, we don't need cognitive science at
all. We do need a theory of the social function of institutionalized
representational systems, similar to an account of, say, tools.
>Relate my perspective to my charges of your innatism. Your
>assumption, and that of most people, seems to be that our
>representation system is innate. That is an implicit assumption of
But I don't think this at all. In some sense my principal
"representation system" is English, for example. That is hardly
innate: it was socialization molded me into an English machine. It
turned other people into French, German, Chinese machines. Among these,
more specialized training produced mathematics and, say, quantum-theory
machines, systems that are competent to use the specialized symbolic
tools and the measuring apparatus that is involved with them.
So I don't think of *any* representational system as innate. Perhaps
the brain at birth is not really a representation-using system at all,
and only comes to be one when it yoked by socialization into moving in
certain ways and not others. Then we should study rules for the use
of English expressions, say, without mentioning brains.
>If, however, I am correct in saying that we modify our information
>representation system, then the activity and effectiveness of science
>becomes far easier to see.
But I agree entirely that we modify our representational systems. Why
not just speak at the cultural level about public representational
systems and capacities to operate with them, however? Then the history
of representational systems is broadly a cultural study, like the
history of other institutionalized tools and artifacts.
>Sure, you can drop the neural discovery from your account. But that
>amnounts to deciding that you will take the process of scientific
>discovery as being forever mysterious.
What mystery is solved by including irrelevant implementation details?
Why not speak as you did above, at the level of concepts as public
representational systems used in the interpretation of equally public
measurement operations?
>A number represents only that number. A number, together with a
>causal chain connecting the number to a measuring procedure might
>represent a position on a line. Three numbers with suitable causal
>connections might represent a position in space.
I would not say a number represents itself, or anything, for that
matter. True we can use numerals or, if you like, the numbers they
denote, to represent things. Then this representational function is a
social institutional matter, like the function of tables. Apart from
our practices a number doesn't represent anything.
More another time. But I wish you would see that I don't believe in
innatism. I believe what's literally in the head is basically
irrelevant to the analysis of the social medium in which public
representational systems have their being.
Anders N Weinstein
But it really is a solid desk in front of you (I assume -- you could be
fooled by an imitation). It is solid in the ordinary sense if it keeps
its macroscopic shape, you can rest your coffee cup, etc. It is a desk
because it was manufactured for that purpose. Why do you think atomic
theory is even remotely relevant to the question of whether there's a
solid desk in front of you, in the everyday sense of the words? (If I
am about to climb on top of the desk to put up a window shade, and I ask,
is this desk solid?", it would be some kind of idiotism to respond
"no, there are spaces between the molecules.")
So, in Gibson's realistic terms, you *have* acquired the "first-order"
information that there's a desk in front of you, which information is
really there in the light, and, I would myself add, which has a content
that can be objectively true.
Why are you thinking of this everyday statement as comparable to those
of a scientific theory? It is as if you think it is a useful but
inadequate representation of the micro-world of atoms. Better to think
of it as a perfectly accurate representation of the macro-world of
solid desks.
I think I can imagine an objection along these lines: why not then
think of witch theory as a perfectly adequate representation of the
world of witches? For a realist, the answer must be flat-footed --
because there are no witches. At least that is what I believe. If there
were some reason for believing there are no solid desks in the ordinary
sense, then we would have to say the same thing for desks. But such a
reason has not been provided.
H. M. Hubey
ande...@pitt.edu (Anders N Weinstein) writes:
>When I see a desk, I don't interpret a signal. I can't, because I can't
>see any signal and so of course I can't apply an interpretation to it.
Your "seeing" is interpretation. We have stereo vision, so we can
only focus on a single spot. In order grasp the whole picture
the eyes scan the area, and the complete picture/scenery
is created via interpretation in your brain. Secondly, even if you
say simply that you are seeing something with some legs and a flat
top etc, it is still being interpreted. If you say you see a desk
that's double interpretation. There are levels and levels.
>Maybe you could say that my cortex "interprets" a signal, but that's only
>metaphor. It could not be literally true, although it's a useful way
>of talking. It's a bit like saying the immune system interprets allergens
>as invaders, a useful figure of speech.
In the same way as "I see a desk", yes. It is a useful way of
talking. But it doesn't mean that just because you can say "I simply
see a desk" that no interpretation or signal processing is going on.
It seems that when people are discussing perception itself, more
careful attention should be paid to what is going on instead of
skipping all the steps and jumping to some final result which because
of its apparent simplicity ignores/evades/disguises what is
going on behind the scene.
>>i(t)=info to be sent ASCII,picture,sound etc In the brain/mind
>> Meaning is here,and
>> not necessarily
>> Fodor's LOT.
>Well first, Stalling's categories are not all so well defined.
Much better defined than some of the words being tossed around
on these newsgroups. Naturally Stallings trips over the same
problem as everyone else when he gets to
"meaning/interpretation/semantics".
>image or words on the page. But I was talking about seeing a desk,
>not seeing a picture of a desk or a written description of a desk.
It doesn't matter much. The brain can create depth from lighting
and shading clues on 2D pictures, and create wholes from parts.
That's how neural networks work.
>Even if we talk of signals here, we have to distinguish the signal
>as EM radiation from the information it contains. Say it contains the
>information that there is a desk in front of me. Then when I see that
>there is a desk in front of me, I am simply receiving or detecting the
>information contained in the signal, not applying an interpretation to it
>as if it were a representation.
Our five senses can all be thought of as processing signals. What is
the big difference between someone tapping on someone's back, or
tom-toms sending messages, or telegraph/telephone. They all have
to be interpreted, and they are all signals with which we perceive
the world. WE even get signals on how hot/cold it is, how objects
feel when we touch them etc. Think of that as passive signaling.
Desk is passively signaling because the light waves bounce off.
A rock standing under the sun for hours also passively tells us
that it's hot, if we touch it. We are all able to tell all this
without paying attention to the processes that create it because
we are used to talking about end results. We perceive the world
thru signals via our senses. All these signals have to be interpreted
by our brain/mind, even heat,cold, wind, sounds, noises, etc.
So when you are talking about not interpreting the signal from the
desk, we can ask "what signal?" Light hits it and bounces off in
all directions. Some of it towards our eyes. There are various
wavelenghts there. We call them color: interpretation. The light
waves that bounce off, say nothing about dimensionality. We interpet
these light waves to create a 3D image in our minds. And it's not
easy. Those working in artificial vision can probably attest to it.
If all you have is the data in the optic array, you need some
serious interpretation to be able to tell apart objects from simple
lines. When is a line a line and when is it an edge of a 3D image?
Very difficult question. If this problem were licked our
manufacturing/industrial facilities would become like farms
in developed countries; about 5% of the population would be able to
run all the factories. The simple fact is that it's extremely
difficult to be able to find an object in a box, and then pick it
up and the orient it somehow on some machine so that it can be
worked on. That's essentially what a huge number of people in
factories all over the world do. They pick up an object from a box,
put it on some vice-like holder and then push some button. Then pick
up the object again and put it in another box.
>If what my senses tell me is that there's a desk in front of me, I don't
>have to interpret it further.
it has already been interpreted, many times over. This is the
"hard problem", not discussing things that don't matter. Solving
scientific problems is like making decisions; if you make all the
little decisions that you have to make, you'll never be faced with
a big decision. If you can solve all the little problems in
cognition that come your way, you'll never have this "hard problem"
of "intelligence/consciousness/etc/etc". They are hard problems
only because we barely have handles on the smaller problems of which
they are a part. That's why these hard problems are for philosophers
and not for computer scientists or engineers.
>Because receipt of information *by means of* an information-bearing
>signal constitutes perception of an object in the world. It is not
>perception *of* the signal itself as an object which might be
>interpreted.
Methinks you are using the word perception to mean reception. That
is also a part of signal processing, but perception is something
else. Like the other hard problems, and others like cognition,
consciousness, awareness, intelligence, they are just words people
make up because they have nothing better to do, and as we learn
more and more about it the old terminology no longer seems to
suffice or no longer seems germane. So either new words phrases
are made up or new problems are unearthed. Things will become
clearer as we learn more and more.
Neil Rickert
In <50ah75$r...@usenet.srv.cis.pitt.edu> ande...@pitt.edu (Anders N Weinstein) writes:
>In article <509i2n$d...@ux.cs.niu.edu>, Neil Rickert <ric...@cs.niu.edu> wrote:
>>In <5082to$k...@usenet.srv.cis.pitt.edu> ande...@pitt.edu (Anders N Weinstein) writes:
>>When we come to the scientist, there is a public representation
>>system. This can only work if adequately supported by the private
> ^^^^^^^^
>>representation systems of the scientists. The public representation
>^^^^^^^^^^^^^^^^^^^^^^^
>>system depends on concepts. In effect, the community of scientists
>>represents current information as measurements (of mass, temperature,
>>velocity, pressure, etc) or as categorizations (identification of
>>chemical elements, members of species, etc).
>While it seems you have a very stimulating theory here, I can't help
>noticing what looks like a key point of disagreement. I take
>Wittgenstein and others to have taught us that the only thing that
>matters for epistemology or philosophy of science is the public
>representational systems.
I must say this is a surprising remark.
This is, after all, an AI newsgroup. Are you suggesting that the way
to do AI is to completely ignore the details of the AI system, and to
only discuss the public discourse the AI system will conduct? That
sure seems a strange way of doing AI.
Peter Lupton
I've been away on holiday again, so this response is a little tardy. However,
better late than never (perhaps).
In article: <4vr38c$4...@usenet.srv.cis.pitt.edu> ande...@pitt.edu (Anders N Weinstein) writes:
>
> In article <643507...@luptonpj.demon.co.uk>,
> Peter Lupton <lup...@luptonpj.demon.co.uk> wrote:
> >In article: <4vkmhv$d...@usenet.srv.cis.pitt.edu> ande...@pitt.edu (Anders N Weinstein)
writes:
> >>
> >I do think that what you seem to call 'objective reference' is brought about
> >by the fact that we are actual mechanisms and that much of our language is
> >'from *here* on out', as it were. This strikes me as the most natural form
> >of reference - forms which aren't 'here on out' are, in comparison, quite
> >sophisticated. Do we know how to identify a place on Earth without
> >exploiting some indexical? That is, that we can identify our own Galaxy/
> >Sun/Planet? And what if there isn't one unique universe as some think? Our
> >language is learned, as it were, from the local to the increasingly wide-ranging.
> ...
> >That indexicals don't show themselves in practice is not really true, is it?
>
> I agree that indexicality is basic, pervasive and ineliminable. I take this
> to support externalism and the idea that a cognitive subject must be an
> embodied being-in-the-world, not a detached brain, since only such a thing
> can *express* an indexical thought.
Where we differ isn't in our assessment of the importance of indexicality,
but in the meaning of the term 'externalism', and what is supported and what
not. More of that later.
> But Searle uses peculiar *self-referential* indexicals to build an
> internalist theory. His idea, for example, is that perceptual states are
> *self-referential* and refer to the world only as "the distal cause of
> this very experience" -- in that way the sense or mode of presentation
> is world-independent although the reference is not. It is only *this use*
> of indexicality -- peculiar self-referential ones to support an internalist
> theory of intentionality -- that I was opposed to.
So far as I'm concerned, that theory may have a place in some/many
cases or not. I doubt it can do as much work as Searle wants it to do.
> >What I will insist on, however, is that the representational structures
> >created in parsimony will, necessarily, (for our sort of world) include
> >far more than what falls within the boundary of the organism per se.
> >Now if you are suggesting that the organism can't distinguish those
> >parts of its own representational states which are self-representational
> >from those which aren't, I'd have to know why. And if the answer is
> >no more than the *assertion* that the subject must be able to *see* it,
> >then I'll be extraordinarily unimpressed.
>
> Well I don't know what you mean if you think *I* am "distinguishing"
> features of representational states buried in my lateral geniculate nuclei,
> say. These are not objects for me.
>
> I can of course distinguish among thoughts, and the like, various
> self-referential elements.
>
> >One of the ways, for example, that parsimony would represent things
> >is as a 3-d manifold. But not just a 3-d manifold, because visual
> >sensory data are special - they are plainly (and parsimony will
> >produce this) associated with perspective mappings. Which means that
>
> Not so obvious. In one sense there clearly is a point of view in
> visual consciousness. On the other hand if we speak Gibson-wise, we need
> not appeal to any representational states with less than three-dimensional
> content (G's "layout" of environmental surfaces). In particular, we
> need not view the job of the visual system as inverting a perspective
> mapping, rather, detecting information about layout and other properties,
> including the position of the subject.
I accept that at some level of abstraction we can avoid talking about
things of less than 3d content. But *that* isn't our discussion here,
is it? Our discussion isn't about whether it's possible to avoid certain
things but whether it's possible for sub-personal accounts to account for
personal access to the wider world. And my contention is that is *is* possible
and I'm giving you a mechanism showing *how* it's possible - pointing out
that it might be, on your account, unnecessary, is beside the point
of this discussion.
Nor is it relevant whether it's *obvious* or not. The point I'm
making is a dual point - firstly to do, empirically, with what our
sensory stream of impulses are. The second point is essentially
a mathemtical or theoretical point about what parsimony will deliver
in such cases. Whether this result (that a 3d representational structure
is a consequence of parsimony applied to our sensory impulses) is
*obvious* or not is hardly relevant. The relevant issue is just
whether it's true or not. And, so far as I can tell, it's true.
> >the 3-d representation is incomplete: it needs a distinguished point
> >and orientation to define that very perspective point. And that
> >perspective point and orientation moves with motor responses etc.
> >Now this is just one of the many roles for 'I' that comes in the
> >representational process. When I say that the bleeding edge is represented,
> >what I mean is just that there are a whole range of such cases, all of which
> >converge on the same represented sub-structure - each of them versions
> >of 'I'.
>
> I don't deny that the I is represented in all of this. As Gibson emphasized,
> patterns in the optic array bear information about the subject at the
> same time as about the environment.
To repeat, now you're just changing the subject. We're not discussing
alternative accounts - we're discussing whether sub-personal accounts
*can* lift themselves to the wider world. And my contention is *that* they
can and I'm showing *how* they can.
> >I don't think that it could be the case that the organism must *actually*
> >see and touch, since I don't see that I could tell the difference between
> >what's happening to me now and what would happen to me if the universe ended
> >at my skin but the sensory neurons fired just the same.
> >
> >Now you could say at this point that the two explanations would just be totally
> >different. But I don't really think that'd wash. It would turn out to be
>
> *What* two explanations? In the second case there is nothing to explain
> in the representational manner.
Sorry, I thought that was obvious. In the one case there is my experience
as produced by contact with the external world, in the other case there is
my experience as produced without contact with any external world. There
are two subtley different explanations for each case.
> >the *same* explanation in both cases except that you'd insist in one case
> >in using the term 'aware' and in the other case you'd insist on using the
> >term 'seems to be aware' and such-like. That apart, the explanations would be
> >identical.
>
> Well this is a standard internalist intuition.
I don't think it is. Internalism is the doctrine that meanings are
envrironment independent. Exeternalism is the denial of that: the assertion
that meaning is environment dependent. You adopt a particularly strong form
of externalism, the assertion that *without* an environment, there is no
meaning of any sort: 'a worthless hunk of metal'. So we have a gradation
from internalism to externalism of the weaker sort to externalism of the
stronger sort.
The intuition I give is not internalist per se, it is perfectly
compatible with externalism of the weaker sort.
Your strong form of externalism plainly entails extenalism, since 'no meaning'
is an environmental dependence. But the converse (that environmental
dependence entails meaninglessness in the no-environment case) has not
been shown to be true, and I can't see any good reason to think it is
true. Further, I think our intuitions are strongly that it isn't true.
I think, intuitively speaking, there are very good reasons to think
that our intuitions shift us from internalism to externalism. I also
think that there are very good reasons to accept that our intuitions
shift us from strong externalism. It seems to me that Descartes intial
insight that, if I'm fooled well enough, then I'm fooled, is as strong
today as it ever was. We have to rephrase it for modern tastes, but it is
still there, as strong as ever.
> .......But I don't think a
> mere brain would even *seem* to be aware.
OK. I think that intuition is against you here and I'm not sure what else
we have to call on. My intuitions which drive me from internalism and
also drive me from strong externalism.
Would your position change if, instead of talking about a 'brain-universe'
(a universe which is exactly my brain) we talked about a 'body-universe'
(a universe which is exactly my body). So that, for example, when I breath in,
air is generated at the boundary of the universe and when I breath out,
air is destroyed. So that, when I look at my hand, there is no causal link
from hand to eye - there is merely the generation of appropriate photons
entering my eye. I just raise the question because you seem very keen on
the bogy as a whole. I'm just wondering whether bodies are enough for your
purposes and, if so, why you think that. You could even have the ambient
optical array if you wanted - just that the ambient optical array is never
caused by any distal structures because either there aren't any or (as in the
case of my foot) they never cause ambient optical effects.
> I think *personal* awareness
> stands to a brain as economic value to a coin -- in different contexts,
> like pieces of metal have different value, and without the right
> surroundings, it's just a worthless hunk of metal. Similarly, a brain
> on Twin Earth can be part of a person who has one sort of thought, a
> homologous brain on Earth can be part of a person with a different
> psychology (different thoughts), and a brain that was always in a vat
> would be just a brain, and no person would be constituted.
Well, I appreciate your re-statement of your position. But, correct me if
I'm wrong, but I thought we were discussing to what extent sub-personal
structure can generate access to the wider world and I was doing this
by pointing out that sub-personally, we'll have a 3-d representational
structure which is incomplete because it stands in need of a perspective
point (a position and direction) in order to account parsimoniously for
sensory impulses. And that was, so far as I can tell, a pretty strong
form of 'I', one that can be subsequently re-inforced by other considerations
if it was felt necessary.
Your response has been to say it wasn't obvious and to repeat
your denial of my claim.
> You have to think of the intentional level as globally, not locally
> supervenient on the natural scientific description of the body and brain.
>
> You must also remember that even subjective seemings are in the
> higher-level world, the world of intentionality, so the content of
> seemings switches too as you move from the Earth case to the Twin Earth
> case. Things *seem* differently to both people.
But they don't seem different in a way that requires explanation other
than mere location and indexicality picks out. To Earthman it seems as
though gold is the essence of this stuff, to Twin-Earthman it seems as
though gold is the essence of this stuff. I don't see what else I'm
supposed to have to worry about.
You seem to think that expternalism per se is a problem for me. I fail
to see where. I *think* the reason is that you conflate externalism with
strong externalism. So far as I can tell, strong externaism is not
logically equivalent to externalism, although it does entail it.
> >What's necessary is the system of representation I've been stressing,
> >together with the observation that such representational structures
> >must also represent onself as part (and not the whole) of that
> >representational structure.
>
> I agree that this is necessary for full-blown conceptual
> representation. (By the way, I get the point from P.F. Strawson's
> reading of Kant's Transcendental Deduction in _The Bounds of Sense_)
What isn't clear is why you feel that an external world is required
rather than just the seeming of an external world. Here I think
intuition is very strongly against you. You might like to argue that,
since intuition is strongly against internalism, intuition must be
for your strong form of externalism. But this won't wash, because
the intuitions are for externalism of the weaker, not the stronger, sort.
> >Fair enough. Searle is obviously too dramatic for your tastes. But I
> >thought the problem you raised was how sub-personal structures could
> >sub-personally lift themselves to the wide world. My use of Searle in
> >a box was to dramatise the inferential processes as explicit inferential
> >processes which needn't be done as explicit inferential processes.
>
> I don't think the problem is just stale drama. I thought, perhaps wrongly,
> that you were trading on a confusion of what Searle qua person in his
> own right can do (posit an external world outside the room) and what
> Searle qua program can do (manipulate meaningless symbols). Your
> picture must be that some of the things Searle *qua program* does
> constitute "positing an external causal structure" to account for
> patterns in sense data, sense data which Robbie does not represent
> (nothing in his belief box refers to them). I continue to believe that
> Searle qua program is always blind and never does come to cognize or even
> posit an external world, while Robbie is a being in the world and does
> not have to "posit" it.
I thought what I was trying to do was to account for how Robbie, qua
being in the external world, would come to posit an external world in
sub-personal terms. Now I don't think that talk of Robbie not having
to posit an external world and just being a being in the external world
is going to prove adequate for my purposes.
The point is this. If you adopt a highly abstract position, that's
fine and dandy. I have no problem with some-one doing that. Indeed,
in my work (program specification, design and implementation) I do it all
the time. But then I pay a price. And the price is that there are certain
questions which, although they can be raised in the context of the abstraction,
can't be answered in the context of that abstraction. In the context of
that abstraction, one really just has to shrug. There is absolutely no
reason to suppose that abstract views are explanatorily closed - and typically,
in my experience, they aren't.
Two sorts of failure of explanatory closure are typically initial conditions
and error conditions. If a Coke machine fails to deliver Coke, then, at the
abstract level (coins in, press button, Coke out) all one gets is non-determinism
at the abstract level: sometimes the Coke can comes out, sometimes it doesn't.
If the question is pressed (at the abstract level) *Why* didn't the Coke can come out?
then the answer (at the abrstract level) must be: Look, we're working at the
abstract leve and some things can't be explained at the abstract level. Tough.
It would be quite wrong to start attempting to 'explain' such errors by
introducing phrases such as 'mis-coining'.
Similarly for intial conditions. Abstractly, Coke machines just appear as a
factory output. In order to explain how Coke cans are made, the abstract level
will not satisfactorily answer the question.
So it is with Robbie as a fully-fledged intentionial subject. We can still,
in this case, press the question: 'How does Robbie come to be?' There are two
questions here - one evolutionary and one to do with child development (from
before birth, if necessary).
It is this notion of explanatory non-closure that your abstract view can't/
souldn't hide and I find that I still stand in need of explanation. And an
explanation typically would involve moving to a lower level, of course. So
what I'm pressing here is the question: How does the sub-personal account
bring about, in aggregate, the personal view. And my account here is that,
in parsimony, we see the way in which sub-personal structures combine to form
representational states in which they, as an aggregate, feature as a small
part of the whole. Now it seems to me that this just is what it is for Robbie
to be developing a notion of 'I'.
> >> No I don't mean much more than that, but all those things require
> >> taking the relevant system to be much larger than a brain, it seems to me.
> >> An embodied organism in a material and social environment, for example.
> >> If you ascribe to a brain the ability to represent "that bird is fat",
> >> I think it must be derivative from the ascription to a person who could
> >> point at a bird. Content devolves on brain states from the top down and
> >> the outside in, it seems to me.
> >
> >I think all you are saying is that we can be uncertain whether we mean.
>
> Not at all. Rather that what we mean is not determined solely by what's in
> us. In fact in many cases we can be certain what we mean.
>
> I think the problem with such arguments is this. Note first, I am a
> kind of property dualist (actually a pluralist). I am this kind of
> dualist about the relation between economic properties and material
> ones, and in particular I don't think the former supervene locally on
> the latter. That should be uncontroversial. Externalists think a
> similar relation holds between intentional contents of thoughts
> (including subjective phenomenology) and structural states of brains.
>
> Now the important point is that this non-local supervenience *includes*
> both the content of "how things seem to the subject" as well as the content
> of second-order states which refer to first order states. For example,
> Oscar and Twin-Oscar might both be said to know what they mean.
>
> The real issue is not the causal role of brains in making the higher level
> possible, but the powerful illusion of autonomy, the illusion that content
> must be determinable entirely by resources solely within myself. In this
> way the conscious subject would not be dependent on anything and could,
> like Descartes, hope to justify all its knowledge from a vantage point of
> suspended belief in anything else.
>
> I take it this is a chimera, and a person qua epistemic subject is
> consitituted in dependence on the world and other people.
So that we really do have an answer to Descartes? Only not the sort he
was after. That, as a result of profound philosophical analysis we have
made the discovery that there really must be an external world after
all (and not just it's seeming) because I'm a subject? Not 'I think,
therefore I am' but 'I, therefore a world'. Stepping over the ontological
argument, as it were.
My point is, however, that there is a world of difference between
looking for absolute certainty ala Descartes and just rejecting the
concern altogether. It seems to me that you are jumping from one failed
attempt at certainty to another. Why not just accept that absolute
certainty is not for us? Set that obscure and useless epistemic point
to one side and move on to more worth-while questions. A while ago we
were discussing 'epistemic disaster'. There is, in my mind, no epistemic
disaster in just saying that absolute certainty is not for us. Fiddling
about pretending that we couldn't be fooled about the external world is idle.
--
Cheers,
Pete Lupton
Anders N Weinstein
>careful attention should be paid to what is going on instead of
>skipping all the steps and jumping to some final result which because
>of its apparent simplicity ignores/evades/disguises what is
>going on behind the scene.
But I think perception consists *entirely* in what you are calling "the
final result". At the epistemological or mental level there simply is
nothing at all "behind the scenes" -- it's not an epistemic or
cognitive event at all *until* that "final result" occurs. That result
is the starting point, although further cognitive processes might be
based on it. This has to be if we are concerned with epistemology,
with a subject's justification for his or her beliefs.
>It doesn't matter much. The brain can create depth from lighting
>and shading clues on 2D pictures, and create wholes from parts.
>That's how neural networks work.
Speak Gibson's way, and you can say instead: patterns of light and
shade in the light can specify depth, and the visual system -- perhaps
a neural network -- can simply detect this depth information. Then,
whatever its internal structure, the whole network can be viewed as a
complex transducer for natural depth information -- it just converts
this information from one form to another, it doesn't create it.
>we are used to talking about end results. We perceive the world
>thru signals via our senses. All these signals have to be interpreted
>by our brain/mind, even heat,cold, wind, sounds, noises, etc.
Part of this can be said if you talk Gibson's way -- the world naturally
gives off information-bearing signals. I'm not sure what it adds to
say these signals must be interpreted. If the signal naturally carries
certain information, why not just the information is detected without
interpretation?
>So when you are talking about not interpreting the signal from the
>desk, we can ask "what signal?" Light hits it and bounces off in
>all directions. Some of it towards our eyes. There are various
>wavelenghts there. We call them color: interpretation. The light
Not if the signals naturally contains information about color.
>waves that bounce off, say nothing about dimensionality. We interpet
This is not obvious -- on Gibson's view a dynamic pattern -- one obtained
by moving through the array of light -- *does* contain information about
the relative layout of surfaces. So depth information does not need to
be added to it.
>these light waves to create a 3D image in our minds. And it's not
>easy. Those working in artificial vision can probably attest to it.
I don't think it is easy. But the right conceptualization is important.
>If all you have is the data in the optic array, you need some
But what is controversial is precisely what the data in the optic array
is -- what it's information content should be taken to be. In one sense
the optic array contains whatever data that a perceiver can get out of
it.
>serious interpretation to be able to tell apart objects from simple
>lines. When is a line a line and when is it an edge of a 3D image?
>Very difficult question. If this problem were licked our
It might be a bit easier for a visual system that can move. Nothing
determines that only static snapshots should be viewed as the input to
visual processing, as opposed to dynamic patterns actively obtained.
Anyway, the fact that it is hard does not mean that interpretation
is involved. It just means it is hard. But no matter how hard or
complex it is, nothing prevents us from viewing the whole process at a
high level as one of information transduction or amplification.
>it has already been interpreted, many times over. This is the
*What* has been interpreted? You say the original signal. But the
question is what the *information* in that signal is. If the
information in is the same as the information out, no interpretation
has taken place.
>they are a part. That's why these hard problems are for philosophers
>and not for computer scientists or engineers.
Probably why I prefer philosophy to engineering.
Anders N Weinstein
In article <50b2m8$e...@ux.cs.niu.edu>, Neil Rickert <ric...@cs.niu.edu> wrote:
> . I take
>>Wittgenstein and others to have taught us that the only thing that
>>matters for epistemology or philosophy of science is the public
>>representational systems.
>
>
>This is, after all, an AI newsgroup. Are you suggesting that the way
Well it's a little bit of an accident that I wound up posting to this
group. But it is still an AI *philosophy* group.
>to do AI is to completely ignore the details of the AI system, and to
>only discuss the public discourse the AI system will conduct? That
>sure seems a strange way of doing AI.
No, but I think that's the way to do philosophy of language, philosophy
of mind, and philosophy of science.
Note also that some of this is in any case necessary for AI if the AI
system is going to participate in community life and discourse.
More to the point I have a very definite philosophical view about what
AI is really about. It is not really about mental states or the mind,
it is about the sub-personal structures that *implement* or make
possible the capacities possession of which constitutes intelligence or
mindfulness.
For many concepts there is a reasonably clear distinction between the "what
it is to be" and the "what enables it to be" questions. What it is for a
car to have good pickup is for it to be able to accelerate quickly. What
enables it is the collaborative working of all the mechanisms inside the car.
What it is for someone to have concepts or to understand the world
scientifically is for them to have an interlocking set of capacities --
chiefly, I think, the capacity to articulate, make inferences, argue,
perform measuring operations, and the like. What enables or makes this
possible is the functioning of neural circuitry. But the circuitry that
makes intellectual activity possible does not itself perform the
intellectual operations.
Anders N Weinstein
In article <334502...@luptonpj.demon.co.uk>,
Peter Lupton <lup...@luptonpj.demon.co.uk> wrote:
>
>I accept that at some level of abstraction we can avoid talking about
>things of less than 3d content. But *that* isn't our discussion here,
>is it? Our discussion isn't about whether it's possible to avoid certain
>things but whether it's possible for sub-personal accounts to account for
>personal access to the wider world. And my contention is that is *is* possible
>and I'm giving you a mechanism showing *how* it's possible - pointing out
>that it might be, on your account, unnecessary, is beside the point
>of this discussion.
Fair enough. I still have my doubts, about which more below.
>Nor is it relevant whether it's *obvious* or not. The point I'm
>making is a dual point - firstly to do, empirically, with what our
>sensory stream of impulses are. The second point is essentially
But I wanted to emphasize that it is not a straightforward empirical
question how our "sensory stream of impulses" should be conceptualized
for the purposes of a higher-level functional inquiry. For example, the
(Humean) presupposition that it should be treated as a series of static
disconnected snapshots has been and continues to be widespread, even
though this itself is in no way a simple empirical result. Many
cogntive scientists -- Zenon Pylyshyn, say, or others working on
"animate vision" -- accept Gibson's insight that temporally extended
patterns might just as well be viewed as the information bearing
inputs, in which case richer information might be available.
So, again, I didn't want to *just* grant you the idea that the
input only bears information about 2D array, and must be enriched by
postulation of a third dimension -- even if this postulation is
justified on the grounds of parsimony. (How does this account for the
information delivered via a compound eye, which does not produce 2-d
images, btw?)
>To repeat, now you're just changing the subject. We're not discussing
>alternative accounts - we're discussing whether sub-personal accounts
>*can* lift themselves to the wider world. And my contention is *that* they
>can and I'm showing *how* they can.
All right. But I think we're going in circles. You say that a parsimony
guided system will in effect form the hypothesis of a 3'd world of
persistent reidentifiable objects and the simultaneous presence of the I
as occupying a changing location and viewpoing within this spatial world.
In this way, it obtains a very parsimonious means of anticipating
future stimulation, much more compact then simply enumerating the stimulations
themselves.
But my beef has always really been this. As long as you think of the
*function* of the 3d spatio-temporal representation as one of predicting
stimuli, or experience in the "thin sense" -- basically Quine's point of
view too -- I see a danger of phenomenalism about 3-d objects (perhaps
a holistic phenomenalism, as in some moments in Quine, in which it is
the whole system of belief which only makes contact with the world at
a boundary of thin experience). The danger is that *as far as the system
is concerned* what it is to be an external object is simply to be a
permanent possibility of obtaining stimulation.
As long as the stimulation accords with the predictions of the system,
it will of course "take" itself to be justified. This will happen for a
brain in a vat as well as for an embodied creature in the world. But
the difficulty is seeing how, from the point of view of the system, its
posits could have any content which would be objectively false if the
system were a brain in a vat. I think you are impelled to a verificationist
conception of the meaning of these posits according to which they are equally
*true* in the case of a brain in a vat -- even if there is no external
world. Otherwise, the meaning you ascribe would be irrelevant to the
functional role of the system.
But I take it that this means it never really does lift itself to posit
an external reality, at best it always only posits a certain order and
connection among its proximal thin experience.
Now you are perfectly free to help yourself to an unexplicated notion
of "positing an external reality" which has a content that is not
really explained within your functional theory. You can say that the
system is functionally identical in a brain in a vat and in a person in
the world, and yet that only the latter has posited the city of
Pittsburgh to induce order among its experiences. Then you are just
helping yourself to something you do not explain -- the power of the
system to enter a state that answers for its correctness to things
where a certain three rivers meet in Pennsylvania. Then I would point
out simply that you are just helping yourself to this unexplained
power, which is where all the work is being done.
I.e it seems that the only significance the "posit of an external reality"
could have for the system is exhausted by its consequences for proximal
stimulation, which could be the same even if there were no Pennsylvania.
Well this is a traditional idea -- phenomenalism, positivism, some
varieties of pragmatism all seem to hve this consequence. But I think it
is totally unacceptable, for in that case no objective reference is
ever really secured.
I don't have this problem, since I don't believe the external world was
ever "posited". I follow the existentialists in eliminating modern
epistemological worries through metaphysics. That is, in thinking a
cognitive subject is constituted as always already understanding a
world.
>I think, intuitively speaking, there are very good reasons to think
>that our intuitions shift us from internalism to externalism. I also
I actually can't see why you are any variety of externalist. Remember
that externalism concerns intentional content (sense, intension), not
merely reference. Searle has a theory according to which intentional
content is internally determined, even though of course reference
varies. I would have thought your theory is similar, even if the
details are different. I wonder how your supposed externalism
manifests itself? Doesn't the brain in a vat make exactly the same
posits as the brain of the person in the world, on your view, so that
both arrive at the same parsimonious system for anticipation of thin
experience?
>think that there are very good reasons to accept that our intuitions
>shift us from strong externalism. It seems to me that Descartes intial
>insight that, if I'm fooled well enough, then I'm fooled, is as strong
>today as it ever was. We have to rephrase it for modern tastes, but it is
>still there, as strong as ever.
I admit it is strong. But we can resist.
>OK. I think that intuition is against you here and I'm not sure what else
>we have to call on. My intuitions which drive me from internalism and
>also drive me from strong externalism.
I believe there is more than just "intuitions". My intuitions are shaped
by theory. For example, a line of thought that runs in outline like this:
Intuitions (sense data) without concepts are blind (Kant). But concepts
are essentially normative, in effect, subject to rules governing their
combination in judgments and inferences (Kant, Frege). But our handle on these
rules comes via consideration of rules for the use of public linguistic
expressions (Wittgenstein, Carnap, analytic philosophy generally). But
the rules governing use of linguistic expressions are in the first
instance embodied in customs or social practices abroad in the culture
(later Wittgenstein, Sellars). An individual masters these rules
implicitly in being trained to use symbols in accordance with the use of
his or her fellows (Wittgenstein, Quine, Sellars).
Later the individual can come to reason silently (Sellars). But when we
talk about inner symbolic inference, we talk about it by reference to
the same common standard we use in evaluating and assessing public
linguistic performances. (Burge) Because the standard has its being in
social space, even supposedly private reasoning is always plotted
against the (evolving) public standard when reason-giving explanation
is being done.
On the existentialist side we have Heidegger, say, on the need for what
he called "the world", also a kind of pre-existing social normative
structure, as a context without which the understanding subject (H's
"Dasein") can not exist.
In any case, I think that it is more than just mere intuition that
teaches us there is no subject without a world. It concerns rather the
conditions for having a concept, and the idea that the understanding
implicit in concepts is abroad in social space before being in the
individual mind.
>Would your position change if, instead of talking about a 'brain-universe'
>(a universe which is exactly my brain) we talked about a 'body-universe'
>(a universe which is exactly my body). So that, for example, when I breath in,
Well maybe, it seems like a rather far fetched case. For example what happens
when the body tries to move in space -- does it really move, or only
seem to? Can it bump into objects?
>air is generated at the boundary of the universe and when I breath out,
>air is destroyed. So that, when I look at my hand, there is no causal link
>from hand to eye - there is merely the generation of appropriate photons
>entering my eye. I just raise the question because you seem very keen on
>the bogy as a whole. I'm just wondering whether bodies are enough for your
^^^^
body, I presume (a bogy to you?)
>purposes and, if so, why you think that.
Mainly because I think the thing that thinks must be a living thing, that
psychological activities are one species of vital operations, like digestion,
and that you don't have a living thing unless you have a body and an
environment in which it does its living. Psychology is world-involving
because biology is.
Perhaps a brain in the vat world could be considered a degenerate case of
a living thing in an environment, but it is far from the normal context in
which we know how to apply our concepts of thinking, feeling, desiring,
willing, etc.
>What isn't clear is why you feel that an external world is required
>rather than just the seeming of an external world.
It's not clear whether Kant achieved that either. But the seeming of
an external world might require the being of a world, if its required
for the assignment of content.
>intuition is very strongly against you. You might like to argue that,
>since intuition is strongly against internalism, intuition must be
>for your strong form of externalism. But this won't wash, because
>the intuitions are for externalism of the weaker, not the stronger, sort.
I am not sure that intution is so clear on this point. Look at it this
way: suppose my whole world is a hallucination. Well then we have two
levels of the appearance-reality distinction to consider (a point
noted, in effect, by most idealist thinkers). In one sense the whole
world I know is appearance. But we can also ask the question of what is
real and what is appearance *inside* that hallucinated world, just as
we can talk about what is appearance and reality inside a story, or a
dream, or a virtual reality game.
For example, inside the hallucinated world, it might be that it appears
to me that there's a house in front of me, but when I walk around to
the side, I see that it's only a facade. I have corrected a belief
about my phenomenal world, even if the whole thing is hallucination
from a wider vantage point. Inside the hallucinated world it might
appear that I am on a flat motionless body which is the center of the
universe, yet with inquiry it could turn out that it is just another
planet in a solar system. And so on.
Now I maintain that inside this supposedly hallucinated world, I will
*at least* ascribe content third-personally to others using externalist
principles, in accordance with the Putnam and Burge arguments. That is
why there intuitions get a grip on so many people. Further, if I
discover there is a brain inside my head and come to study it, I will
ascribe content to events in that brain in accordance with externalist
principles as well (and puzzle over there relation to my conscious states).
So externalism will be true of other people's content, and of neural
content. And I will never be in a position to apply an interpretation
to my *own* present conscious states -- I am in them, but I don't view them
as objects of interpretation.
Now as a phenomenologist, I don't believe in a god's eye view, and so I
am really *only* interested in things that are true *inside* the world
I can experience, the world of offices and houses and planets, other
people, and myself as one among many. And externalism is not just true
but, I think, intuitive in this domain.
>So it is with Robbie as a fully-fledged intentionial subject. We can still,
>in this case, press the question: 'How does Robbie come to be?' There are two
>questions here - one evolutionary and one to do with child development (from
>before birth, if necessary).
Agreed.
>It is this notion of explanatory non-closure that your abstract view can't/
>souldn't hide and I find that I still stand in need of explanation. And an
>explanation typically would involve moving to a lower level, of course. So
>what I'm pressing here is the question: How does the sub-personal account
>bring about, in aggregate, the personal view. And my account here is that,
>in parsimony, we see the way in which sub-personal structures combine to form
>representational states in which they, as an aggregate, feature as a small
>part of the whole. Now it seems to me that this just is what it is for Robbie
>to be developing a notion of 'I'.
OK, your position seems very well stated here to me. In some sense I have
to admit that my outlook is basically circular, like many philosophical
positions -- you can buy it as a whole for the benefits it brings, but
it is hard to actually argue for each piece of it against someone who
resists its appeal.
That said, I think your idea involves an identification of the personal
and sub-personal levels that I want to resist. You say you are moving to
a lower level, which is fine with me, but your remarks at the end suggest
an unholy mixing of the levels by my lights.
In some sense of course we can say that the sub-personal entities doing
their stuff and the person-level consciousness of the world emerging
are coeval and intimately related, just as one nation mobilizing for
war, or an economy entering a recession is coeval and bound up with the
millions of individual actions that constitute these things. Better, an
"analog" dynamical system (neural network) coming via training to
instantiate a digital symbolic device -- it is moving from being one
kind of thing to another.
But beyond that I want to keep the notional distinction of levels. You
can have the idea of a sub-personal explanation of how Robbie comes to
be, but the point remains that Robbie is never in a position to choose
on reasoned grounds whether or not to posit an external world. Robbie
as epistemic agent doesn't really exist until the sub-personal entites
have done their stuff, in your terms, and then the implicit posit has
been made for him by something else.
Also, what Robbie comes to understand, and the reason the sub-personal
entities are doing what they are doing, is largely due to the fact that
there *already* is an understanding of the world abroad in the culture,
and that people are taking measures to transmit this to Robbie -- they
teach him to walk, to fetch things, to dress, be polite, sit at tables,
eat with knives and forks, etc. Again, Robbie does not -- and cannot --
learn anything if he must, like a little Descartes, *first* establish
the trustworthiness of his teachers. No his mind is just *stuffed* with
these understandings, and it is only much later that he might, little
by little, be in a position to question and revise some of it.
On the other hand, they *never* teach him that there is a
three-dimensional world of objects or that the ground can generally be
relied on when one walks. But this is not because Robbie is such a
genius that he has hit upon these things in the solitude of his own
conscience, nor because they are innate. It is rather because these
things simply do not have to be represented at all in order to act in
the world. (What lies at the bottom of the language game is not belief,
but our acting in a certain way). Contents like these are *very late*
achievements for Robbie, even if in your sense they are early implicit
achievements for Robbie's brain.
So, to come back to an earlier point, positing an external world is not
a reasoning Robbie can be held responsible for. Perhaps Robbie is
fortunate that his brain can do it for him, but Robbie's reasoning only
comes on to the scene once that "unconscious posit" is in place. And
then that posit does not enter into the explanation of Robbie's
justification for beliefs.
>So that we really do have an answer to Descartes? Only not the sort he
>was after. That, as a result of profound philosophical analysis we have
>made the discovery that there really must be an external world after
>all (and not just it's seeming) because I'm a subject? Not 'I think,
>therefore I am' but 'I, therefore a world'. Stepping over the ontological
>argument, as it were.
Well this is a staple of existentialist thought. See Richardson,
_Existential Epistemology_, or the discussion of Descartes in
Heidegger's _Being and Time_. There is an interesting discussion in
the analytic tradition in McDowell, "Singular Thought and the Extent
of Inner Space" in Pettit and McDowell, eds., _Subject, Thought and
Context_.
>My point is, however, that there is a world of difference between
>looking for absolute certainty ala Descartes and just rejecting the
>concern altogether. ...
>to one side and move on to more worth-while questions. A while ago we
>were discussing 'epistemic disaster'. There is, in my mind, no epistemic
>disaster in just saying that absolute certainty is not for us.
Indeed not. The disaster that threatens is not loss of certainty, it is
loss of content or aboutness at all.
CDJ
In article <hubey.8...@pegasus.montclair.edu>,
hu...@pegasus.montclair.edu (H. M. Hubey) wrote:
>ande...@pitt.edu (Anders N Weinstein) writes:
>
>>When I see a desk, I don't interpret a signal. I can't, because I can't
>>see any signal and so of course I can't apply an interpretation to it.
>
>Your "seeing" is interpretation. We have stereo vision, so we can
>only focus on a single spot. In order grasp the whole picture
>the eyes scan the area, and the complete picture/scenery
>is created via interpretation in your brain. Secondly, even if you
>say simply that you are seeing something with some legs and a flat
>top etc, it is still being interpreted. If you say you see a desk
>that's double interpretation. There are levels and levels.
I remember a year or two ago: A co-student of mine went around for a solid
week asking questions of the form: "When I see an apple, do I see the apple,
or do I just see the _surface_ of the apple?"
This is one situation, among various others, in which it is good to take
Quine's advice not to look for a more familiar language, as there is none. (I
think that this is an almost-quote from very early in W&O.)
CDJ
Neil Rickert
In <50ae42$r...@usenet.srv.cis.pitt.edu> ande...@pitt.edu (Anders N Weinstein) writes:
>>From my perspective you are making too much of the distinction
>>between the neural level and the scientist level. The neurons form a
>>community of cooperating individuals, and communicate with neural
>>signals. The scientists form a community of cooperating individuals,
>>and communicate with concepts.
>No doubt there are lots of analogies. I have no problem with exploiting
>the analogies and using the metaphors, even formal isomorphisms between
>the two processes. But I think many cognitive scientists are carried
>away by the intellectualization of sub-personal processes.
I am not trying to intellectualize sub-personal processes. Perhaps
you think that because of the way I am using `representation', which
you take to apply only to the personal level. You should take me as
using it in the sense you referred to as `carrying' the information.
I don't mean to imply anything intellectual about it.
>The Chomskian idea that there is a "logical problem" of language
>acquisition, a problem of how acquired tacit meta-theory relates to
>available evidence about the language, is a paradigm of the confusion.
>There simply is no such problem. First, of course, mastering a language
>need not involve internally representing a metatheory for it.
I'm not a Chomsky fan. I agree that he overintellectualizes the
problem. I think it a mistake to assume that the brain has a theory
of language, in any sense comparable to our usual understanding of
'theory'.
> But more
>importantly, a cognitive subject literally has no evidence propositions
>or capacity to base inferential moves upon them until it has been given
>a language in which to express these propositions and conduct its
>intellectual operations.
If anything, you should be accusing me of under-intellectualizing
science. It is my contention that a great deal of science is done
before there are propositions or inferences.
Andrzej Pindor
In article <507um0$k...@usenet.srv.cis.pitt.edu>,
>analogy between what a scientist does with datal contents and
>theoretical contents -- where there is in a sense something to
>interpret, namely the data -- and what the brain does in initial
>concept development, which does *not* involve interpreting anything
>"data" like.
Does this "concept development" take place when the signals from transducers
(senses) reach the brain or perhaps it needs no external input at all?
In the latter case you would be suggesting that we can develop concepts
without any sensual (empirical) contact with the reality. In the former case,
the formation of the concepts _on the basis of sensory signals_ can very well
be called "interpretting the signals as data". If you prefer to use other
words to describe the process, this is fine with me, but then the disgreement
is basically about terminology, not the essence.
>........Rather evolving detectors for that same information, which
>is a non-representational process. After the detectors have evolved one
>might of course speak of the information they detect as data.
>
I do not see why you insist on calling "evolving detectors" as a non-
representational process. After all, "evolving detectors" is driven by certain
aspects of the information - different information will lead to evolving
different detectors, right? Apparently in a certain experiment kittens
brought up in an environment consisting of vertical lines only were later
unable to see horizontal lines - they have not evolved detectors for them.
So perhaps again the disagreement is a terminological one? Perhaps you simply
have an aversion to the word "representation"?
>I want the terminology to suggest that proximal patterns of ambient
>light "carry information" whether or not anyone detects it. But the
This sounds very much like ascribing to "the God's eye view" notion of
reality and hence it would be a very misleading terminology. Can you
explain what good would it do, in particular considering the fact that you
yourself seem to dislike "the God's eye view" notion of reality (at least you
have said so on several occasions)?
>patterns are not themselves data, rather one can only talk of data
Why can't one talk of the "patterns' as data? Matter of terminology again?
>after a detector for the information has evolved (and even after that,
>only once this detector functions to implement the input side to a
>concept-using, inference making system).
>
>This is in contrast to the scientific case, where something can fairly
>be said to be interpreted, since the data there is conceptualized.
What about formation of new concepts? Is the data which leads to a formation
of a new concept not "data" before the concept is formed?