specific vs generic dependence, and their examples

[Marijke Keet]

Hi All,

I'm (still) trying to make sense out of the additions to BFO v1.1, in
particular w.r.t. specific and generic dependence. Such a distinction
makes sense, but how it is defined (well, the string of text) in the OWL
file now and the corresponding examples cause me more confusion than it
solves. Writing it up resulted in a longer and font-layouted piece that
is not very readable in plain-text email, so I've put it here:

http://keet.wordpress.com/2008/07/09/bfo%e2%80%99s-specific-and-generic-dependence-and-generalising-progress-in-essential-and-mandatory-parts/

Any suggestions---or even better: solutions---are most welcome.

Best regards,
Marijke

C. Maria Keet
KRDB Research Centre
Faculty of Computer Science
Free University of Bozen-Bolzano
Piazza Domenicani 3
39100 Bozen-Bolzano
Italy
tel: +39 04710 161287
fax: +39 04710 16009
email: ke...@inf.unibz.it <mailto:ke...@inf.unibz.it>
web: http://www.inf.unibz.it/krdb/
home: http://www.meteck.org <http://www.meteck.org/>

[Pierre Grenon]

Hiya,

quick answer

I think part of your problem(s?) is actually exposed in your analysis.
You complain that an SDC changes in value and so on and that this
causes some inconsistency. As you say, the liquidity of the blood has
a value. Saying that the liquidity of the blood changes in value is
not to say that the blood has distinct liquidities. It could be the
theory, but it's not BFO. Let me explain in the context of the blood
example.

A blood sample A has a liquidity L. A is an indepednent continuant. L
is a SDC. L inheres in A at all times L exists. (Perhaps also L is the
sort of SDC that inheres in A in just a part of A's lifetime, but
that's another story.) The point is first that if L is A's liquidity,
it is the liquidity of no other entity whatsoever. It is specific to A
and it depends on A for its existence.

Now at different moments of the existence of A it may be that L has
different values. At point in time, it may be that L has the same
value as L' where L' is the liquidity of blood sample B. But it is
still L that A has, independently of the value of L. B does not have L
and A does not have L', even if at a point or another L and L' happen
to have the same value.

So part of your problem seems to be that you reason on the value of an
SDC and not SDCs themselves. With this distinction in mind, perhaps
the notion of SDC will be a bit clearer.

I'm not sure I understood your complaints about GDC

cheers
pierre

[Matthew Pocock]

On Wednesday 09 July 2008, Pierre Grenon wrote:
> Hiya,
>
> quick answer

quick response :)

> A blood sample A has a liquidity L. A is an indepednent continuant. L
> is a SDC. L inheres in A at all times L exists. (Perhaps also L is the
> sort of SDC that inheres in A in just a part of A's lifetime, but
> that's another story.) The point is first that if L is A's liquidity,
> it is the liquidity of no other entity whatsoever. It is specific to A
> and it depends on A for its existence.
>
> Now at different moments of the existence of A it may be that L has
> different values. At point in time, it may be that L has the same
> value as L' where L' is the liquidity of blood sample B. But it is
> still L that A has, independently of the value of L. B does not have L
> and A does not have L', even if at a point or another L and L' happen
> to have the same value.

OK - can you please explain to me what the difference is betwen the liquidity
of the blood, and the value of the liquidity? If the value of the liquidity
is not the liquidity, then where does this value go in BFO? I hope we can
agree that the liquidity and the value of the liquidity (if these are
different things) are different to a quantification (observation) that is
also different to the representation of that as a number in some units.

Matthew

[Cristian Cocos]

> A blood sample A has a liquidity L. A is an indepednent continuant. L

> is a SDC. L inheres in A at all times L exists. ... The point is first

> that if L is A's liquidity, it is the liquidity of no other entity
> whatsoever. It is specific to A and it depends on A for its existence.

Trope theory strikes again :-)

Say you split that blood sample in half (fiat split or bona fide split). What liquidities might those halves have? L1 and L2? I assume that L1 & L2 have no connection whatsoever to L, right? ... except perhaps that they may have the same "value," is that so?

C

[lampucerka]

c'mon, it so easy: if you split the blood sample B with liquidity L
into two parts, B1 and B2, with liquidities L1 and L2, respectively, so
that B1 constitutes an x-th part of B (and B2 an (1-x)-th part), then L1
constitutes an x-th part of L (and likewise for L2). surely both L1 and
L2 have *the same* value, at least asymptotically. who said values of
qualities are specifically dependent?

vQ

[Pierre Grenon]

On Wed, Jul 9, 2008 at 8:34 PM, Cristian Cocos <cri...@ieee.org> wrote:
>
>> A blood sample A has a liquidity L. A is an indepednent continuant. L
>> is a SDC. L inheres in A at all times L exists. ... The point is first
>> that if L is A's liquidity, it is the liquidity of no other entity
>> whatsoever. It is specific to A and it depends on A for its existence.
>
> Trope theory strikes again :-)

still the gospell, isn't it?

> Say you split that blood sample in half (fiat split or bona fide split). What liquidities might those halves have? L1 and L2? I assume that L1 & L2 have no connection whatsoever to L, right? ... except perhaps that they may have the same "value," is that so?

Yeah, according to me, which is not saying much, each half of the
sample would have its own liquidity. You can make it more fancy and
say it's only true if the separation is bona fide, maybe it's
warranted to make this qualification, I don't know. There may be a
mereotopological story to tell regarding the connection between the
sample parts. It could be long winded... If it's a bona fide
separation, that would be a story similar to that of a cake cut in
parts...

Now these liquidities could have the same value, but I'm not sure that
it has to be the case.

> C

[Pierre Grenon]

Maybe it has to do with the example... I just don't see why liquidity
would have to be taken as something homogeneous like that. But if
that's the case wioth liquidity there's no reason to take it for
granted that it would have to be the case with other quantifiable
properties

> vQ
>
>
>
> >
>

[lampucerka]

gosh, that was a *joke*. perhaps a bad one, but no worse than the rest...
vQ

[Matthew Pocock]

On Wednesday 09 July 2008, Pierre Grenon wrote:
> On Wed, Jul 9, 2008 at 8:34 PM, Cristian Cocos <cri...@ieee.org> wrote:

> >> A blood sample A has a liquidity L. A is an indepednent continuant. L
> >> is a SDC. L inheres in A at all times L exists. ... The point is first
> >> that if L is A's liquidity, it is the liquidity of no other entity
> >> whatsoever. It is specific to A and it depends on A for its existence.
> >
> > Trope theory strikes again :-)
>

> still the gospell, isn't it?
>

> > Say you split that blood sample in half (fiat split or bona fide split).
> > What liquidities might those halves have? L1 and L2? I assume that L1 &
> > L2 have no connection whatsoever to L, right? ... except perhaps that
> > they may have the same "value," is that so?
>

> Yeah, according to me, which is not saying much, each half of the
> sample would have its own liquidity. You can make it more fancy and
> say it's only true if the separation is bona fide, maybe it's
> warranted to make this qualification, I don't know. There may be a
> mereotopological story to tell regarding the connection between the
> sample parts. It could be long winded... If it's a bona fide
> separation, that would be a story similar to that of a cake cut in
> parts...
>
> Now these liquidities could have the same value, but I'm not sure that
> it has to be the case.

The same value instance, or two equivalent value instances? And where does

this value go in BFO?

Matthew

>
> > C
>
>

[Pierre Grenon]

apparently I missed the list, not that it matters much there's also a
clarification below

On Wed, Jul 9, 2008 at 9:11 PM, Pierre Grenon <pierre...@gmail.com> wrote:
>>> Now these liquidities could have the same value, but I'm not sure that
>>> it has to be the case.
>>
>> The same value instance, or two equivalent value instances? And where does
>> this value go in BFO?
>

> I meant perhaps they could have the same 'value instance' (although I
> don't like the term)
>
> BFO has no value,

humm.. seems very telling to me as a way of phrasing it. I mean there
is no theory of value and how they attach to their bearers in BFO, of
course.

> that would be part of GROSS, the Great Realist
> Ontological Suggested Scheme... I'm just making this up obviously (I
> say that for Germans or people who have stayed too long in Germany and
> who like me can't spot what's supposed to be a joke) I mean it's part
> of the big picture, it has to be attached to BFO, it's just not in
> BFO. I really think we went through this
>
>> Matthew
>>
>>>
>>> > C
>>>
>>> >>
>>
>>
>

[Matthew Pocock]

On Wednesday 09 July 2008, Pierre Grenon wrote:

> apparently I missed the list, not that it matters much there's also a
> clarification below
>
> On Wed, Jul 9, 2008 at 9:11 PM, Pierre Grenon <pierre...@gmail.com>
wrote:
> >>> Now these liquidities could have the same value, but I'm not sure that
> >>> it has to be the case.
> >>
> >> The same value instance, or two equivalent value instances? And where
> >> does this value go in BFO?
> >
> > I meant perhaps they could have the same 'value instance' (although I
> > don't like the term)
> >
> > BFO has no value,
>
> humm.. seems very telling to me as a way of phrasing it. I mean there
> is no theory of value and how they attach to their bearers in BFO, of
> course.

So, this is possibly the root of the confusion. People expect to be able to
say things like 'the height of Matthew is 5ft7' but using the BFO way, we can
say 'Matthew exists' and 'Matthew's height exists' but we can't say that the
height in question can be represented as 5ft7, or that I am the same height
as my sister. We can't describe in any meaningful way how my actual height
value (of 5ft7) is related to the random variable that is observations of my
height. Also, we can't meaningfully seperate out the height from its
representation in feet and inches, or in millimeters.

Or to put it another way, most science is about quantitative models
of 'values'. These 'values' are at least as 'real' as the continuants and
qualities that BFO recognises - after all, they are the only things we have
any possibility of actually observing, so where do they go in a realist
ontology?

Matthew

[lampucerka]

Pierre Grenon wrote:
>
>> BFO has no value,
>>
>
> humm.. seems very telling to me as a way of phrasing it. I mean there
> is no theory of value and how they attach to their bearers in BFO, of
> course.
>

eager to find 'bfo has no value' hashed by google...

>> that would be part of GROSS, the Great Realist
>> Ontological Suggested Scheme... I'm just making this up obviously (I
>> say that for Germans or people who have stayed too long in Germany and
>> who like me can't spot what's supposed to be a joke) I mean it's part
>> of the big picture, it has to be attached to BFO, it's just not in
>> BFO. I really think we went through this
>>
>>

just moved and already complaining about germans? they paid your bread,
didn't they?

one (very serious, like others) way to solve the 'value of liquidity'
problem is to call for quales (a la dolce): L1, the liquidity of B1, is
an entity (a quality) distinct from L2, the liquidity of B2, even if
they have the same liquidity quale, i.e., a position in the space of
values for liquidity qualities. ('have'? 'the same'? 'quale'? please
read, e.g., [1] before complaining. they actually say 'so we
distinguish', 'so when we say', etc., without too much insistence on
that that's how R is, bless them for that.)

vQ

[1] http://wonderweb.semanticweb.org/deliverables/D17.shtml and related
pubs, e.g., Gangemi et al. Sweetening ..., EKAW 2002

[Bill Hogan]

I find this discussion fascinating.  One question I myself had when I looked at one of Ingovar Johansson's papers on determinables and determinates (http://hem.passagen.se/ijohansson/ontology6.htm), was: is "one foot" a universal, of which there are millions of instances of lengths which inhere in some ruler?

If the relationship between a determinate and determinable is "is-a", as Johansson suggests, then I think "one foot" is-a "length".

So, if we have two rulers, R1 and R2, whose lengths respectively are L1 and L2, then if L1 has a value of one foot and L2 has a value of one foot, do we have two instances F1 and F2 of the universal "one foot" that are related somehow to L1 and L2 (value_of?), or are L1 and L2 themselves the instances of the universal "one foot" which is a sub-type of "length"? 

And what happens when we talk about disparate objects both with a length of one foot, for example a fish and a piece of tile? Do their lengths if the same instantiate the same length universal?  Or is there something fundamentally different between fish-lengthness and tile-lengthness?

Or is there a completely different approach to the ontology of determinables and determinates that makes better sense of all this somehow?

Bill

[Alan Ruttenberg]

On Jul 9, 2008, at 9:40 PM, Matthew Pocock wrote:

So, this is possibly the root of the confusion. People expect to be able to 

say things like 'the height of Matthew is 5ft7'

Except that it is exceedingly improbable that the height of Matthew is 5ft7.

It is almost certainly a bit more or less, and variable over time.

but using the BFO way, we can 

say 'Matthew exists' and 'Matthew's height exists' but we can't say that the 

height in question can be represented as 5ft7

Well, recently we can say that we can make a measurement of Matthew's height, that it results in an information artifact, that the information_artifact is_measurement of (height quality that inheres in Matthew at a certain time)

Information artifacts are generically dependent continuants.

That instance of height measurement information artifact

 has_value 67 

 in_units inches

 of_dimension length

 

Some work to do to put this into an ontology to share, product of http://neurocommons.org/page/First_IAO_workshop 

or that I am the same height as my sister.

Certainly not true. But perhaps some theory of this if you deal in approximate equality.

However it could be true that a measurement of each yields an equivalent measurement information artifact. Such an artifact, if in the same units, or if transformed by some process into another with the same units, could have a value that could be compared.

We can't describe in any meaningful way how my actual height 

value (of 5ft7) is related to the random variable that is observations of my 

height.

Actually, we probably can't know the actual height, but will do OK with the observations (measurements) as described above.

Also, we can't meaningfully seperate out the height from its 

representation in feet and inches, or in millimeters.

The height itself would be a determinate quality instance. It's specific value is elusive. (Cocos rule: any measurement is wrong) One can have classes of determinate quality instances that are in certain ranges:

Class: Determinate heights between 5'6 and 5'8

Class: Determinate heights between 5'6.5 and 5'7.5

Likely that the instance of determinate height quality instance is instance_if of those two.

The "between 5'6.5 and 5'7.5" is part of how we could name that class. However that class could have other names, ones which mentioned the centimeters.

A representation is an information artifact. So we can separate that from the determinate quality instance this, and say it is "about" the quality instance.

And, of course, there are the measurements.

Or to put it another way, most

s/most/some/

science is about quantitative models 

of 'values'. These 'values' are at least as 'real' as the continuants and 

qualities that BFO recognises

Agreed. We recognize them as information artifacts, which are generically dependent continuants. Such entities inhere in such as dram, paper and ink, LCD screens, and brains.

- after all, they are the only things we have  any possibility of actually observing,

Oops - it is the qualities we observe. The results of the observations are the values, which are information artifacts.

so where do they go in a realist  ontology?

Are you getting some hints from the above? This deserves a longer write up, which will follow.

See also, 

[1] http://hem.passagen.se/ijohansson/information2.PDF

[1] http://hem.passagen.se/ijohansson/ontology6.htm

[Marijke Keet]

Hi Pierre,

It is exactly that you talk about a blood *sample* that has a liquidity, then I agree---but the example in the OWL file does not take the setting of particular blood samples.

In addition, regarding

> So part of your problem seems to be that you reason on the value of an
> SDC and not SDCs themselves. With this distinction in mind, perhaps
> the notion of SDC will be a bit clearer.

yes, because the definition talks about instances. an instance of "Liquidity" without specifying its value seems odd to me.

I don't have a real issue with GDC (though it is flexible in how to interpret the example)

cheers,
marijke

[Marijke Keet]

well, the discussion of having values, i.e., permitting data types, in
an ontology is not new, and adding them may easily complicate
integration -- say, one ontology has height in integer-centimeters, the
other one has it in real-feet -- as is well-known in conceptual
model-based database integration.
So it is fair to say that BFO will not (never?) include any datatypes &
values aspects?

as for trope-background, and comparing it with DOLCE's qualities etc,
I'll check up the trope things first before commenting (but at least
it's good to know that the SDC/GDC motivation has something to do with
the trope discussion--maybe that can be added in the annotation filed in
the OWL file?). for DOLCE's qualities and qualia, however, I'd rather
suggest more recent material, being wonderweb deliverable D18 and
"qualities in formal ontolgy" [1]

[1] http://www.loa-cnr.it/Publications.html

cheers,
marijke

[lampucerka]

Marijke Keet wrote:
> for DOLCE's qualities and qualia, however, I'd rather
> suggest more recent material, being wonderweb deliverable D18 and
> "qualities in formal ontolgy" [1]
>
> [1] http://www.loa-cnr.it/Publications.html
>
>

indeed, although the remarks about quales have the same wording as those
published in Gangemi et al., Sweetening..., EKAW2002.

vQ

[Phillip Lord]

>>>>> "PG" == Pierre Grenon <pierre...@gmail.com> writes:

PG> I think part of your problem(s?) is actually exposed in your analysis.
PG> You complain that an SDC changes in value and so on and that this causes
PG> some inconsistency. As you say, the liquidity of the blood has a value.
PG> Saying that the liquidity of the blood changes in value is not to say
PG> that the blood has distinct liquidities. It could be the theory, but
PG> it's not BFO. Let me explain in the context of the blood example.

PG> A blood sample A has a liquidity L. A is an indepednent continuant. L is
PG> a SDC. L inheres in A at all times L exists. (Perhaps also L is the sort
PG> of SDC that inheres in A in just a part of A's lifetime, but that's
PG> another story.) The point is first that if L is A's liquidity, it is the
PG> liquidity of no other entity whatsoever. It is specific to A and it
PG> depends on A for its existence.

I think that you have missed some key entities here.

A blood sample A has a liquidity L.

L is a SDC.

Then there is the mathematical construct $l = r/u$ (or whatever it is -- sorry
my fluid dynamics are not very good) which describes liquidity. I call this
m(L). This is a GDC.

Then there is a measurement that we can make of liquidity. This is an GDC. I
am not sure whether this is a measurement of L or m(L). It's not actually the
value of L or m(L).

Then if we take a set of measurements (call this S(M(L)) )this gives us an
mean value. These are also a GDC. By asserting a distribution of these
measurements, this give us an average value which is a mean with a measure of
variance.(bar n of S(M(L)) ) This average is also a GDC.

The variance is a measure of how far the mean value we have is likely to be
from the real average. I am not sure whether this statistical "real average"
is related to the value of L which is in BFO-speak a unit of reality, so I
don't know whether it's a SDC or a GDC.

All of which is pretty confusing. Given that, for most of science, we are
actually interested in the mathematical construct and that the only thing we
can actually have a value for is bar n of S(M(L)), it seems most of the
essential things to model are GDCs. The only thing that is a SDC is the one
thing that we can't ever have a value for.

So, I don't think that your example is a clear cut as you think.

Phil

[Phillip Lord]

>>>>> "w" == waku <lampu...@gmail.com> writes:

w> c'mon, it so easy: if you split the blood sample B with liquidity L into
w> two parts, B1 and B2, with liquidities L1 and L2, respectively, so that
w> B1 constitutes an x-th part of B (and B2 an (1-x)-th part), then L1
w> constitutes an x-th part of L (and likewise for L2). surely both L1 and
w> L2 have *the same* value, at least asymptotically. who said values of
w> qualities are specifically dependent?

If the I have a blood sample taken from a patient. I then remove a small
amount to measure it's liquidity. Is the original sample the same sample, with
a SDC L? Or do we have new Ls.

Besides which is liquidity not a feature of blood, the stuff in general,
rather than a specific sample of it?

Phil

[Phillip Lord]

>>>>> "Alan" == Alan Ruttenberg <alanrut...@gmail.com> writes:

Alan> On Jul 9, 2008, at 9:40 PM, Matthew Pocock wrote:

>> So, this is possibly the root of the confusion. People expect to be able
>> to say things like 'the height of Matthew is 5ft7'

Alan> Except that it is exceedingly improbable that the height of Matthew is
Alan> 5ft7. It is almost certainly a bit more or less, and variable over
Alan> time.

Then we not only need to know about values, but measurements, means,
distributions, models of distributions, averages and variances.

Alan> That instance of height measurement information artifact
Alan> has_value 67 in_units inches of_dimension length

Alan> Some work to do to put this into an ontology to share, product of
Alan> http://neurocommons.org/page/First_IAO_workshop

>> or that I am the same height as my sister.

Alan> Certainly not true. But perhaps some theory of this if you deal in
Alan> approximate equality.

If height is a continuous quality which can have any value expressible by an
integer, then equality can only ever be approximate, usually to the resolution
of the measurement. We use the words "same" or "equal" because they are
useful, though, and we still need to.

However it could be true that a measurement of

Alan> each yields an equivalent measurement information artifact. Such an
Alan> artifact, if in the same units, or if transformed by some process into
Alan> another with the same units, could have a value that could be
Alan> compared.

>> We can't describe in any meaningful way how my actual height value (of
>> 5ft7) is related to the random variable that is observations of my
>> height.

Alan> Actually, we probably can't know the actual height, but will do OK
Alan> with the observations (measurements) as described above.

How is the "real height" related to the statistical "real average"?

>> science is about quantitative models of 'values'. These 'values' are at
>> least as 'real' as the continuants and qualities that BFO recognises

Alan> Agreed. We recognize them as information artifacts, which are
Alan> generically dependent continuants. Such entities inhere in such as
Alan> dram, paper and ink, LCD screens, and brains.

So, for every quality or SDC, we need to have an equivalent term under the GDC
which describes measurements of those? Along with an equivalent term
describing the mathematical model of these. Along with an equivalent term
describing the "real average" which is the statistical model.

I think we are going to end up with a rather complex ontology if we follow
this line.

Phil

[Marijke Keet]

As for you examples, Alan, how are things proceeding on, say, a separate
measurement ontology that then can be linked to bfo for the
bioinformatician implementers? Then we can keep the two things separate,
or, as in dolce, have a separate branch in the ontology.

>> science is about quantitative models
>>
>> of 'values'. These 'values' are at least as 'real' as the continuants
>> and
>>
>> qualities that BFO recognises
>>
>
> Agreed. We recognize them as information artifacts, which are
> generically dependent continuants. Such entities inhere in such as
> dram, paper and ink, LCD screens, and brains.

I'm fine with the GDCs, but this still leaves open good, realistic,
examples for SDCs, in the context of qualities, I presume.
Would the following be BFO-allowed too, given a basic notion of specific
dependence? Assume that we do not have brain/body transplants, then
brain is an essential part of human. for this example, each instance of
brain must always be related to the same instance of human.
coincidentally, it is also the other way around, where each instance of
human must have as part the same instance of brain during its lifetime.

>
>> - after all, they are the only things we have any possibility of
>> actually observing,
>>
>
> Oops - it is the qualities we observe. The results of the observations
> are the values, which are information artifacts.

"information artifacts"... in computer science that [the observed
values] is just *data* (cf. structured data that is bumped up to be
"information" and then the next step with analysis & patterns et al we
get to "knowledge"). but this as an aside.

best,
marijke

[Alan Ruttenberg]

On Jul 10, 2008, at 11:28 AM, Phillip Lord wrote:

"Alan" == Alan Ruttenberg <alanrut...@gmail.com> writes:

  Alan> On Jul 9, 2008, at 9:40 PM, Matthew Pocock wrote:

So, this is possibly the root of the confusion. People expect to be able

to say things like 'the height of Matthew is 5ft7'

  Alan> Except that it is exceedingly improbable that the height of Matthew is

  Alan> 5ft7. It is almost certainly a bit more or less, and variable over

  Alan> time.

Then we not only need to know about values, but measurements, means,

distributions, models of distributions, averages and variances.

An ontology of the sort that is useful for quantitative experiments will need these. All of these are information artifacts, some of which are produced by obi:data transformations.

  Alan> That instance of height measurement information artifact

  Alan>   has_value 67 in_units inches of_dimension length

  Alan> Some work to do to put this into an ontology to share, product of

  Alan> http://neurocommons.org/page/First_IAO_workshop

or that I am the same height as my sister.

  Alan> Certainly not true. But perhaps some theory of this if you deal in

  Alan> approximate equality. 

If height is a continuous quality which can have any value expressible by an

integer, then equality can only ever be approximate, usually to the resolution

of the measurement. We use the words "same" or "equal" because they are

useful, though, and we still need to. 

Sure. My premise has been that our work on ontology is towards being able to represent clearly what was meant when something is said, so that in a communication between computers there is no confusion about what is meant. 

However it could be true that a measurement of

  Alan> each yields an equivalent measurement information artifact. Such an

  Alan> artifact, if in the same units, or if transformed by some process into

  Alan> another with the same units, could have a value that could be

  Alan> compared.

We can't describe in any meaningful way how my actual height value (of

5ft7) is related to the random variable that is observations of my

height.

  Alan> Actually, we probably can't know the actual height, but will do OK

  Alan> with the observations (measurements) as described above.

How is the "real height" related to the statistical "real average"? 

it is a proxy for the real height (is_proxy_for will be part of RO2).

The proxy relationship captures the relationship between things where measurement of one is taken to be measurement of another. We read the average (measure) and take it for what we would get by actual measuring of the actual thing. We understand, in each is_proxy_for step that there are all sorts of caveats - exactly the kind of caveats a good experimentalist/statistician understands. Often in representation of scientific result, these caveats have no place. With is_proxy_of we know where to ask probing questions. 

science is about quantitative models of 'values'. These 'values' are at

least as 'real' as the continuants and qualities that BFO recognises

  Alan> Agreed. We recognize them as information artifacts, which are

  Alan> generically dependent continuants. Such entities inhere in such as

  Alan> dram, paper and ink, LCD screens, and brains.

So, for every quality or SDC, we need to have an equivalent term under the GDC

which describes measurements of those?

They exist. Whether we need to name them, or record them as anonymous expressions is a separate matter.

Along with an equivalent term describing the mathematical model of these.

A mathematical model is a different sort of thing. The values (information) may be the results of modeling.

Along with an equivalent term describing the "real average" which is the statistical model. 

"Real average"?

I think we are going to end up with a rather complex ontology if we follow this line. 

Ontology is complicated. Don't confuse what you will need to actually record with what the ontology has. In fact, I believe that at then end of doing good ontology we will land up with a much simpler situation, in practice, than we have now  for a given level of desired quality. That's because we will be able to be clear about which inferences can be made safely and correctly, be able to say compactly only what is necessary, and have the rest inferred.

-Alan

Phil

[lampucerka]

cool tricks, alan. we probably can't know the actual height, so instead
of saying 'the hight is x y-units', we say 'we have made an observation
which yielded an information artifact which is a measurement of the
actual height and has value x in y-units'.

issue 1: can we know the actual value of the measurement?

issue 2: seems like you suggest a departure from the 'represent the
reality directly' dogma; instead of making claims about entities (e.g.,
the height of matthew), we make statements about our observations of the
entities (e.g., our measurements of matthew's height).

issue 3: we probably can't know the actual height of matthew (or
maybe: we can know the height -- it's *the* height of matthew -- but
not its value?); amazing that we can know for sure that there are, as
real entities on their own, specifically dependent continuants, among
them qualities such as matthew's height, roles, functions, etc.
according to your model, we measure the height (which is real) and
obtain an information artefact (which is real) that has a value (which
is real?), but does the height also have a value (which is real) which
the value of the information artefact serves as a proxy of, or is there
no such entity? if there is, is it not desirable to be able to
represent it, even if we probably can't know it? if we believe it is
there, the fact that we can't represent it in the chosen framework is
evidence for a weakness in the model, no?

>
>> Also, we can't meaningfully seperate out the height from its
>>
>> representation in feet and inches, or in millimeters.
>>
>
> The height itself would be a determinate quality instance. It's
> specific value is elusive. (Cocos rule: any measurement is wrong) One
> can have classes of determinate quality instances that are in certain
> ranges:
>
> Class: Determinate heights between 5'6 and 5'8
> Class: Determinate heights between 5'6.5 and 5'7.5
>
> Likely that the instance of determinate height quality instance is
> instance_if of those two.
>
> The "between 5'6.5 and 5'7.5" is part of how we could name that class.
> However that class could have other names, ones which mentioned the
> centimeters.
>
> A representation is an information artifact. So we can separate that
> from the determinate quality instance this, and say it is "about" the
> quality instance.

ok. then a representation of an information artifact -- a
representation of a measurement, as you discuss above, instead of a
representation of the actual value -- is an information artifact about
an information artifact. so we start departing from representing the
reality as it is, and represent it indirectly via intermediate artefacts.

vQ

[Alan Ruttenberg]

On Jul 10, 2008, at 11:43 AM, Marijke Keet wrote:
>
> As for you examples, Alan, how are things proceeding on, say, a
> separate
> measurement ontology that then can be linked to bfo for the
> bioinformatician implementers? Then we can keep the two things
> separate,
> or, as in dolce, have a separate branch in the ontology.

The information artifact ontology will deal with the upper level
entities such as datum, is_measurement_of, relations to units and
dimensions, etc.

OBI will use import this ontology and then use the terms in it to
more specifically define protocol applications, instruments, specific
sorts of measurements and the like. OBI itself, uses other foundry
ontologies wherever possible, so that we don't have obi:cell - rather
it uses the term for the cell ontology.

>
>>> science is about quantitative models
>>>
>>> of 'values'. These 'values' are at least as 'real' as the
>>> continuants
>>> and
>>>
>>> qualities that BFO recognises
>>>
>>
>> Agreed. We recognize them as information artifacts, which are
>> generically dependent continuants. Such entities inhere in such as
>> dram, paper and ink, LCD screens, and brains.
>
> I'm fine with the GDCs, but this still leaves open good, realistic,
> examples for SDCs, in the context of qualities, I presume.
> Would the following be BFO-allowed too, given a basic notion of
> specific
> dependence? Assume that we do not have brain/body transplants, then
> brain is an essential part of human. for this example, each
> instance of
> brain must always be related to the same instance of human.
> coincidentally, it is also the other way around, where each
> instance of
> human must have as part the same instance of brain during its
> lifetime.

I think that this sort of question is difficult because it impinges
on a controversial area of ontology.
However you have made it somewhat easy by not mentioning a dependent
continuant anywhere.

What BFO provides is the ability to say, when you mean it, that there
are dependent continuants that are specifically dependent, i.e.
existentially dependent on their bearer. Outside BFO one must
determine what the bearers are, what the dependent continuants are,
and whether they are specific or generic, or something else altogether.

>>> - after all, they are the only things we have any possibility of
>>> actually observing,
>> Oops - it is the qualities we observe. The results of the
>> observations
>> are the values, which are information artifacts.
>
> "information artifacts"... in computer science that [the observed
> values] is just *data* (cf. structured data that is bumped up to be
> "information" and then the next step with analysis & patterns et al we
> get to "knowledge"). but this as an aside.

Every datum is an information artifact. Not all information artifacts
are data. For instance, the wording of the poem "Leaves of grass".

Regards,
Alan

[lampucerka]

Alan Ruttenberg wrote:
>
> Every datum is an information artifact. Not all information artifacts
> are data. For instance, the wording of the poem "Leaves of grass".
>

perhaps you should start with this:
http://plato.stanford.edu/entries/information-semantic/

vQ

[Alan Ruttenberg]

You suggest the impossible - I have already started ;-)

But thanks for the pointer - I will have a look. Do you have in mind
some particular section that you think speaks to specifics of what
has been proposed, either in favor of, or in contradiction to?

-Alan

[lampucerka]

Alan Ruttenberg wrote:
> You suggest the impossible - I have already started ;-)
>
> But thanks for the pointer - I will have a look. Do you have in mind
> some particular section that you think speaks to specifics of what
> has been proposed, either in favor of, or in contradiction to?
>
>

the part i had in mind was this:

"
*The General Definition of Information (GDI):*
σ is an instance of information, understood as semantic content, if and
only if:

(GDI.1) σ consists of one or more /data/;

(GDI.2) the data in σ are /well-formed/;

(GDI.3) the well-formed data in σ are /meaningful/.

"

if we think of information as consisting (rather than being) data, then
you should perhaps say that no information *is* data, even if *all*
information consists of data. (or maybe information consists of data
and can also be data itself?)

if we instead interpret the above as saying that information *is* data,
only that some additional properties are considered (e.g., gdi.3), then
clearly *all* information is data.

anyway, your claim that not all information is data seemed to me an
ungrounded, ad hoc proposition. just that.

vQ

[Alan Ruttenberg]

"value of the measurement".

If you mean do we know what is produced in a measurement process?
Answer yes: A datum of the sort I have described. The "actual value"
is, by definition, that part of the datum which is the object of the
relation I labeled "value:"

If you mean something other than this could you clarify for me what
you are referring to?

> issue 2: seems like you suggest a departure from the 'represent the
> reality directly' dogma; instead of making claims about entities
> (e.g.,
> the height of matthew), we make statements about our observations
> of the
> entities (e.g., our measurements of matthew's height).

Data is part of reality, but importantly a different part of reality.
There is Matthew's height, and there are also measurements of
Matthew's height. Both real. Moreover, there is the "audit trail to
reality" in our representation of measurement data, is_measurement_of
relates data to a thing in reality. In OBI, a measurement datum
is_output_of and OBI:assay. Thus we have connections from the data to
the process in which it was created, and to that which it is about.
Moreover with the is_proxy_for relationship we can account for such
things as the fact that a weight scale isn't actually measuring the
mass directly, but for instance the current through a stress
sensitive resistor (that current being a proxy for the weight) and
that value undergoes a data transformation on the weigh to reporting
the weight in grams. (a different data transformation when reporting
mass in ounces).

We can make statements about either Matthew's height or the
measurement of it but we need to understand what we are saying and
take responsibility for making the statements. In the GO,
measurements, or sets of measurements are "believed" and promoted to
statements about reality. They may be wrong, but they have taken
responsibility for their statements by making corrections when
necessary, presenting a consistent view, etc. When I say "they" it is
important to note that the they, in this, case, is a large body of
competent scientists who represent that the GO is their consensus.

In recording the results of experiments we may not wish to
immediately make such commitments. However, by evaluating many
experiments scientists come to think they have exposed reality then
we represent it as such.

> issue 3: we probably can't know the actual height of matthew (or
> maybe: we can know the height -- it's *the* height of matthew -- but
> not its value?); amazing that we can know for sure that there are, as
> real entities on their own, specifically dependent continuants, among
> them qualities such as matthew's height, roles, functions, etc.
> according to your model, we measure the height (which is real) and
> obtain an information artefact (which is real) that has a value (which
> is real?), but does the height also have a value (which is real) which
> the value of the information artefact serves as a proxy of, or is
> there
> no such entity?

BFO says there is such an entity.

> if there is, is it not desirable to be able to represent it, even
> if we probably can't know it?

We can represent it. For example, an instance of PATO:0000119 that
inheres in Matthew at a certain time is such an entity.

> if we believe it is there, the fact that we can't represent it in
> the chosen framework is
> evidence for a weakness in the model, no?

It would be, if it were the case.

>>> Also, we can't meaningfully seperate out the height from its
>>>
>>> representation in feet and inches, or in millimeters.
>>>
>>
>> The height itself would be a determinate quality instance. It's
>> specific value is elusive. (Cocos rule: any measurement is wrong) One
>> can have classes of determinate quality instances that are in certain
>> ranges:
>>
>> Class: Determinate heights between 5'6 and 5'8
>> Class: Determinate heights between 5'6.5 and 5'7.5
>>
>> Likely that the instance of determinate height quality instance is
>> instance_if of those two.
>>
>> The "between 5'6.5 and 5'7.5" is part of how we could name that
>> class.
>> However that class could have other names, ones which mentioned the
>> centimeters.
>>
>> A representation is an information artifact. So we can separate that
>> from the determinate quality instance this, and say it is "about" the
>> quality instance.
>
> ok. then a representation of an information artifact -- a
> representation of a measurement, as you discuss above, instead of a
> representation of the actual value

Using the wording "actual value" makes this difficult to parse and
somewhat confusing. Instead we have entities which are the
determinate quality instances - they represent the real thing.

We also have a datum which is_measurent_of such an instance.

Although, in language, we sometimes use the word "represent" in both
these case, these are different relationships.

> -- is an information artifact about
> an information artifact. so we start departing from representing the
> reality as it is, and represent it indirectly via intermediate
> artefacts.

As I've said. We choose what we represent. The ontology gives us the
means to be precise about what we are representing.

>
> vQ
>
>
> >

[David Decraene]

I'm probably opening the can of worms even further:
but here is a brief summary how LinKBase (upper level is based upon
BFO) considers UNITs, values and properties:

PROPERTY
equals BFO quality, inheres in independent continuants

PROPERTY STATE
PROPERTY with given value restriction (qualitative or quantitive):
normal length, increased length, 5cm length.
red colour
also subsumed by (specific) dependent continuant, debate on whether
they are child (ISA) of the property is ongoing

UNIT
specific type of PROPERTY STATE (ISA KIND-OF-QUANTITY STATE ISA
PROPERTY STATE)
interpretation: each UNIT is a standardized kind-of-quantity (ignoring
language/syntax connotation that 'METER' is abstract (1 meter),
interpreting (focusing on semantics instead) as one given standardized
value of length):

* meter is a specific & quantified length state
* kilogram is a specific & quantified weight state.

PROPERTY STATE's can be related to other property states with a
relation called 'HAS-REALISED-P-STATE.
consider the possiblity to use qualified cardinality restrictions with
this relation, then it might be possible to say (not entirely
expressible in OWL):
<CENTIMETER> HAS-REALISED-P-STATE METER, VALUE,(=0.01)
<POUND> HAS-REALISED-P-STATE GRAM, VALUE,(=453.59237)

Hope this contributes in any way
David Decraene

[Pierre Grenon]

Hiya Marijke

>> Agreed. We recognize them as information artifacts, which are
>> generically dependent continuants. Such entities inhere in such as
>> dram, paper and ink, LCD screens, and brains.
>
> I'm fine with the GDCs, but this still leaves open good, realistic,
> examples for SDCs, in the context of qualities, I presume.
> Would the following be BFO-allowed too, given a basic notion of specific
> dependence? Assume that we do not have brain/body transplants, then
> brain is an essential part of human. for this example, each instance of
> brain must always be related to the same instance of human.
> coincidentally, it is also the other way around, where each instance of
> human must have as part the same instance of brain during its lifetime.

i'm not sure whether BFO forces any decision on this

but to relate to the discussion of SDC, make sure you understand that
for BFO the brain falls under IndependentContinuant as it is a (fiat,
I guess) part of the body, it does not inhere in it. the brain is not
an SDC in BFO. Don't take SDC to mean the class of entities that are
(specifically) dependent on something. There are claims of existential
dependence in BFO beyond the category of so-called
DependentContinuant, e.g. between processes and their particiapants
although processes are not SDC nor DC for that matter.

inherence is not parthood and your discussion of temporal relations breaks here

BFO would allow that an SDC is shorter lived than its bearer, but it
would not allow an SDC to be longer lived than its bearer or that the
two lifespan only overlap. It would allow that both an SDC and its
bearer are temporally coincidental. In any event, the SDC lifespan has
to be contained (properly or not) in its bearer's lifespan.

An example of temporal coincidence could be a tomato and its color
while an example of proper temporal subsumption (bearer's lifespan
contains the SDC's lifespan) could be a tomato and its price.

pierre

[Pierre Grenon]

>
> eager to find 'bfo has no value' hashed by google...

it will only become a concern when cited in a wikipedia article

>>> that would be part of GROSS, the Great Realist
>>> Ontological Suggested Scheme... I'm just making this up obviously (I
>>> say that for Germans or people who have stayed too long in Germany and
>>> who like me can't spot what's supposed to be a joke) I mean it's part
>>> of the big picture, it has to be attached to BFO, it's just not in
>>> BFO. I really think we went through this
>>>
>>>
>
> just moved and already complaining about germans? they paid your bread,
> didn't they?

I was on a diet

> one (very serious, like others) way to solve the 'value of liquidity'
> problem is to call for quales (a la dolce): L1, the liquidity of B1, is
> an entity (a quality) distinct from L2, the liquidity of B2, even if
> they have the same liquidity quale, i.e., a position in the space of
> values for liquidity qualities. ('have'? 'the same'? 'quale'? please
> read, e.g., [1] before complaining. they actually say 'so we
> distinguish', 'so when we say', etc., without too much insistence on
> that that's how R is, bless them for that.)

Yes, the mechanism is very similar for all I can remeber of DOLCE.

I've said in the past already, i think, that you could attach the
conceptual space theory of DOLCE to BFO although the favored treatment
might be slightly dissimilar and based also on considerations
developed by Ingvar Johanson (mentioned somewhere else in this
thread).

In the context of OBI or OBO, say, we would want in fact some thorough
mechanism to deal wth quantities and so on and an able theory of
units. I seem to remember it's more or less the idea of OBO to have it
as a module of its own, so to speak. I haven't looked at it for a
while...

p

[lampucerka]

Pierre Grenon wrote:
>> eager to find 'bfo has no value' hashed by google...
>>
>
> it will only become a concern when cited in a wikipedia article
>
>

am writing it *now*

>>>> that would be part of GROSS, the Great Realist
>>>> Ontological Suggested Scheme... I'm just making this up obviously (I
>>>> say that for Germans or people who have stayed too long in Germany and
>>>> who like me can't spot what's supposed to be a joke) I mean it's part
>>>> of the big picture, it has to be attached to BFO, it's just not in
>>>> BFO. I really think we went through this
>>>>
>>>>
>>>>
>> just moved and already complaining about germans? they paid your bread,
>> didn't they?
>>
>
> I was on a diet
>
>

you looked pretty hungry watching chicks ;)

>> one (very serious, like others) way to solve the 'value of liquidity'
>> problem is to call for quales (a la dolce): L1, the liquidity of B1, is
>> an entity (a quality) distinct from L2, the liquidity of B2, even if
>> they have the same liquidity quale, i.e., a position in the space of
>> values for liquidity qualities. ('have'? 'the same'? 'quale'? please
>> read, e.g., [1] before complaining. they actually say 'so we
>> distinguish', 'so when we say', etc., without too much insistence on
>> that that's how R is, bless them for that.)
>>
>
> Yes, the mechanism is very similar for all I can remeber of DOLCE.
>
> I've said in the past already, i think, that you could attach the
> conceptual space theory of DOLCE to BFO although the favored treatment
> might be slightly dissimilar and based also on considerations
> developed by Ingvar Johanson (mentioned somewhere else in this
> thread).
>
> In the context of OBI or OBO, say, we would want in fact some thorough
> mechanism to deal wth quantities and so on and an able theory of
> units. I seem to remember it's more or less the idea of OBO to have it
> as a module of its own, so to speak. I haven't looked at it for a
> while...
>

alan is certainly the authority here. see his recent comments here.

vQ

[Marijke Keet]

Alan Ruttenberg wrote:

>
> On Jul 10, 2008, at 11:43 AM, Marijke Keet wrote:
>
>>
>> As for you examples, Alan, how are things proceeding on, say, a
>> separate
>> measurement ontology that then can be linked to bfo for the
>> bioinformatician implementers? Then we can keep the two things
>> separate,
>> or, as in dolce, have a separate branch in the ontology.
>
>
> The information artifact ontology will deal with the upper level
> entities such as datum, is_measurement_of, relations to units and
> dimensions, etc.

this is developed de novo?

given the premise, both are specifically dependent. (omitting that info
from the example was by intention, because I was fishing if you would
have said yes or no instead of avoiding answering the question).
you have a generic dependence with, say, your heart: you have to have
one, but not necessarily always the same one during your lifespan.
likewise, for a (working) human heart: it must be part of a human, but
not necessarily the same one.

best,
marijke

[Alan Ruttenberg]

On Jul 10, 2008, at 12:18 PM, lampucerka wrote:
> Alan Ruttenberg wrote:
>> You suggest the impossible - I have already started ;-)
>>
>> But thanks for the pointer - I will have a look. Do you have in mind
>> some particular section that you think speaks to specifics of what
>> has been proposed, either in favor of, or in contradiction to?
> the part i had in mind was this:
>
> "
> *The General Definition of Information (GDI):*
> σ is an instance of information, understood as semantic content, if
> and
> only if:
>
> (GDI.1) σ consists of one or more /data/;
>
> (GDI.2) the data in σ are /well-formed/;
>
> (GDI.3) the well-formed data in σ are /meaningful/.
>
> "
>
> if we think of information as consisting (rather than being) data,
> then
> you should perhaps say that no information *is* data, even if *all*
> information consists of data. (or maybe information consists of data
> and can also be data itself?)

First, to clarify: I am not speaking of an information (mass noun)
ontology. I am speaking of an Information Artifact (count noun)
Ontology.

Now, so that we can communicate effectively, could you give me a
couple of sentences on what you mean when you use the word
"information" and "data"? Probably time to synchronize on what we
mean by our words.

What we mean by datum is the sort of thing one writes on a paper or
stores on a hard disk that is the result of a measurement or some
other recording of fact.

Information artifact's current description is:

> An information entity is a generically dependent continuant that
> originates with a sentient - either by a person thinking/
> communicating, or by a machine that was designed to have a function
> to produce/communicate information.

(I apologize for the current imprecision in our documentation - this
is a work in progress. Suggestions are welcome)

By what we intend, when we speak of information artifact, an example
of an information artifact that is not a datum is the wording of the
poem "Leaves of Grass". It is generically dependent because people
can remember the same exact thing (i.e. the same entity can inhere in
parts of many different brains).

[yes, I know, some will make mistakes - this does not invalidate the
point].

> if we instead interpret the above as saying that information *is*
> data,
> only that some additional properties are considered (e.g., gdi.3),
> then
> clearly *all* information is data.
>
> anyway, your claim that not all information is data seemed to me an
> ungrounded, ad hoc proposition. just that.

I am trying to create definitions, not use existing ones. If I was
suggesting what words or terms mean I might be making a proposition.

>
> vQ
>
>
>
> >

[Phillip Lord]

>>>>> "Alan" == Alan Ruttenberg <alanrut...@gmail.com> writes:

>> Then we not only need to know about values, but measurements, means,
>> distributions, models of distributions, averages and variances.

Alan> An ontology of the sort that is useful for quantitative experiments
Alan> will need these. All of these are information artifacts, some of which
Alan> are produced by obi:data transformations.

Yes, I know. The question I raise, then, is whether we will ever use the
qualities, since we only ever have values for measurements.

>> If height is a continuous quality which can have any value expressible by
>> an integer, then equality can only ever be approximate, usually to the
>> resolution of the measurement. We use the words "same" or "equal" because
>> they are useful, though, and we still need to.

Alan> Sure. My premise has been that our work on ontology is towards being
Alan> able to represent clearly what was meant when something is said, so
Alan> that in a communication between computers there is no confusion about
Alan> what is meant.

Computational ontologies are one way of achieving this, yes, I'd agree.

Alan> Actually, we probably can't know the actual height, but
>> will do OK
Alan> with the observations (measurements) as described above.
>>
>> How is the "real height" related to the statistical "real average"?

Alan> it is a proxy for the real height (is_proxy_for will be part of RO2).

Alan> The proxy relationship captures the relationship between things where
Alan> measurement of one is taken to be measurement of another. We read the
Alan> average (measure) and take it for what we would get by actual
Alan> measuring of the actual thing.

No, you have misunderstood.

Take the question: what is the average height of the male adult population of
the UK.

So, I measure the height of 100 people. I determine that these 100
measurements have a normal distribution (or rather than they approximate the
normal distribution). I then take the mean of these 100 measurements. This
then gives me a value which is a measure of the population average; say 1.60
+/- 30 m.

This measurement is based on by selecting a statistical model which is
appropriate to the data. We have a statistical estimation of the population
average. In almost all cases within science, we can only ever have an
estimation of the average (the exception would be for generated data sets).

But the population average, in this case, is statistical abstraction which we
are estimating. It's also something that you can rarely or never measure. Is
that what BFO is talking about?

Alan> We understand, in each is_proxy_for step that there are all sorts of
Alan> caveats - exactly the kind of caveats a good
Alan> experimentalist/statistician understands. Often in representation of
Alan> scientific result, these caveats have no place. With is_proxy_of we
Alan> know where to ask probing questions.

I would also have worries about your terminology. If I measure your height 10
times and take the average, it's just a measurement of your height not a proxy
for it.

Compare this with measuring the weight of a stone; this is a proxy for
measuring the mass, I think.

>>>>
>>>> science is about quantitative models of 'values'. These 'values' are at
>>>> least as 'real' as the continuants and qualities that BFO recognises
>>
Alan> Agreed. We recognize them as information artifacts, which are
Alan> generically dependent continuants. Such entities inhere in
>> such as
Alan> dram, paper and ink, LCD screens, and brains.

>> So, for every quality or SDC, we need to have an equivalent term under
>> the GDC which describes measurements of those?

Alan> They exist. Whether we need to name them, or record them as anonymous
Alan> expressions is a separate matter.

Everything else that you have said leads me to suspect that you think they do
need to be named.

>> Along with an equivalent term describing the mathematical model of these.

Alan> A mathematical model is a different sort of thing. The values
Alan> (information) may be the results of modeling.

>> Along with an equivalent term describing the "real average" which is the
>> statistical model.

Alan> "Real average"?

>> I think we are going to end up with a rather complex ontology if we
>> follow this line.

Alan> Ontology is complicated. Don't confuse what you will need to actually
Alan> record with what the ontology has. In fact, I believe that at then end
Alan> of doing good ontology we will land up with a much simpler situation,
Alan> in practice, than we have now for a given level of desired quality.

This is a truism; simple is good right?

Phil

[Marijke Keet]

Hi Pierre,

>>>Agreed. We recognize them as information artifacts, which are
>>>generically dependent continuants. Such entities inhere in such as
>>>dram, paper and ink, LCD screens, and brains.
>>>
>>>
>>I'm fine with the GDCs, but this still leaves open good, realistic,
>>examples for SDCs, in the context of qualities, I presume.
>>Would the following be BFO-allowed too, given a basic notion of specific
>>dependence? Assume that we do not have brain/body transplants, then
>>brain is an essential part of human. for this example, each instance of
>>brain must always be related to the same instance of human.
>>coincidentally, it is also the other way around, where each instance of
>>human must have as part the same instance of brain during its lifetime.
>>
>>
>
>i'm not sure whether BFO forces any decision on this
>
>but to relate to the discussion of SDC, make sure you understand that
>for BFO the brain falls under IndependentContinuant as it is a (fiat,
>I guess) part of the body, it does not inhere in it. the brain is not
>an SDC in BFO. Don't take SDC to mean the class of entities that are
>(specifically) dependent on something. There are claims of existential
>dependence in BFO beyond the category of so-called
>DependentContinuant, e.g. between processes and their particiapants
>although processes are not SDC nor DC for that matter.
>
>inherence is not parthood and your discussion of temporal relations breaks here
>
>

true, and I never intended that, I was pondering how far I could strech
the generalization of the essential vs mandatory participation that
captures specific and generic dependence in a more general sense, not
only applying it to inheres_in or part_of (and its corresponding general
formalization we take as basis---only after that we apply it to
part-whole relations)

>BFO would allow that an SDC is shorter lived than its bearer, but it
>would not allow an SDC to be longer lived than its bearer or that the
>two lifespan only overlap. It would allow that both an SDC and its
>bearer are temporally coincidental. In any event, the SDC lifespan has
>to be contained (properly or not) in its bearer's lifespan.
>
>

clearly, we have exactly those constraints on the life cycles as well
(with 4 theoretical options).

best,
marijke

[Alan Ruttenberg]

On Jul 10, 2008, at 4:39 PM, Marijke Keet wrote:

> Alan Ruttenberg wrote:
>> On Jul 10, 2008, at 11:43 AM, Marijke Keet wrote:
>>> As for you examples, Alan, how are things proceeding on, say, a
>>> separate
>>> measurement ontology that then can be linked to bfo for the
>>> bioinformatician implementers? Then we can keep the two things
>>> separate,
>>> or, as in dolce, have a separate branch in the ontology.
>> The information artifact ontology will deal with the upper level
>> entities such as datum, is_measurement_of, relations to units and
>> dimensions, etc.
>
> this is developed de novo?

Yes. Want to help?

both what?

> are specifically dependent. (omitting that info from the example
> was by intention, because I was fishing if you would have said yes
> or no instead of avoiding answering the question).
> you have a generic dependence with, say, your heart: you have to
> have one, but not necessarily always the same one during your
> lifespan. likewise, for a (working) human heart: it must be part of
> a human, but not necessarily the same one.

I think the sense of dependent you are using is not the one BFO uses.
This is not uncommon - dependence is a general terms and in BFO it is
used for one sort of dependence. That sense is to relate an
independent continuant and a dependent continuant. Dependent
continuants are things like functions, qualities, dispositions,
information - things which don't make sense as existing except by
virtue of something else (necessary but not sufficient). So there can
be no shape without the shape being the shape of something.

There are other kinds of existential dependence - for instance the
dependence of a boundary on the thing it is a boundary of, or the
existential dependence of a process on its participants. But neither
of these kinds of dependence are what is meant by the word in
"Dependent continuant".

In your example the two entities are the brain and the body. Each of
these, in BFO, are independent continuants.

Regards,
Alan

[Alan Ruttenberg]

On Jul 10, 2008, at 4:49 PM, Phillip Lord wrote:

>
>>>>>> "Alan" == Alan Ruttenberg <alanrut...@gmail.com> writes:
>
>>> Then we not only need to know about values, but measurements, means,
>>> distributions, models of distributions, averages and variances.
>
> Alan> An ontology of the sort that is useful for quantitative
> experiments
> Alan> will need these. All of these are information artifacts,
> some of which
> Alan> are produced by obi:data transformations.
>
> Yes, I know. The question I raise, then, is whether we will ever
> use the
> qualities, since we only ever have values for measurements.

Indeed we may never instantiate them. However it is important, for
the coherence of the ontology that they present, and may be necessary
for certain inferences to be correctly made, by virtue of the axioms
that mention the classes.

Ok. Different question.

> So, I measure the height of 100 people. I determine that these 100
> measurements have a normal distribution (or rather than they
> approximate the
> normal distribution). I then take the mean of these 100
> measurements. This
> then gives me a value which is a measure of the population average;
> say 1.60
> +/- 30 m.
>
> This measurement is based on by selecting a statistical model which is
> appropriate to the data. We have a statistical estimation of the
> population
> average. In almost all cases within science, we can only ever have an
> estimation of the average (the exception would be for generated
> data sets).
>
> But the population average, in this case, is statistical
> abstraction which we
> are estimating. It's also something that you can rarely or never
> measure. Is
> that what BFO is talking about?

I am not sure that I think something as the average height of a
certain population is something real of the same sort that the height
of an individual is. I am confident that the information entities: a)
The mean measured height of 100 people b) the mean of three
measurements of one person's height. Both of these are the result of
data transformations of multiple datums that are measurements of the
same sort of thing.

So, to summarize, what you call the "real average" (technically the
average) is something I would give ontological status to other than
it being an idea.

> Alan> We understand, in each is_proxy_for step that there are all
> sorts of
> Alan> caveats - exactly the kind of caveats a good
> Alan> experimentalist/statistician understands. Often in
> representation of
> Alan> scientific result, these caveats have no place. With
> is_proxy_of we
> Alan> know where to ask probing questions.
>
> I would also have worries about your terminology. If I measure your
> height 10
> times and take the average, it's just a measurement of your height
> not a proxy
> for it.

Perhaps you are right - it is a bit too close. The canonical case
for proxy relation when it was conceived was when representing a
chromium release assay (a kind of cytoxicity assay). In that case
radioactivity is a proxy for the number of cells that died. Will
think more about that.

> Compare this with measuring the weight of a stone; this is a proxy for
> measuring the mass, I think.

Yes. Note weight and height are both qualities. A recent question
asked is what the domain and range of the is_proxy_for relation for.
On first glance it would seem to be entity, in both cases. Needs some
more thought. Got any cases that might suggest otherwise?

Complex is bad, right?

-Alan

>
> Phil
>
> >

[Phillip Lord]

>>>>> "Alan" == Alan Ruttenberg <alanrut...@gmail.com> writes:

>>>> Then we not only need to know about values, but measurements, means,
>>>> distributions, models of distributions, averages and variances.
>>
Alan> An ontology of the sort that is useful for quantitative
>> experiments
Alan> will need these. All of these are information artifacts,
>> some of which
Alan> are produced by obi:data transformations.
>>
>> Yes, I know. The question I raise, then, is whether we will ever use the
>> qualities, since we only ever have values for measurements.

Alan> Indeed we may never instantiate them.

If they are not instantiable, then they have no instances in reality. If they
have no instances in reality, then they should not be in BFO, which is my
question.

Alan> However it is important, for the coherence of the ontology that they
Alan> present, and may be necessary for certain inferences to be correctly
Alan> made, by virtue of the axioms that mention the classes.

Such as...

>> No, you have misunderstood.
>>
>> Take the question: what is the average height of the male adult
>> population of the UK.

Alan> Ok. Different question.

My apologies. I should carried with the same example.

>> So, I measure the height of 100 people. I determine that these 100
>> measurements have a normal distribution (or rather than they approximate
>> the normal distribution). I then take the mean of these 100 measurements.
>> This then gives me a value which is a measure of the population average;
>> say 1.60 +/- 30 m.
>>
>> This measurement is based on by selecting a statistical model which is
>> appropriate to the data. We have a statistical estimation of the
>> population average. In almost all cases within science, we can only ever
>> have an estimation of the average (the exception would be for generated
>> data sets).
>>
>> But the population average, in this case, is statistical abstraction
>> which we are estimating. It's also something that you can rarely or never
>> measure. Is that what BFO is talking about?

Alan> I am not sure that I think something as the average height of a
Alan> certain population is something real of the same sort that the height
Alan> of an individual is.

It's a quality of an agregate I think. The same questions apply to averages
of height of a person I think.

Alan> I am confident that the information entities: a) The mean measured
Alan> height of 100 people b) the mean of three measurements of one person's
Alan> height. Both of these are the result of data transformations of
Alan> multiple datums that are measurements of the same sort of thing.

Again, these are estimates of an average. In case a) where we measure 100
people's height to get a population average, what we really want is to measure
the entire population. In case b) where we want to control against variability
in height over time (for example), we want to measure over all time.

The actual average you don't measure, you only estimate. This is why you have
error bars.

>> I would also have worries about your terminology. If I measure your
>> height 10 times and take the average, it's just a measurement of your
>> height not a proxy for it.

Alan> Perhaps you are right - it is a bit too close. The canonical case for
Alan> proxy relation when it was conceived was when representing a chromium
Alan> release assay (a kind of cytoxicity assay). In that case radioactivity
Alan> is a proxy for the number of cells that died. Will think more about
Alan> that.

I agree. I don't think this type of relationship should be confused with a
measurement.

>> Compare this with measuring the weight of a stone; this is a proxy for
>> measuring the mass, I think.

Alan> Yes. Note weight and height are both qualities. A recent question
Alan> asked is what the domain and range of the is_proxy_for relation for.
Alan> On first glance it would seem to be entity, in both cases. Needs some
Alan> more thought. Got any cases that might suggest otherwise?

Can you restate the question? A case that suggests otherwise would have to be
a counter example; given that all things are entities, it's not possible to
give a counter example.

Phil

[Cristian Cocos]

On Jul 10, 2008, at 12:43 PM, Marijke Keet wrote:

> I'm fine with the GDCs, but this still leaves open good, realistic,
> examples for SDCs, in the context of qualities, I presume.

What I could never explain myself is how anyone can be fine with *SDCs*.
SDCs are solely necessary in virtue of BFO allegiance to tropes: if one
drops tropes, one has to dump SDCs.

So why not dump SDCs? That would make things much simpler: This pencil I am
holding has a length, a *particular* length (say 7 cm). This is an
individual in OWL terminology. This length is the same as the length of the
pen on the desk--same length particular/instance. This is something that
GDCs accommodate very well. No need for further length tropes that, in turn,
have "values" etc. No need to discern between the length of the pencil and
the length of the pen by postulating ("recognizing") another entity (length
tropes) that are "specific" and "unique" etc., and that "inhere" in ...
wherever they may inhere.

Yes, but what sort of a horrible person does not find "the particular
redness of this shirt" as compelling evidence in favor of the existence of
tropes? I, for one, ain't. And so do many others (namely mainstream
contemporary philosophy).

Of course, dumping tropes means having to wrestle with all the pro-trope
literature, and with all the passions surrounding it. We'd be assaulted with
replies of the type "read this, this, this, and that" -- to which we can
also reply with "read this, this, this, and that" as proof to the contrary.

At any rate, what matters most is that BFO curators *themselves* decided to
include GDCs in BFO--which, as far as I sense, they regard as a grey area
(which, again, they shouldn't). Hence, we cannot do *without* GDCs. Why not
work *only* with GDCs then? A possible answer is that we have to stick to
trope orthodoxy due to different (and higher) metaphysical reasons. But
while admitting GDCs is definitely a departure from trope orthodoxy, there
are good indications that tropes do not even fulfill the aims that prompted
their introduction in the first place (explain "similarity" among others).

Hence why not GDCs all the way?

C

[lampucerka]

Cristian Cocos wrote:
> On Jul 10, 2008, at 12:43 PM, Marijke Keet wrote:
>
>
>> I'm fine with the GDCs, but this still leaves open good, realistic,
>> examples for SDCs, in the context of qualities, I presume.
>>
>
> What I could never explain myself is how anyone can be fine with *SDCs*.
> SDCs are solely necessary in virtue of BFO allegiance to tropes: if one
> drops tropes, one has to dump SDCs.
>
> So why not dump SDCs? That would make things much simpler: This pencil I am
> holding has a length, a *particular* length (say 7 cm). This is an
> individual in OWL terminology. This length is the same as the length of the
> pen on the desk--same length particular/instance. This is something that
> GDCs accommodate very well. No need for further length tropes that, in turn,
> have "values" etc. No need to discern between the length of the pencil and
> the length of the pen by postulating ("recognizing") another entity (length
> tropes) that are "specific" and "unique" etc., and that "inhere" in ...
> wherever they may inhere.
>
>

the length already is a bit of a mysterious world; what happens to the
length of the pencil when it gets shorter?
what happens to a particular length when all objects having this length
cease to exist, and then new objects of this (?) length appear? does
the length persist through periods when there is nothing for it to
inhere in, and if not, how can it be the same length before and after?
and if it is not (ie, if *they* are not) the same length, does a piece
of rubber have a distinct length than the original one after i extend it
and then let it shrink, even if i'd say it has the same length?

oh my, it all get's so simple once you deny values.

> Yes, but what sort of a horrible person does not find "the particular
> redness of this shirt" as compelling evidence in favor of the existence of
> tropes? I, for one, ain't. And so do many others (namely mainstream
> contemporary philosophy).
>
> Of course, dumping tropes means having to wrestle with all the pro-trope
> literature, and with all the passions surrounding it. We'd be assaulted with
> replies of the type "read this, this, this, and that" -- to which we can
> also reply with "read this, this, this, and that" as proof to the contrary.
>

... while perhaps we should focus on something that *really* matters.
for example, how well this or that model (*model*) of reality suits our
particular needs, irrespectively of what armchair speculation tells us
about what there is and what there is not.

> At any rate, what matters most is that BFO curators *themselves* decided to
> include GDCs in BFO

someone has to do this -- and it's usually the boss who decides, and
that's not a bad principle, is it?
the question here is about how decisions are justified. but "'cause
this is how reality is" is just funny.

> --which, as far as I sense, they regard as a grey area
> (which, again, they shouldn't). Hence, we cannot do *without* GDCs. Why not
> work *only* with GDCs then? A possible answer is that we have to stick to
> trope orthodoxy due to different (and higher) metaphysical reasons. But
> while admitting GDCs is definitely a departure from trope orthodoxy, there
> are good indications that tropes do not even fulfill the aims that prompted
> their introduction in the first place (explain "similarity" among others).
>
> Hence why not GDCs all the way?
>

hereby i suggest MDC -- mixed-dependency continuants (or perhaps
mysteriously dependent continuants) -- can we work out the semantics?
i'm pretty sure they exist.

vQ

[Cristian Cocos]

> the length already is a bit of a mysterious world; what happens to the
> length of the pencil when it gets shorter?

You mean "what happens with 7 cm when the pencil gets shorter?" Well, 7 cm
stays 7 cm of course ( :-) ), only it won't be the length of the pencil
anymore. The (new) length of the pencil will be 6 cm, say. The pencil has
different lengths at different times, *not* different values for its one and
only, unique and specific, length trope. There is no such thing as the
length of the pencil that "stays with it" ("inheres") at all times only it
has different "values." The "length" is not an individuating factor for the
pencil. The pencil has to be regarded as independent of its length.

(Yeah, it would take me some time to rephrase that w/o appeal to syntagms
like "its length" and the like, which is quite sneaky as it may make some
(actually quite many, as I came to realize) take it too literally and
postulate some specific and unique length literally *attached* to the
pencil, some entity that *belongs* to it in a unique and specific way etc.)

C

[Alan Ruttenberg]

On Jul 10, 2008, at 9:00 PM, Cristian Cocos wrote:

> So why not dump SDCs? That would make things much simpler: This
> pencil I am
> holding has a length, a *particular* length (say 7 cm). This is an
> individual in OWL terminology. This length is the same as the
> length of the
> pen on the desk--same length particular/instance.

The problem starts here. The issue is that your statement is
factually incorrect. Manufactured objects, of macroscopic scales
vary from item to item. That pen on the table is *not* the same
length as the one in your hand. Moreover, the one in your hand isn't
the same length as when you put it down on the table. It is shorter,
most likely, because it is colder.

Not paying attention to stuff that follows from incorrect statement
<<NEXT

> This is something that GDCs accommodate very well. No need for
> further length tropes that, in turn, have "values" etc. No need to
> discern between the length of the pencil and
> the length of the pen by postulating ("recognizing") another entity
> (length
> tropes) that are "specific" and "unique" etc., and that "inhere"
> in ...
> wherever they may inhere.

NEXT

>
> Yes, but what sort of a horrible person does not find "the particular
> redness of this shirt" as compelling evidence in favor of the
> existence of
> tropes? I, for one, ain't. And so do many others (namely mainstream
> contemporary philosophy).

But you give no alternative. What do the mainstream philosophers
offer in return.
(parenthetically, while some may consider this a matter which has
currency in the philosophical world, I, personally, am not part of
that environs. I'm just trying to build stuff on premises that
aren't obviously wrong. So, skipping now over the portion of the
conversation involving nothing but the relationships of philosophers
to each <<OTHER

OTHER

Oops - that's it.
By for now.
-Alan

ps. << EOF is a perl joke.

>
> C
>
>
> >

[Alan Ruttenberg]

On Jul 10, 2008, at 11:36 PM, Cristian Cocos wrote:

>
>> the length already is a bit of a mysterious world; what happens
>> to the
>> length of the pencil when it gets shorter?
>
> You mean "what happens with 7 cm when the pencil gets shorter?"
> Well, 7 cm
> stays 7 cm of course ( :-) ), only it won't be the length of the
> pencil
> anymore.

You haven't answered what happens if here is no pencil that has that
length.
The current view, which attempts to ensure that every entity can be
traced to something in reality, would say that the instance ceases to
exist.

> The (new) length of the pencil will be 6 cm, say. The pencil has
> different lengths at different times, *not* different values for
> its one and
> only, unique and specific, length trope.

BFO doesn't state this, fwiw. Perhaps you won't like the alternative,
but here it is: There is a determinable quality: length of the pencil
(the pencil necessarily has some length for all it's life) and there
are the determinate quality instances, each of which are the specific
length the pencil has at each instance: a new one of these each time
the length changes. (This is Ingvar's story, I think I cited the
paper earlier). Not that this view doesn't have its own issues...

-Alan

[Alan Ruttenberg]

On Jul 10, 2008, at 6:47 PM, Phillip Lord wrote:

>>> Yes, I know. The question I raise, then, is whether we will ever
>>> use the
>>> qualities, since we only ever have values for measurements.
>
> Alan> Indeed we may never instantiate them.
>
> If they are not instantiable, then they have no instances in
> reality. If they
> have no instances in reality, then they should not be in BFO, which
> is my
> question.

I meant: It may be the case that we never instantiate them. Not: we
may not ever instantiate them.

> Alan> However it is important, for the coherence of the ontology
> that they
> Alan> present, and may be necessary for certain inferences to be
> correctly
> Alan> made, by virtue of the axioms that mention the classes.
>
> Such as...

I don't have an example, for now of this case. But consider the case
of protein, which is analogous. One never, or virtually never, refer
to a single instance of a protein. Does that mean the class is not
necessary? We certainly need those classes if we are to describe
protein binding, instance of which we understand to be cases where
instances of protein are attached to each other.

Maybe. Here's why I say maybe. I think first about possible
counterexamples: Are there other quantities that I can compute from a
set of measurements of qualities of individuals which don't plausibly
seem to be qualities of an aggregate. Here's one. Suppose I measure
the length of the longest thread that is pulled from any fabric of
the clothes each person wears. Is that average a quality of the
aggregate? I don't find that gives me the warm and fuzzy: Yup, that's
a quality of the aggregate.

Then the positive side: What would I consider a legitimate quality of
an aggregate? Immediately temperature comes to mind.

Now, is the average height of the people in the population the sort
of entity that is closer to a temperature, or to the average thread
length. And to that question, the answer isn't too clear. So I
continue to thing about it and I say maybe.

> Alan> I am confident that the information entities: a) The mean
> measured
> Alan> height of 100 people b) the mean of three measurements of
> one person's
> Alan> height. Both of these are the result of data
> transformations of
> Alan> multiple datums that are measurements of the same sort of
> thing.
>
> Again, these are estimates of an average. In case a) where we
> measure 100
> people's height to get a population average, what we really want is
> to measure
> the entire population. In case b) where we want to control against
> variability
> in height over time (for example), we want to measure over all time.
>
> The actual average you don't measure, you only estimate. This is
> why you have
> error bars.

It's why you can have error bars. Whether you note error bars depends
on how careful a scientist you are. BTW, I place error bars firmly in
the area of information artifact.

> Alan> A recent question

> Alan> asked is what the domain and range of the is_proxy_for
> relation for.
> Alan> On first glance it would seem to be entity, in both cases.
> Needs some
> Alan> more thought. Got any cases that might suggest otherwise?
>
> Can you restate the question? A case that suggests otherwise would
> have to be
> a counter example; given that all things are entities, it's not
> possible to
> give a counter example.

What I meant was, are there kinds of thing that don't seem to make
sense as subject or object of is_proxy_for, therefore disagreeing
with the hypothesis that the domain and range are entity.

-Alan

>

[Alan Ruttenberg]

On Jul 10, 2008, at 10:06 PM, lampucerka wrote:

At any rate, what matters most is that BFO curators *themselves* decided to

include GDCs in BFO

someone has to do this -- and it's usually the boss who decides, and

that's not a bad principle, is it?

the question here is about how decisions are justified.  but "'cause

this is how reality is" is just funny.

Since I was there from the beginning of the GDC story, I'm happy to relate as much about how this decision was made as you like. While sometimes arguments that involve BFO sound like they are truncated with "cause this is how reality is", they aren't decided like that among us (I guess I am one of them now) and I don't find explanations such as the above satisfactory in any way.

FWIW, I think such is said as deliberately provocative, in a (perhaps wrongheaded) attempt to get the other person thinking harder. Moreover, getting someone else thinking harder about these issues *does not* mean that the expected result of that thinking will only have been successful when the other party agrees. Instead, all of "us" consider these issues hard to figure out, and are on the lookout for others who can provide insight that might help make progress in sorting them out.

A bit of background on the GCD decision. The driver was the need to represent things such as experimental data, file formats, written protocols, study designs, and a number of other entities that OBI concerns itself with. What are these sorts of things? What are they for real? (not just: how can I make a data structure that can hold something about them - there is a premise that figuring out what things are for real will lead to representations that make it much easier to integrate data)

One aspect that drew attention was that of perfect copyability. Unlike the copying exhibited in the sequential prints of a lithograph, where each differs in many small ways, the copy of a file on my hard disk has some element of it that is exactly the same as copy on my backup drive. 

Following this observation two possibilities arose: 1) The things that are the same are instances of a class (or raising the stakes, a universal ;-) or  2) Something else.

Here's a consequence of #1. If I have  PDF file that I distribute, the set of those PDF files form a class.  So we have a class and the instances. Which of these captures the shared sameness? Has to be the class - there is no other choice. So, given that, one would have a situation where whenever one produced a new PDF a new class was born. Some, Barry particularly, and most of the OBIans considered this distasteful. That PDF file was an instance and those copies were ... well ... something. Oh: instance can't have instances.

OTOH, Pierre and perhaps Cristian (can't exactly remember) insisted that despite the distastefulness, the conclusion that the PDF was a class was the only reasonable one. Sameness that manifests at different locations = Universal. End of story.

Having these two options and having to resolve which to take is a no win situation. No one will be happy. Here the boss (Barry) stepped in and said we will try GDC. My view was (and continues to be) that it's worth a try. By try, I mean see if we can work consistently within this framework, what consequences arise, whether the "universal view" keeps calling, etc.

The argument about whether classes for PDFs or instance *is not over*. It continues and new ways of attempting to clearly distinguish GDCs from classes arise and are debated. But not in a way that blocks further work, which would be bad. A current argument for distinguishing them is based on the idea that PDFs are more "accidental" than universals are and they are more complex in their structure - one can't describe them in hierarchies with relatively simple differentia defining each branch. So this argues for them being more instance like. A counter argument to this view is that if one re-scales time things which seem deliberate now seem more accidental and unique. A nature segue has been an ongoing discussion of whether species are instances or classes.

Moreover, in my view, the exact sameness aspect of GDCs seems to be present in some other SDCs. One *can* argue that the diploid quality of one cell is exactly the same as the diploid quality of another cell (ignoring the spatiotemporal aspects). So should diploid be considered GDC? Some other sort of thing? Again, this discussion continues. 

Finally there is the issue of entities that seem to be qualities of GDCs. The compressibility of a file is one example. This would seem to be a "quality" of the file. However qualities are specifically dependent on independent continuants, and a file is dependent continuant. You can't have, according to the current definitions,  dependent continuants dependent on other dependent continuants. 

So the bottom line is that certainly decisions are not justified by "cause this is how reality is". Thinking about representation and ontology is hard. All of "us" feel unsatisfied because there are things that we don't feel we have good explanations of. But many of us are actively involved in projects that need solutions now, and for which the thinking behind BFO which does seem less problematic now can have serious benefits. So time is balanced between pushing forward on projects with the best we have, and on the other hand, with trying to sort out the problems that remain. If either stops, then this becomes an unsuccessful endeavor.

I hope this gives you a better feel for the dynamic of the development, and gives you some comfort that decisions don't happen capriciously, or without quite a bit of deliberation, or with overconfidence that we have *the* answer.

Regards,

Alan 

[Matthew Pocock]

> > The actual average you don't measure, you only estimate. This is
> > why you have
> > error bars.
>

> It's why you can have error bars. Whether you note error bars depends
> on how careful a scientist you are. BTW, I place error bars firmly in
> the area of information artifact.

Error bars are a statement about your confidence, or (if you are a bayesian
not a frequentist) your belief in the actual value of the random variable you
have measured. I have no opinion about if they are informatoin or generically
dependent - that is utterly asside from what you know because you have been
presented with error bars vs not being presented with them.

Matthew

[Cristian Cocos]

> The problem starts here. The issue is that your statement is
> factually incorrect. Manufactured objects, of macroscopic scales
> vary from item to item. That pen on the table is *not* the same
> length as the one in your hand. Moreover, the one in your hand isn't
> the same length as when you put it down on the table. It is shorter,
> most likely, because it is colder.

Forget about the pen on the desk. So you're saying that there's no other
object, fiat or bona fide, artifact or natural, anywhere in the universe,
that at this very time has the same length as the pencil that I am holding
now? I think that's a very bold statement, which needs further arguing for
(as far as I am concerned, a *lot* of arguing for).

(Note that exact "copyability" is precisely the reason behind GDCs. Exact
copyability *is* something that *can* be found in nature, and not just among
"informational objects.")

> > Yes, but what sort of a horrible person does not find "the particular
> > redness of this shirt" as compelling evidence in favor of the
> > existence of tropes? I, for one, ain't. And so do many others (namely
> > mainstream contemporary philosophy).
>
> But you give no alternative.

Alternative to what? To regarding the redness of this shirt as something
only *it* can have? Surely other shirts, or patches of material, can have
*exactly* the same color? If, however, by "particular redness" you mean
something that this shirt has by virtue of being *this shirt*, then you've
moved away from taking just color into consideration: what you're doing is
(illegitimately) moving the focus from color (wavelength?) itself, to the
pair <this shirt, color>. Now *that* is, of course, something that no other
thing out there has, agreed. But that's hardly talk about *just* color.

C

[Alan Ruttenberg]

On Jul 11, 2008, at 7:03 AM, Cristian Cocos wrote:

>> The problem starts here. The issue is that your statement is
>> factually incorrect. Manufactured objects, of macroscopic scales
>> vary from item to item. That pen on the table is *not* the same
>> length as the one in your hand. Moreover, the one in your hand isn't
>> the same length as when you put it down on the table. It is shorter,
>> most likely, because it is colder.
>
> Forget about the pen on the desk. So you're saying that there's no
> other
> object, fiat or bona fide, artifact or natural, anywhere in the
> universe,
> that at this very time has the same length as the pencil that I am
> holding
> now?

I don't know. Maybe. Maybe not. A *macroscopic* scales, I think it is
pretty unlikely, and I suspect that it wouldn't be too hard to argue.

But in another email I noted concerns about qualities where this
question is more legitimately raised: The diploid quality of a cell,
for example. An on the microscopic level there are other questions -
the quality that that represents the ratio of neutrons to protons in
an atom of argon, for example.

> I think that's a very bold statement, which needs further arguing
> for (as far as I am concerned, a *lot* of arguing for).
>
> (Note that exact "copyability" is precisely the reason behind GDCs.
> Exact
> copyability *is* something that *can* be found in nature, and not
> just among
> "informational objects.")

At our workshop, there was some argumentation around this, and we
decided to more narrowly define the scope of our ontology work to be
information artifacts. Some argued that the natural entities that
exhibited perfect copyability, such as the sequence of amino acids
that defines the primary structure of a protein, carried information.
Others disagreed that such entities should be considered
information. So the information artifact ontology is isn't an
ontology of all GDCs, just a subset containing some of those that we
know we will need.

>
>>> Yes, but what sort of a horrible person does not find "the
>>> particular
>>> redness of this shirt" as compelling evidence in favor of the
>>> existence of tropes? I, for one, ain't. And so do many others
>>> (namely
>>> mainstream contemporary philosophy).
>>
>> But you give no alternative.
>
> Alternative to what? To regarding the redness of this shirt as
> something
> only *it* can have?

Not to *regarding* it, but to the representational consequences of
the fact that, like in the case of the pen, it is exceedingly
unlikely that two shirts will have *exactly* the same color.

> Surely other shirts,
probably not, in fact.

> or patches of material,
natural or fiat? If fiat, I'm not so interested only because of games
one might try to play around making the fiat criteria exactly this.

> can have *exactly* the same color?

We should, in any case, avoid using color for too much longer,
because of the shaky ground "color" as a quality, rests on.

> If, however, by "particular redness" you mean
> something that this shirt has by virtue of being *this shirt*,

I don't. But, as I've said, such issues do arise with qualities like
diploid.

> then you've
> moved away from taking just color into consideration: what you're
> doing is
> (illegitimately) moving the focus from color (wavelength?)

Most objective would be something like spectrum of light under a
standard spectrum of illumination.

> itself, to the
> pair <this shirt, color>. Now *that* is, of course, something that
> no other
> thing out there has, agreed. But that's hardly talk about *just*
> color.

I agree.

-Alan

[Alan Ruttenberg]

On Jul 11, 2008, at 2:58 AM, Matthew Pocock wrote:

>
>>> The actual average you don't measure, you only estimate. This is
>>> why you have
>>> error bars.
>>
>> It's why you can have error bars. Whether you note error bars depends
>> on how careful a scientist you are. BTW, I place error bars firmly in
>> the area of information artifact.
>
> Error bars are a statement about your confidence, or (if you are a
> bayesian
> not a frequentist) your belief in the actual value of the random
> variable you
> have measured.

Only if you are the sort of person that uses that information as a
basis for belief. For many they are the rendering of the results of a
data transformation whose input is a set of measurement data. The way
you've said it, it wouldn't be an error bar if the author didn't
actually use that information in the way you state.

> I have no opinion about if they are informatoin or generically

> dependent - that is utterly aside from what you know because you

> have been presented with error bars vs not being presented with them.

And what you know because you have been presented with error bars is
utterly aside from what you learn from them. The latter involves many
other things, such as how you have been educated.

-Alan

>
> Matthew
>
> >

[lampucerka]

thanks for the detailed explanation.

now, if your argument could be reduced to something like 'we decide to
include X in ontology O because we found it most convenient to do so',
i'd be pretty happy to accept that as a valid justification. (if in
addition you could convincingly argue why the decision was the most
convenient, even better). you discuss models, and it matters more which
model works better for the purpose than which model fits best the
reality (which seems an unsolvable issue at tthe level of universals,
gdcs, and the like).

but it does not seem to be exactly the case; there is still somewhere
there sneaking in the argument that this is how things are, so we must
represent them this way. well, the problem is that you can't -- or tell
me how you can -- know how things are (when it comes to the existence of
universals, gdcs, tropes, and all those fancy would-be-there entities),
and at this level you really ground the discussions in what fits your
view of the world, rather than in how the world really is.

you will never know how it is with pdf files -- will you? -- whether
there is a class for each pdf file which is then instantiated on every
hard drive (etc) where the file resides, or if the situation is
different. and once you can't know it, does it really matter how it
is? would it not be enough to examine if this or that representation works?

it's fine that in your internal discussions you use arguments other than
'this is how reality is'. would be better if there were more public
statement about design decisions made in BFO acknowledging usage
arguments as opposed to 'cause this is how it is' bluff. the latter has
been severely overexploited on this list

vQ

[Alan Ruttenberg]

I'm sorry, I don't see your point. You asked for detail and it was
given. It is now a part of the public record. If there are specific
edits to specific BFO related web pages that you would like to
suggest, please let me know and we could discuss those. As far as
simply giving an argument: "cause this is how it is'", I've already
expressed my opinion about such argumentation style. While I can't
control what others do, feel free to flame me if I try it. I've a
couple of other comments in line.

Otherwise I am now, personally, more interested in working again on
the problems at hand than the meta.

Regards,
Alan

On Jul 11, 2008, at 10:33 AM, lampucerka wrote:

>
> thanks for the detailed explanation.
>
> now, if your argument could be reduced to something like 'we decide to
> include X in ontology O because we found it most convenient to do so',
> i'd be pretty happy to accept that as a valid justification. (if in
> addition you could convincingly argue why the decision was the most
> convenient, even better). you discuss models, and it matters more
> which
> model works better for the purpose than which model fits best the
> reality (which seems an unsolvable issue at tthe level of universals,
> gdcs, and the like).

I happen think that representation of reality will work better for
the purposes I (and many others) have, and so I see aiming to
working better for purpose as one of several ways of understanding
reality. But I own this as my personal view.

> but it does not seem to be exactly the case; there is still somewhere
> there sneaking in the argument that this is how things are,

It is not sneaking. It is what we strive for.

> so we must
> represent them this way. well, the problem is that you can't -- or
> tell
> me how you can -- know how things are (when it comes to the
> existence of
> universals, gdcs, tropes, and all those fancy would-be-there
> entities),
> and at this level you really ground the discussions in what fits your
> view of the world, rather than in how the world really is.
>
> you will never know how it is with pdf files -- will you? -- whether
> there is a class for each pdf file which is then instantiated on every
> hard drive (etc) where the file resides, or if the situation is
> different. and once you can't know it, does it really matter how it
> is?

You see this is the kind of conclusion that is damaging. It does
matter, particularly if you are of the opinion, that things will work
better if they represent reality more closely. Your reasoning opens
the doors to having no principles whatsoever. If you want to pursue
this avenue I have no problem with that. It's just not what *this*
project does.

> would it not be enough to examine if this or that representation
> works?

What does "works" mean? I will tell you what I strive towards: You
pick up any source of data or knowledge that is represented using
ontologies based on the principles of BFO, add it to what you have
already, make a query, and get the answers. Compare the current
situation where one has to do *much* more work than that to do
anything resembling such a situation.

[Stefan Schulz]

Hi all,

what I am missing here is to granularity and scale are not addressed
in this discussion, e.g. two pencils can be described as being of the
same length in one scale (or granularity level) and of different
length in another one.
Here again, there is a confusion about the meaning of the word
"value", as it has two different senses:

1. values of qualities that inhere in something
2. values of measurements

on *1*:
Values of (non-discrete) quality instances that inhere in particulars
(e.g. Matthew's height) :
- They have an infinite definition and correspond to poins on
a real scale, e.g. 184.03726430098672… mm
- due to external physical factors (temperature, pressure, movement)
they vary across time, i.e. Matthew's exact height does rarely have
EXACTLY the same values comparing two instances in time,
- they are granularity-independent
- they cannot be known exactly

on *2*
Values of measurement :
- They depend on a given measurement technique
- accordingly, they exhibit a certain accuracy of measurement and a scale
- their values are intervals, according to the measurement accuracy
(e.g. if you use a ruler with a cm scale, then the (integer) measurement
"184 cm" stands for the interval
]183.5cm ; 184.5cm] ),
just as the (integer) measurement "72 inches" represents the interval
]71.5´´ ; 72.5´´]
- According to the measurement technique we can only state that
the (real) value of something is located in such an interval
(with a certain probability)

* on dependence *

- Matthew's height depends on Matthews body (specific dependence)
- For each instance in time there is exactly one value of
Matthew's height (generic dependence)
- For Matthews lifetime there is exactly one height-time function
(specific dependence)
- Any statement of quantifying Matthews height (or comparing it to
someone else's) does not refer to the height value itself (we cannot know
exactly) but rather to a measurement value which represents an interval
that is supposed to contain the real value of Matthew's height.
This interval depends on granularity and scale. So is it possible that the
two people have the same height (measurement) on an inch scale (e.g. 72 ´´)
but different weights on a centimeter scale (184cm vs. 185cm).

Comments ?

Best regards

Stefan

--
Stefan SCHULZ (apl. Prof. Dr. med.)
Universitätsklinikum - Institut für Medizinische
Biometrie und Medizinische Informatik
Stefan-Meier-Strasse 26 D-79104 Freiburg
[home: Eschholzstr. 70 D-79115 Freiburg]
+49 (0)761 2036725, 2049089, FAX 2036711
http://purl.org/steschu
[stsc...@uni-freiburg.de], Skype: stschulz

[Cristian Cocos]

> You haven't answered what happens if here is no pencil that has that
> length.

Yeah, but then I'd have to reveal my true (and hideous) Platonist (or, at
least, non-immanent realist) nature--that is, if I haven't already :-)) (I
do expect to be crucified as a dangerous non-immanent realist, if either
Barry, Ingvar, or Pierre read some of my previous emails :-) )

But then, if admitting GDCs is not a definite step in the direction of
non-immanent realism, I don't know what it is.

> > The (new) length of the pencil will be 6 cm, say. The pencil has
> > different lengths at different times, *not* different values for
> > its one and only, unique and specific, length trope.
>
> BFO doesn't state this, fwiw. Perhaps you won't like the alternative,

You're right, I don't. Looks very farfetched and obscure. Prima facie looks
to be quite the balancing act, strikingly ad hoc. At any rate, I'd have to
read Ingvar's paper in detail before pronouncing myself with confidence. So
far, from your description below, I don't see the need for the "determinable
quality" thingie. Will give it a thorough reading though.

> but here it is: There is a determinable quality: length of the pencil

> (the pencil necessarily has some length for all its life) and there

> are the determinate quality instances, each of which are the specific
> length the pencil has at each instance: a new one of these each time
> the length changes. (This is Ingvar's story, I think I cited the
> paper earlier). Not that this view doesn't have its own issues...

C

[Cristian Cocos]

> -----Original Message-----
> From: Alan Ruttenberg [mailto:alanrut...@gmail.com]
> Sent: Friday, July 11, 2008 8:37
> >
> > Forget about the pen on the desk. So you're saying that there's no
> > other object, fiat or bona fide, artifact or natural, anywhere in the
> > universe, that at this very time has the same length as the pencil that
I am
> > holding now?
>

> I don't know. Maybe. Maybe not. At *macroscopic* scales, I think it is

> pretty unlikely, and I suspect that it wouldn't be too hard to argue.

I disagree. I think the burden of proof is on those who maintain the
*principled* impossibility of an exact match, at *any time* in the existence
of the Universe. All it takes is a *single* counterexample, and a trope
theory grounded in the "postulate" of impossibility of an exact match goes
down the drain. At any rate, I wouldn't be pinning my hopes on rescuing
trope theory via maintaining that no two (or more) macroscopic objects in
the Universe can *as a matter of principle* have the same mass (say) at the
same time.

I don't, however, think that that's the route trope theorists would choose
in defending the idea of abstract particulars/tropes. My guess is that their
emphasis would be on "this-ness," rather than "exact match-ness". And that,
unless backed by some serious mysticism, won't fly.

> > (Note that exact "copyability" is precisely the reason behind GDCs.
> > Exact copyability *is* something that *can* be found in nature, and not
> > just among "informational objects.")
>
> At our workshop, there was some argumentation around this, and we
> decided to more narrowly define the scope of our ontology work to be
> information artifacts. Some argued that the natural entities that
> exhibited perfect copyability, such as the sequence of amino acids
> that defines the primary structure of a protein, carried information.
> Others disagreed that such entities should be considered
> information. So the information artifact ontology is isn't an
> ontology of all GDCs, just a subset containing some of those that we
> know we will need.

Yes, but my drift concerned the exact copyability of objects *other than*
informational (be they artifactual or not), like simple everyday life
macroscopic objects. In such mundane cases, denying exact
replicability/copyability AS A MATTER OF *PRINCIPLE*, is a *very* strong
statement.

> > Alternative to what? To regarding the redness of this shirt as
> > something only *it* can have?
>
> Not to *regarding* it, but to the representational consequences of
> the fact that, like in the case of the pen, it is exceedingly
> unlikely that two shirts will have *exactly* the same color.

Unlikely maybe, but not impossible. And that's what matters. To rescue
tropes in *this* manner you need ironclad certainty. Mere unlikelyhood won't
do, however strong. You need to deny the possibility of exact matches *as a
matter of principle*.

> > Surely other shirts,
> probably not, in fact.

What about tearing the shirt in two? So the shirt had initially a particular
color (light spectrum under standard white light, or however you may want to
define it). Would you say that the two halves have now, all of a sudden,
different colors? Just because they are no longer contiguous?

C

[Alan Ruttenberg]

On Jul 11, 2008, at 8:00 PM, Cristian Cocos wrote:

>> -----Original Message-----
>> From: Alan Ruttenberg [mailto:alanrut...@gmail.com]
>> Sent: Friday, July 11, 2008 8:37
>>>
>>> Forget about the pen on the desk. So you're saying that there's no
>>> other object, fiat or bona fide, artifact or natural, anywhere in
>>> the
>>> universe, that at this very time has the same length as the
>>> pencil that
> I am
>>> holding now?
>>
>> I don't know. Maybe. Maybe not. At *macroscopic* scales, I think
>> it is
>> pretty unlikely, and I suspect that it wouldn't be too hard to argue.
>
> I disagree. I think the burden of proof is on those who maintain the
> *principled* impossibility of an exact match, at *any time* in the
> existence
> of the Universe. All it takes is a *single* counterexample, and a
> trope
> theory grounded in the "postulate" of impossibility of an exact
> match goes
> down the drain.

Sorry don't understand. I didn't postulate impossibility of an exact
match, just how unlikely it was. Suppose there was a match among the
many different. Any reason this should be considered something other
than an accident?

> At any rate, I wouldn't be pinning my hopes on rescuing
> trope theory via maintaining that no two (or more) macroscopic
> objects in
> the Universe can *as a matter of principle* have the same mass
> (say) at the
> same time.
>
> I don't, however, think that that's the route trope theorists would
> choose
> in defending the idea of abstract particulars/tropes. My guess is
> that their
> emphasis would be on "this-ness," rather than "exact match-ness".

Seems that there are both in the BFO sense of quality - the spatial
region of a quality, for example, is defined to occupy the same
spatial region as its bearer
http://www.ifomis.org/Research/IFOMISReports/IFOMIS%20Report%
2006_2003.pdf
(A 58)

But solely relying on location as a criterion isn't very persuasive -
I've agreed with you on that.

Moreover, exactly which qualities inhere in which entities is not
obvious to me. This suggests thinking carefully around this matter.
Does the shape of the ball inhere in the ball or the surface of it?

> And that, unless backed by some serious mysticism, won't fly.
>
>>> (Note that exact "copyability" is precisely the reason behind GDCs.
>>> Exact copyability *is* something that *can* be found in nature,
>>> and not
>>> just among "informational objects.")
>>
>> At our workshop, there was some argumentation around this, and we
>> decided to more narrowly define the scope of our ontology work to be
>> information artifacts. Some argued that the natural entities that
>> exhibited perfect copyability, such as the sequence of amino acids
>> that defines the primary structure of a protein, carried information.
>> Others disagreed that such entities should be considered
>> information. So the information artifact ontology is isn't an
>> ontology of all GDCs, just a subset containing some of those that we
>> know we will need.
>
> Yes, but my drift concerned the exact copyability of objects *other
> than*
> informational (be they artifactual or not), like simple everyday life
> macroscopic objects. In such mundane cases, denying exact
> replicability/copyability AS A MATTER OF *PRINCIPLE*, is a *very*
> strong
> statement.

Look, you are aiming at the edge case - I gave you a gift of an easy
case: diploid. Work your arguments with that. I'm curious what
insight you might bring examining this case.

>>> Alternative to what? To regarding the redness of this shirt as
>>> something only *it* can have?
>>
>> Not to *regarding* it, but to the representational consequences of
>> the fact that, like in the case of the pen, it is exceedingly
>> unlikely that two shirts will have *exactly* the same color.
>
> Unlikely maybe, but not impossible. And that's what matters.

Why? It is you who has extremized an observation of the extreme
unlikelihood of certain qualities of two macroscopic qualities being
the same into an absolute, not I.

> To rescue
> tropes in *this* manner you need ironclad certainty. Mere
> unlikelyhood won't
> do, however strong. You need to deny the possibility of exact
> matches *as a
> matter of principle*.

Why? So what if there happen to be a handful of GDCs in 10^80 SDCs?
What does this tell us?

>>> Surely other shirts,
>> probably not, in fact.
>
> What about tearing the shirt in two? So the shirt had initially a
> particular
> color (light spectrum under standard white light, or however you
> may want to
> define it). Would you say that the two halves have now, all of a
> sudden,
> different colors? Just because they are no longer contiguous?

It is the idea of the redness of this shirt that is flawed. A
convenient example for purposes of discourse, but not really sensible
when it comes down to close examination. Let's suppose we equate it
with the number of everything-but-red-absorbing dye molecules,
divided by the area of the shirt (a charitable definition) Variations
in the density (number of molecules per surface area) across the
shirt will again make it exceedingly unlikely that the same density
will be found for any two fragments.

Don't like this definition of redness of shirt? Offer me another one.

Again: I handed you "diploid". Crucify the idea with that.

-Alan

>
> C
>

[Alan Ruttenberg]

On Jul 11, 2008, at 11:30 AM, Stefan Schulz wrote:
> Hi all,
>
> what I am missing here is to granularity and scale are not addressed
> in this discussion, e.g. two pencils can be described as being of the
> same length in one scale (or granularity level) and of different
> length in another one.
> Here again, there is a confusion about the meaning of the word
> "value", as it has two different senses:
>
> 1. values of qualities that inhere in something
> 2. values of measurements
>
> on *1*:
> Values of (non-discrete) quality instances that inhere in particulars
> (e.g. Matthew's height) :

> - They have an infinite definition and correspond to points on

> a real scale, e.g. 184.03726430098672… mm

Having trouble parsing this point.

> - due to external physical factors (temperature, pressure,
> movement)
> they vary across time, i.e. Matthew's exact height does
> rarely have
> EXACTLY the same values comparing two instances in time,
> - they are granularity-independent
> - they cannot be known exactly

Yes.

>
> on *2*
> Values of measurement :
> - They depend on a given measurement technique
> - accordingly, they exhibit a certain accuracy of measurement
> and a scale
> - their values are intervals, according to the measurement
> accuracy
> (e.g. if you use a ruler with a cm scale, then the (integer)
> measurement
> "184 cm" stands for the interval
> ]183.5cm ; 184.5cm] ),
> just as the (integer) measurement "72 inches" represents the
> interval
> ]71.5´´ ; 72.5´´]
> - According to the measurement technique we can only state that
> the (real) value of something is located in such an interval
> (with a certain probability)

This is to my thinking other than the precision you assign to the
bounds of the interval. I'd suspect quite a bit of variation,
depending on the scale, perception of the measurer, etc.

>
> * on dependence *
>
> - Matthew's height depends on Matthews body (specific dependence)
> - For each instance in time there is exactly one value of
> Matthew's height (generic dependence)

Don't know this sense of generic dependence. But according to your
*2* above, special relativity brings some question about this - are
multiple, simultaneous measurements of height in different frames of
reference a counterexample? Probably. For your *1* the question is
harder, but my initial reaction is that this holds - only one
instance, riding along in the same frame of reference as the bearer.

> - For Matthews lifetime there is exactly one height-time function
> (specific dependence)

Why is this not a simple consequence of the previous. Seems more like
a statement about time.

> - Any statement of quantifying Matthews height (or comparing it to
> someone else's) does not refer to the height value itself (we
> cannot know
> exactly) but rather to a measurement value which represents an
> interval
> that is supposed to contain the real value of Matthew's height.

Three things here: 1) I don't know what the sense of "refer" is in
the above sentence. Similarly, the sense of "represents".
2) "Any statement" would seem to refer to quite a large range of
possibilities. So large, that unless the proposition is true by
definition, it raises suspicion.
3) When you say comparison, do you mean comparison of the heights
(sense *1*) or the measurements of the heights (sense *2*). It seems
that these are two different cases. The comparison of *1* heights can
be made without the necessity of comparing the
*2* measurements of the heights.

> This interval depends on granularity and scale.

new sense of "depends". Probably better to use a different word.

> So is it possible that the two people have the same height
> (measurement) on an inch scale (e.g. 72 ´´) but different weights

height measurements

> on a centimeter scale (184cm vs. 185cm).

I'm not finding the sameness/differentness of rounded measurements
terribly compelling.
>
> Comments ?

Of course ;-)

Best,
Alan

[Stefan Schulz]

>> Values of (non-discrete) quality instances that inhere in particulars
>> (e.g. Matthew's height) :
>> - They have an infinite definition and correspond to points on
>> a real scale, e.g. 184.03726430098672… mm
>
> Having trouble parsing this point.

Probably I am not mastering terms of mathematics in English very well...
I meant that at a given moment the height (length, diameter) of an
object equals the distance between two mathematical points
(Real number scale in Euclidean space). This distance is not
expressible in integer numbers, in contradistinction of what s.o. had
expressend in this discussion

>
>> - due to external physical factors (temperature, pressure, movement)
>> they vary across time, i.e. Matthew's exact height does rarely have
>> EXACTLY the same values comparing two instances in time,
>> - they are granularity-independent
>> - they cannot be known exactly
>
> Yes.
>
>>
>> on *2*
>> Values of measurement :
>> - They depend on a given measurement technique
>> - accordingly, they exhibit a certain accuracy of measurement and a
>> scale
>> - their values are intervals, according to the measurement accuracy
>> (e.g. if you use a ruler with a cm scale, then the (integer)
>> measurement
>> "184 cm" stands for the interval
>> ]183.5cm ; 184.5cm] ),
>> just as the (integer) measurement "72 inches" represents the
>> interval ]71.5´´ ; 72.5´´]
>> - According to the measurement technique we can only state that
>> the (real) value of something is located in such an interval
>> (with a certain probability)
>
> This is to my thinking other than the precision you assign to the bounds of
> the interval. I'd suspect quite a bit of variation, depending on the scale,
> perception of the measurer, etc.

There are two aspects.
First the granularity of your scale, independent of errors of
measurement. The use of a (discrete) centimeter scale for measuring
(non-discrete) length values requires conventions of how to round off
finer-grained values.
Second the measurement technique that always exhibits some measurement
error which has a probability distribution about the true value of the
measured quantity.

>> * on dependence *
>>
>> - Matthew's height depends on Matthews body (specific dependence)
>> - For each instance in time there is exactly one value of
>> Matthew's height (generic dependence)
>
> Don't know this sense of generic dependence. But according to your *2*
> above, special relativity brings some question about this - are multiple,
> simultaneous measurements of height in different frames of reference a
> counterexample? Probably.

So replace "exactly one" by "some". No problem. The important thing is
that height without value does not exist.

> For your *1* the question is harder, but my
> initial reaction is that this holds - only one instance, riding along in the
> same frame of reference as the bearer.

>> - For Matthews lifetime there is exactly one height-time function
>> (specific dependence)
>
> Why is this not a simple consequence of the previous. Seems more like a
> statement about time.
>

Yes it is a simple consequence (but only ignoring special relativity).
I brought this because under this view we get rid of generic dependence.

>> - Any statement of quantifying Matthews height (or comparing it to
>> someone else's) does not refer to the height value itself (we cannot know
>> exactly) but rather to a measurement value which represents an interval
>> that is supposed to contain the real value of Matthew's height.
>
> Three things here: 1) I don't know what the sense of "refer" is in the above
> sentence. Similarly, the sense of "represents".

- "Refer": better "denotes".
- "represents" the relation between a measurement value
(that can always be expressed as an integer value) and
a line segment on a continuous scale.

> 2) "Any statement" would seem to refer to quite a large range of
> possibilities. So large, that unless the proposition is true by definition,
> it raises suspicion.

... under the assumption that the proposition is true, OK.

> 3) When you say comparison, do you mean comparison of the heights (sense
> *1*) or the measurements of the heights (sense *2*). It seems that these are
> two different cases. The comparison of *1* heights can be made without the
> necessity of comparing the *2* measurements of the heights.

You can compare magnitudes without mapping them to measurements, but
you still depend on some measurement technique. For instance, you put
two objects on a laboratory balance and conclude that they have the
same mass. This assertion of sameness is, of course dependent on the
metering precision, probably if you use a higher precision lab balance
you will find a difference.

>> This interval depends on granularity and scale.
>
> new sense of "depends". Probably better to use a different word.

Really. What about "is defined by"

>>
>> So is it possible that the two people have the same height (measurement)
>> on an inch scale (e.g. 72 ´´) but different weights
>
> height measurements

of course

>>
>> on a centimeter scale (184cm vs. 185cm).
>
> I'm not finding the sameness/differentness of rounded measurements terribly
> compelling.

But doesn't you (or the instrument you use) always round off if measuring?

[Alan Ruttenberg]

On Jul 13, 2008, at 8:04 AM, Stefan Schulz wrote:

>>> Values of (non-discrete) quality instances that inhere in
>>> particulars
>>> (e.g. Matthew's height) :
>>> - They have an infinite definition and correspond to points on
>>> a real scale, e.g. 184.03726430098672… mm
>>
>> Having trouble parsing this point.
>
> Probably I am not mastering terms of mathematics in English very
> well...
> I meant that at a given moment the height (length, diameter) of an
> object equals the distance between two mathematical points
> (Real number scale in Euclidean space). This distance is not
> expressible in integer numbers, in contradistinction of what s.o. had
> expressend in this discussion

OK. I think this is justification for them not being able to be known
exactly.

OK. That there is some interval within which you know the answer I
agree with. That it is the specific interval you mention +/- .5 of the
scale, might be a reasonable convention. It wasn't clear from the
presentation which statements were about the nature of values, versus
being example or convention.

> Second the measurement technique that always exhibits some measurement
> error which has a probability distribution about the true value of the
> measured quantity.

This is something about the instrument/method, not so much the value.

>>> * on dependence *
>>>
>>> - Matthew's height depends on Matthews body (specific dependence)
>>> - For each instance in time there is exactly one value of
>>> Matthew's height (generic dependence)
>>
>> Don't know this sense of generic dependence. But according to your
>> *2*
>> above, special relativity brings some question about this - are
>> multiple,
>> simultaneous measurements of height in different frames of
>> reference a
>> counterexample? Probably.
>
> So replace "exactly one" by "some". No problem. The important thing is
> that height without value does not exist.

Ah - missed this point. Well taken. So you mean by generic dependence
that *1* is generically dependent because it is existentially
dependent on there being at least on *2* value always, but not the
same one?

>> For your *1* the question is harder, but my
>> initial reaction is that this holds - only one instance, riding
>> along in the
>> same frame of reference as the bearer.
>
>
>>> - For Matthews lifetime there is exactly one height-time function
>>> (specific dependence)
>>
>> Why is this not a simple consequence of the previous. Seems more
>> like a
>> statement about time.
>>
> Yes it is a simple consequence (but only ignoring special relativity).
> I brought this because under this view we get rid of generic
> dependence.

You mean replacing the many *2* values with a single function instead?
What status does this function have? Seems like it's just a renaming
of the sort that moves an issue out of view but doesn't really change
anything.

>>> - Any statement of quantifying Matthews height (or comparing it to
>>> someone else's) does not refer to the height value itself (we
>>> cannot know
>>> exactly) but rather to a measurement value which represents an
>>> interval
>>> that is supposed to contain the real value of Matthew's height.
>>
>> Three things here: 1) I don't know what the sense of "refer" is in
>> the above
>> sentence. Similarly, the sense of "represents".
>
> - "Refer": better "denotes".
> - "represents" the relation between a measurement value
> (that can always be expressed as an integer value) and
> a line segment on a continuous scale.

I understand "denotes" to be a relation between a name and a thing,
and encompasses intention. Therefore, the statement of quantifying
Matthew's height, when it uses the quantity, can be reasonably seen,
in my mind, to be a name for *2*. But maybe this is a wrong view...

I think I use "represents" in the sense of standing in for the thing
in a more detailed way then via denotation. True things one says using
the representation, when appropriately manipulated according to the
logical system that is used, are also true of the thing that is
represented.

The thing is that in your "represents" you connect two things in the
representation world. I don't see the line segment on a continuous
scale to be part of the same piece of reality that Matthew's *2*
height is.

(note, these comments aren't really a criticism of something - more of
an attempt to understand what you are getting at)

> You can compare magnitudes without mapping them to measurements, but
> you still depend on some measurement technique. For instance, you put
> two objects on a laboratory balance and conclude that they have the
> same mass. This assertion of sameness is, of course dependent on the
> metering precision, probably if you use a higher precision lab balance
> you will find a difference.

The balance is a good example. If you use one of these (image of
fulcrum balance):

[Cristian Cocos]

> > I disagree. I think the burden of proof is on those who maintain the
> > *principled* impossibility of an exact match, at *any time* in the
> > existence of the Universe. All it takes is a *single* counterexample,
> > and a trope theory grounded in the "postulate" of impossibility of
> > an exact match goes down the drain.
>
> Sorry don't understand. I didn't postulate impossibility of an exact
> match,

*You* didn't. Trope theory, on the other hand, doesn't admit (numerically
distinct) identical tropes. It's in the definition of a trope. It's an
analytic truth, if you will. If colors are the same, even for only two
(numerically distinct) objects from the multitude of objects in the
Universe, that would make those two tropes identical, which would be
impossible.

Again, individuation of tropes is a BIG question mark for trope theory, and
that's putting it mildly.

> just how unlikely it was. Suppose there was a match among the
> many different. Any reason this should be considered something other
> than an accident?

Yes, see above.

> > Yes, but my drift concerned the exact copyability of objects *other
> > than* informational (be they artifactual or not), like simple everyday
> > life macroscopic objects. In such mundane cases, denying exact
> > replicability/copyability AS A MATTER OF *PRINCIPLE*, is a *very*
> > strong statement.
>
> Look, you are aiming at the edge case

Not I, trope theory does that.

> - I gave you a gift of an easy
> case: diploid. Work your arguments with that. I'm curious what
> insight you might bring examining this case.

Ploid would be the parent quality, diploid a child quality to ploid. Cell A
has a diploid trope, cell B another diploid trope. The tropes are unique:
the two tropes are not to be confused. What makes them distinct *other than*
the fact that they belong to different cells? Damned if I know.

Hence "diploid" would certainly not be a SDC, but a GDC (hence not a
quality, but an Informational Object ... or whatever). Trouble is, the same
argument can be run for positively all qualities commonly thought to reside
on the SDC branch. Conclusion: do away with SDC, and move everything to the
GDC branch.

... On the other hand, GDCs means no longer immanent realism. It's something
else: Platonism (horribile dictu).

> >> Not to *regarding* it, but to the representational consequences of
> >> the fact that, like in the case of the pen, it is exceedingly
> >> unlikely that two shirts will have *exactly* the same color.
> >
> > Unlikely maybe, but not impossible. And that's what matters.
>
> Why? It is you who has extremized an observation of the extreme
> unlikelihood of certain qualities of two macroscopic qualities being
> the same into an absolute, not I.

Again, not me, trope theory does that.

> > To rescue
> > tropes in *this* manner you need ironclad certainty. Mere
> > unlikelyhood won't
> > do, however strong. You need to deny the possibility of exact
> > matches *as a
> > matter of principle*.
>
> Why? So what if there happen to be a handful of GDCs in 10^80 SDCs?
> What does this tell us?

See above.

> > What about tearing the shirt in two? So the shirt had initially a
> > particular color (light spectrum under standard white light,
> > or however you may want to define it). Would you say that the
> > two halves have now, all of a sudden, different colors? Just
> > because they are no longer contiguous?
>
> It is the idea of the redness of this shirt that is flawed.

Color (color of a tomato, color of a chair etc.) is absolutely, positively,
the favorite/paradigm example given by trope theorists when demonstrating
their theories. It's also a prominent example of a quality in the current
BFO version.

C

[Alan Ruttenberg]

Long story short: "Am I my trope theorist's keeper?"

Longer: Here's my thinking: As I understand qualities in BFO, they are
individuated by sharing the same spatial region as their bearers. In
many cases their *2* values also differ in all or virtually all cases.
Thus their individuation seems justified. As I said, I would consider
cases of rare equality to be accidental.

OTOH, the case of diploid and similar the only justification for
individuation is the (posited) spatial region sharing. In these case I
think there is possibly some further thinking to be done. I don't
think of these cases as the same phenomena as information artifacts -
a single entity that can have many bearers or different bearers at
different times. But, as I said, I share a similar discomfort to your
sentiment: What makes them distinct *other than* the fact that they

belong to different cells? Damned if I know.

If it will bring you comfort, why don't we agree that the trope
theorists are losers. How about we get on to figuring out what a
proper theory is?

-Alan
(too practical to be a philosopher)

[Phillip Lord]

>>>>> "Alan" == Alan Ruttenberg <alanrut...@gmail.com> writes:

Alan> Indeed we may never instantiate them.
>>
>> If they are not instantiable, then they have no instances in
>> reality. If they have no instances in reality, then they should not
>> be in BFO, which is my question.

Alan> I meant: It may be the case that we never instantiate them. Not:
Alan> we may not ever instantiate them.

You're thinking OO style -- BFO wise, it's reality which instantiates
things. More sensible, if we use an open world assumption, lengths exist
whether we know their value or not.

Alan> However it is important, for the coherence of the ontology
>> that they
Alan> present, and may be necessary for certain inferences to be
>> correctly
Alan> made, by virtue of the axioms that mention the classes.
>>
>> Such as...

Alan> I don't have an example, for now of this case. But consider the
Alan> case of protein, which is analogous. One never, or virtually
Alan> never, refer to a single instance of a protein. Does that mean
Alan> the class is not necessary? We certainly need those classes if
Alan> we are to describe protein binding, instance of which we
Alan> understand to be cases where instances of protein are attached
Alan> to each other.

If I have not lost track of this thread, I don't think that this is
analogous. To make this equivalent, you would have to define a binding
property and a property which describes the measurement of binding. Then
you would say "well, we may never know whether two proteins bind in
reality, but every time we measure it, it comes out at this value".

Alan> I am not sure that I think something as the average height
>> of a
Alan> certain population is something real of the same sort that
>> the height
Alan> of an individual is.
>>
>> It's a quality of an agregate I think. The same questions apply to
>> averages of height of a person I think.

Alan> Maybe. Here's why I say maybe. I think first about possible
Alan> counterexamples: Are there other quantities that I can compute
Alan> from a set of measurements of qualities of individuals which
Alan> don't plausibly seem to be qualities of an aggregate. Here's
Alan> one. Suppose I measure the length of the longest thread that is
Alan> pulled from any fabric of the clothes each person wears. Is that
Alan> average a quality of the aggregate? I don't find that gives me
Alan> the warm and fuzzy: Yup, that's a quality of the aggregate.

The length of the longest is different from the average. The length is
not an aggregate quantity, it's a quantity associated with a single
individual thread. The determination of which thread is longest is,
however, a property of the aggregate.

Alan> Then the positive side: What would I consider a legitimate
Alan> quality of an aggregate? Immediately temperature comes to mind.

Alan> Now, is the average height of the people in the population the
Alan> sort of entity that is closer to a temperature, or to the
Alan> average thread length. And to that question, the answer isn't
Alan> too clear. So I continue to thing about it and I say maybe.

Well, I go with the statisticians on this. This is a population average.
If the height of any member of the population changes, the average
changes. I see no way in which this could not be considered to be a
quality of the population which would be an aggregate.

>> Again, these are estimates of an average. In case a) where we
>> measure 100 people's height to get a population average, what we
>> really want is to measure the entire population. In case b) where
>> we want to control against variability in height over time (for
>> example), we want to measure over all time.
>>
>> The actual average you don't measure, you only estimate. This is
>> why you have error bars.

Alan> It's why you can have error bars. Whether you note error bars
Alan> depends on how careful a scientist you are. BTW, I place error
Alan> bars firmly in the area of information artifact.

Error bars are a representation of error. An error is an intrinsic part
of every measurement. I think I would argue that error exists in every
measurement, whether you have know that error or not.

Again, I have to question the use of qualitities like height, if we
can't attach measurements of their value to them. And to attach values,
we have to be able to talk about errors.

My worry, here, that I have expressed before, is that we are going to
find a large part of science stuck under GDC, or worse, duplicated under
GDC.

Alan> A recent question asked is what the domain and range of the
Alan> is_proxy_for

>> relation for.
Alan> On first glance it would seem to be entity, in both cases.
>> Needs some
Alan> more thought. Got any cases that might suggest otherwise?
>>
>> Can you restate the question? A case that suggests otherwise would
>> have to be a counter example; given that all things are entities,
>> it's not possible to give a counter example.

Alan> What I meant was, are there kinds of thing that don't seem to
Alan> make sense as subject or object of is_proxy_for, therefore
Alan> disagreeing with the hypothesis that the domain and range are
Alan> entity.

Ah, okay. Well, the domain and range wouldn't be incorrect then, just
not as specific as they could be. I'll have a think about this. In
practice, I'm not 100% sure what you are trying to achieve with
"is_proxy_for", so I may not have much luck.

Phil

[Phillip Lord]

>>>>> "Alan" == Alan Ruttenberg <alanrut...@gmail.com> writes:

Alan> A bit of background on the GCD decision. The driver was the need
Alan> to represent things such as experimental data, file formats,
Alan> written protocols, study designs, and a number of other entities
Alan> that OBI concerns itself with. What are these sorts of things?
Alan> What are they for real? (not just: how can I make a data
Alan> structure that can hold something about them - there is a
Alan> premise that figuring out what things are for real will lead to
Alan> representations that make it much easier to integrate data)

Well, aside from being very unclear abot what "for real" means, I think
that there is an absence of evidence for this. The hope would be, I
guess, that if two people model the same set of things, they are more
likely to come up with a common representation if those things are
traceable back to reality; but, then, this philosophers seem to have
come up with a very large number of ways of modelling reality. Perhaps
not as many data structures as programmers have, but then there are more
programmers around.

Alan> Here's a consequence of #1. If I have PDF file that I
Alan> distribute, the set of those PDF files form a class. So we have
Alan> a class and the instances. Which of these captures the shared
Alan> sameness? Has to be the class - there is no other choice. So,
Alan> given that, one would have a situation where whenever one
Alan> produced a new PDF a new class was born. Some, Barry
Alan> particularly, and most of the OBIans considered this
Alan> distasteful. That PDF file was an instance and those copies were
Alan> ... well ... something. Oh: instance can't have instances.

Alan> OTOH, Pierre and perhaps Cristian (can't exactly remember)
Alan> insisted that despite the distastefulness, the conclusion that
Alan> the PDF was a class was the only reasonable one. Sameness that
Alan> manifests at different locations = Universal. End of story.

I think that this is where the "audit trail to reality" breaks down.

A PDF file, like all digitial entities is a basically a big number that
we choose to intepret in a given way. So the question is the same as
whether there is more than one instance of 3.

Personally, I think realism has caused problems because we ask pointless
questions like "does 3 really exist". I don't care because I know that
numbers are useful.

Ultimately, the distinction is minimal. Personally, I would have
preferred that BFO had not conflated entities like a PDF with the paper
is is printed on (or the disk it's written on), but ultimately I don't
think it makes that muh difference.

Alan> Moreover, in my view, the exact sameness aspect of GDCs seems to
Alan> be present in some other SDCs. One *can* argue that the diploid
Alan> quality of one cell is exactly the same as the diploid quality
Alan> of another cell (ignoring the spatiotemporal aspects). So should
Alan> diploid be considered GDC? Some other sort of thing? Again, this
Alan> discussion continues.

Alan> Finally there is the issue of entities that seem to be qualities
Alan> of GDCs. The compressibility of a file is one example. This
Alan> would seem to be a "quality" of the file.

Compressibility is an quality of the pattern of the file -- it's
essentially the inverse of the degree of randomness. It's also
independent of the interpretation; so a PDF is a number interpreted in a
certain way; the contents of a PDF depend on this interpretation. The
compressibility does not.

Alan> However qualities are specifically dependent on independent
Alan> continuants, and a file is dependent continuant. You can't have,
Alan> according to the current definitions, dependent continuants
Alan> dependent on other dependent continuants.

Oh dear.

Alan> So the bottom line is that certainly decisions are not justified
Alan> by "cause this is how reality is".

No. The opposite is often true, however, where good models are denied
because the "entities don't exist in reality".

I'd prefer a clear set of compentency questions and a justification of
how the end model fulfils them. Reality or otherwise, I don't
understand.

Phil

[Phillip Lord]

>>>>> "CC" == Cristian Cocos <cri...@ieee.org> writes:

>> The problem starts here. The issue is that your statement is
>> factually incorrect. Manufactured objects, of macroscopic scales
>> vary from item to item. That pen on the table is *not* the same
>> length as the one in your hand. Moreover, the one in your hand
>> isn't the same length as when you put it down on the table. It is
>> shorter, most likely, because it is colder.

CC> Forget about the pen on the desk. So you're saying that there's no
CC> other object, fiat or bona fide, artifact or natural, anywhere in
CC> the universe, that at this very time has the same length as the
CC> pencil that I am holding now? I think that's a very bold
CC> statement, which needs further arguing for (as far as I am
CC> concerned, a *lot* of arguing for).

Alan's done this before. He's trying to use "equal" in the mathematical
sense and apply it to physical models where it doesn't work or have any
good meaning. We (may) have measurements of your pencil which give an
average and error which may overlap with measurements of another pencil
and be equal within some specified degree of error.

At a deeper level, your pencil has a probability of being a certain
length which might overlap with the probabilities of other pencils.

Well, like I say "equal" is not that useful outside mathematics unless
it's more defined.

CC> (Note that exact "copyability" is precisely the reason behind
CC> GDCs. Exact copyability *is* something that *can* be found in
CC> nature, and not just among "informational objects.")

>> > Yes, but what sort of a horrible person does not find "the
>> > particular redness of this shirt" as compelling evidence in favor
>> > of the existence of tropes? I, for one, ain't. And so do many
>> > others (namely mainstream contemporary philosophy).
>>
>> But you give no alternative.

CC> Alternative to what? To regarding the redness of this shirt as
CC> something only *it* can have? Surely other shirts, or patches of
CC> material, can have *exactly* the same color? If, however, by
CC> "particular redness" you mean something that this shirt has by
CC> virtue of being *this shirt*, then you've moved away from taking
CC> just color into consideration: what you're doing is
CC> (illegitimately) moving the focus from color (wavelength?) itself,
CC> to the pair <this shirt, color>.

I'd tend to agree that it confuses the question about equality.

Phil

[Phillip Lord]

>>>>> "Alan" == Alan Ruttenberg <alanrut...@gmail.com> writes:

Alan> On Jul 11, 2008, at 7:03 AM, Cristian Cocos wrote:
>> (Note that exact "copyability" is precisely the reason behind GDCs.
>> Exact copyability *is* something that *can* be found in nature, and
>> not just among "informational objects.")

Alan> At our workshop, there was some argumentation around this, and
Alan> we decided to more narrowly define the scope of our ontology
Alan> work to be information artifacts. Some argued that the natural
Alan> entities that exhibited perfect copyability, such as the
Alan> sequence of amino acids that defines the primary structure of a
Alan> protein, carried information. Others disagreed that such
Alan> entities should be considered information.

DNA clearly carries information. If I send you a clone and you sequence
it, then I have communicated something to you, that you did not have
before.

Alan> I don't. But, as I've said, such issues do arise with qualities
Alan> like diploid.

>> then you've moved away from taking just color into consideration:
>> what you're doing is (illegitimately) moving the focus from color
>> (wavelength?)

Alan> Most objective would be something like spectrum of light under a
Alan> standard spectrum of illumination.

A (reflection) spectrum is a not the same thing as colour. Reflection
spectra admits to a standard definition, while colour depends on the
perception of the spectra. This is why colour models reflect
trichromatic human perception rather 14-chromatic shrimp preception.
PATO, for example, blesses HSV for this reason.

Phil

[Phillip Lord]

>>>>> "Alan" == Alan Ruttenberg <alanrut...@gmail.com> writes:

>> would it not be enough to examine if this or that representation
>> works?

Alan> What does "works" mean? I will tell you what I strive towards:
Alan> You pick up any source of data or knowledge that is represented
Alan> using ontologies based on the principles of BFO, add it to what
Alan> you have already, make a query, and get the answers. Compare the
Alan> current situation where one has to do *much* more work than that
Alan> to do anything resembling such a situation.

Alan

"I want to ask questions and get the answers" does not really advance
the world much beyond "I want it to work". You have to know what answers
and what questions.

I think that the point Waclaw is trying to make is that "representation
of reality" doesn't help with defining these questions. In practice, I
don't think it makes that much difference; Barry thinks that function
and little entites of reality, I think that function is a generalised
model that biologists use to help them understand too much data. We both
end up with function in our ontologies.

In a way, I feel like when I was young and joining the Scouts. In those
days, we promised to do our duty to God. Well, as God doesn't exist, I
was happy to do my duty to him, as these duties were likely to be
minimal. The same is true with "representations of entities in reality".
I doesn't seem to mean much, is highly malleable and all sorts of
strange things turn out to my entities in reality.

There are times, however, when it seems to lead to less optimal
modelling decisions. I hope that "entity in reality" doesn't become
doctrinaire, as I think this would damage the utility of ontologies.

Phil

[Phillip Lord]

>>>>> "SS" == Stefan Schulz <stsc...@uni-freiburg.de> writes:

SS> 1. values of qualities that inhere in something
SS> 2. values of measurements

SS> on *1*: Values of (non-discrete) quality instances that inhere in
SS> particulars (e.g. Matthew's height) :
SS> - They have an infinite definition and correspond to poins on
SS> a real scale, e.g. 184.03726430098672… mm

I think that's our mathematical abstraction. Normal space is not
infinitely divisible.

SS> - due to external physical factors (temperature, pressure,
SS> movement) they vary across time, i.e. Matthew's exact
SS> height does rarely have EXACTLY the same values comparing
SS> two instances in time,
SS> - they are granularity-independent
SS> - they cannot be known exactly

Quality instances of this sort probably are granularity dependent. For
Matt's height, this to be so, but if I ask for Matt's surface area then
it's depends how fine you measure it.

This becomes relevant in ecology---the effective surface of a tree, for
example, gets larger the smaller you are.

A similar argument would be true for height, although it would have less
impact than for surface area.

SS> on *2* Values of measurement :
SS> - They depend on a given measurement technique
SS> - accordingly, they exhibit a certain accuracy of measurement
SS> and a scale
SS> - their values are intervals, according to the measurement
SS> accuracy (e.g. if you use a ruler with a cm scale, then the
SS> (integer) measurement "184 cm" stands for the interval
SS> ]183.5cm ; 184.5cm] ),
SS> just as the (integer) measurement "72 inches" represents
SS> the interval
SS> ]71.5´´ ; 72.5´´]
SS> - According to the measurement technique we can only state
SS> that the (real) value of something is located in such an
SS> interval (with a certain probability)

I think that the latter is less definitional of 2. At a molecular scale
the "real" value of something is located in a interval with a
probabilitity which is potentially larger than the interval defined by
the standard error.

At a larger scale, this is not so much of an issue, since the
uncertainly in the real value is far outweighed by the uncertaintity in
it's measurement.

SS> * on dependence *

SS> - Matthew's height depends on Matthews body (specific dependence)
SS> - For each instance in time there is exactly one value of
SS> Matthew's height (generic dependence)

To a first approximation, although I think it depends on the
granularity.

SS> - For Matthews lifetime there is exactly one height-time function
SS> (specific dependence)
SS> - Any statement of quantifying Matthews height (or comparing it to
SS> someone else's) does not refer to the height value itself (we
SS> cannot know exactly) but rather to a measurement value which
SS> represents an interval that is supposed to contain the real
SS> value of Matthew's height.

Again, I don't think that you can split the two in such a simple manner.
If you accept my point about granularity (which you may not, of course),
then the height depends on the granularity at which you are measureing
it.

I also would question the use of this "real value" --- if we can never
measure it, then what's the point? Why not just talk about measurements.

Phil

[Alan Ruttenberg]

On Jul 15, 2008, at 4:57 AM, Phillip Lord wrote:
>
>>>>>> "Alan" == Alan Ruttenberg <alanrut...@gmail.com> writes:
>
> Alan> Indeed we may never instantiate them.
>>>
>>> If they are not instantiable, then they have no instances in
>>> reality. If they have no instances in reality, then they should not
>>> be in BFO, which is my question.
>
> Alan> I meant: It may be the case that we never instantiate them.
> Not: Alan> we may not ever instantiate them.
>
> You're thinking OO style -- BFO wise, it's reality which
> instantiates things.

I'm thinking of making statements about individuals proteins, which,
in OWL, involves creating individuals in the representation. But loose
words noted.

> More sensible, if we use an open world assumption, lengths exist
> whether we know their value or not.

> Alan> However it is important, for the coherence of the ontology
> that they
> Alan> present, and may be necessary for certain inferences to be
> correctly
> Alan> made, by virtue of the axioms that mention the classes.
>>>
>>> Such as...
>
> Alan> I don't have an example, for now of this case. But consider the
> Alan> case of protein, which is analogous. One never, or virtually
> Alan> never, refer to a single instance of a protein. Does that mean
> Alan> the class is not necessary? We certainly need those classes if
> Alan> we are to describe protein binding, instance of which we
> Alan> understand to be cases where instances of protein are attached
> Alan> to each other.
>
> If I have not lost track of this thread, I don't think that this is
> analogous. To make this equivalent, you would have to define a binding
> property and a property which describes the measurement of binding.
> Then you would say "well, we may never know whether two proteins
> bind in reality, but every time we measure it, it comes out at this
> value".

The comparison here was to motivate why an ontology need refer to
instances, even though one might never make a statement about a
particular protein.

But to pop a level, back to the question of measurement, while I see
the analogy you are making, binding is not the kind of quality
measurement of which might yield an huge number of distinct
possibilities. It is more like the case, considered as a measurement,
of diploid.

Moreover, the question of whether the thing measured was reality or
not hasn't been an issue in this discussion. No one contested that
there was such a thing as the height of a person. The entities in
question are the determinate versus determinable qualities, and
measurements of the determinate qualities. The question doesn't arise
(for me at least) as to whether binding exists, or whether the range
of possible "determinate binding values" is infinite (from quantum
mechanics we expect them to be discrete set).

> Alan> I am not sure that I think something as the average height of a
> Alan> certain population is something real of the same sort that
> the height
> Alan> of an individual is.
>>>
>>> It's a quality of an agregate I think. The same questions apply to
>>> averages of height of a person I think.
>
> Alan> Maybe. Here's why I say maybe. I think first about possible
> Alan> counterexamples: Are there other quantities that I can compute
> Alan> from a set of measurements of qualities of individuals which
> Alan> don't plausibly seem to be qualities of an aggregate. Here's
> Alan> one. Suppose I measure the length of the longest thread that is
> Alan> pulled from any fabric of the clothes each person wears. Is
> that
> Alan> average a quality of the aggregate? I don't find that gives me
> Alan> the warm and fuzzy: Yup, that's a quality of the aggregate.
>
> The length of the longest is different from the average. The length is
> not an aggregate quantity, it's a quantity associated with a single
> individual thread.

Yes.

> The determination of which thread is longest is, however, a property
> of the aggregate.

A determination is the result of a process, not a property of the
thing being studied. That determination might be a measure of some
determinate quality of the aggregate or it might not - that's the
question that isn't resolved in my thinking.

> Alan> Then the positive side: What would I consider a legitimate
> Alan> quality of an aggregate? Immediately temperature comes to mind.
>
> Alan> Now, is the average height of the people in the population the
> Alan> sort of entity that is closer to a temperature, or to the
> Alan> average thread length. And to that question, the answer isn't
> Alan> too clear. So I continue to thing about it and I say maybe.
>
> Well, I go with the statisticians on this. This is a population
> average.
> If the height of any member of the population changes, the average
> changes. I see no way in which this could not be considered to be a
> quality of the population which would be an aggregate.

Why can't it be an information artifact that is the product of a
process that statisticians do?

>>> Again, these are estimates of an average. In case a) where we
>>> measure 100 people's height to get a population average, what we
>>> really want is to measure the entire population. In case b) where
>>> we want to control against variability in height over time (for
>>> example), we want to measure over all time.
>>>
>>> The actual average you don't measure, you only estimate. This is
>>> why you have error bars.
> Alan> It's why you can have error bars. Whether you note error bars
> Alan> depends on how careful a scientist you are. BTW, I place error
> Alan> bars firmly in the area of information artifact.
>
> Error bars are a representation of error. An error is an intrinsic
> part
> of every measurement. I think I would argue that error exists in every
> measurement, whether you have know that error or not.

I would agree that error exists in many measurements (don't know about
all). However the error bar doesn't represent the error of any
particular measurement.

> Again, I have to question the use of qualitities like height, if we
> can't attach measurements of their value to them. And to attach
> values,
> we have to be able to talk about errors.

Who said that we can't measure them and talk about error (estimates)?

> My worry, here, that I have expressed before, is that we are going
> to find a large part of science stuck under GDC, or worse,
> duplicated under GDC.

Don't know. Either it does or doesn't. What to worry about? This
reminds me of the anxiety about too much of OBI landing up under
bfo:Role. But virtually everything about biomedical investigations
that isn't dealing with the "natural" which is the scope of other
ontologies is about social processes, and social processes involve
roles. So this is perfectly reasonable, if not one might first expect.

> Alan> A recent question asked is what the domain and range of the
> Alan> is_proxy_for relation for.
> Alan> On first glance it would seem to be entity, in both cases.
> Needs some
> Alan> more thought. Got any cases that might suggest otherwise?
>>>
>>> Can you restate the question? A case that suggests otherwise would
>>> have to be a counter example; given that all things are entities,
>>> it's not possible to give a counter example.
>
> Alan> What I meant was, are there kinds of thing that don't seem to
> Alan> make sense as subject or object of is_proxy_for, therefore
> Alan> disagreeing with the hypothesis that the domain and range are
> Alan> entity.
>
> Ah, okay. Well, the domain and range wouldn't be incorrect then, just
> not as specific as they could be. I'll have a think about this. In
> practice, I'm not 100% sure what you are trying to achieve with
> "is_proxy_for", so I may not have much luck.

is_proxy_for is to mark theory based connections between the entities
that we intend to measure and the things we actually measure directly.
A typical example is where we might measure time, but in order to do
so we count the number of oscillations of a crystal, and use a formula
to convert count to time. In each case where there is a proxy relation
there is a chance for a systematic error, if the theory is wrong. This
sort of relation, and problems in not explicitly noting it, is, in my
opinion riddled representation of scientific results. We use light
intensity to measure many things that are not light (e.g. presence of
quantity of a protein in a cell). In microarray experiments we measure
mrna using amount of hybridization to a probe as proxy then
radioactivity as proxy for that then fluorescence molecule as proxy
for that then light intensity as proxy for that.

-Alan

[Alan Ruttenberg]

On Jul 15, 2008, at 5:40 AM, Phillip Lord wrote:

>
>>>>>> "Alan" == Alan Ruttenberg <alanrut...@gmail.com> writes:
>
> Alan> A bit of background on the GCD decision. The driver was the
> need
> Alan> to represent things such as experimental data, file formats,
> Alan> written protocols, study designs, and a number of other
> entities
> Alan> that OBI concerns itself with. What are these sorts of things?
> Alan> What are they for real? (not just: how can I make a data
> Alan> structure that can hold something about them - there is a
> Alan> premise that figuring out what things are for real will lead to
> Alan> representations that make it much easier to integrate data)
>

> Well, aside from being very unclear about what "for real" means,

I don't believe you are unclear about this.

> I think that there is an absence of evidence for this.

That's a fair criticism. Time will tell.

> The hope would be, I guess, that if two people model the same set of
> things, they are more
> likely to come up with a common representation if those things are
> traceable back to reality; but, then, this philosophers seem to have
> come up with a very large number of ways of modelling reality. Perhaps
> not as many data structures as programmers have, but then there are
> more
> programmers around.

:)

> Alan> Here's a consequence of #1. If I have PDF file that I
> Alan> distribute, the set of those PDF files form a class. So we have
> Alan> a class and the instances. Which of these captures the shared
> Alan> sameness? Has to be the class - there is no other choice. So,
> Alan> given that, one would have a situation where whenever one
> Alan> produced a new PDF a new class was born. Some, Barry
> Alan> particularly, and most of the OBIans considered this
> Alan> distasteful. That PDF file was an instance and those copies
> were
> Alan> ... well ... something. Oh: instance can't have instances.
>
> Alan> OTOH, Pierre and perhaps Cristian (can't exactly remember)
> Alan> insisted that despite the distastefulness, the conclusion that
> Alan> the PDF was a class was the only reasonable one. Sameness that
> Alan> manifests at different locations = Universal. End of story.
>
> I think that this is where the "audit trail to reality" breaks down.

I don't see how. In both cases the trail leads to the bearer of the
PDF. There isn't any question of whether they are carried by bits of
charge, magnetic fields, etc.

> A PDF file, like all digitial entities is a basically a big number
> that
> we choose to intepret in a given way. So the question is the same as
> whether there is more than one instance of 3.

Given this view every data file (and every ebook) is basically a big
number. I don't find that view to be particularly illuminating or
useful.

> Personally, I think realism has caused problems because we ask
> pointless
> questions like "does 3 really exist". I don't care because I know that
> numbers are useful.

I haven't been devoting very much time to this question. Who do you
see has?

> Ultimately, the distinction is minimal. Personally, I would have
> preferred that BFO had not conflated entities like a PDF with the
> paper
> is is printed on (or the disk it's written on),

It precisely hasn't! There is the paper and disk. These are one sort
of thing. Then there is the information that they bear. That is a
different sort of thing. There is a connection. This is the opposite
of conflation. The question above is whether the thing that is the
information is a class or an instance.

> but ultimately I don't think it makes that much difference.

>
>
> Alan> Moreover, in my view, the exact sameness aspect of GDCs seems
> to
> Alan> be present in some other SDCs. One *can* argue that the diploid
> Alan> quality of one cell is exactly the same as the diploid quality
> Alan> of another cell (ignoring the spatiotemporal aspects). So
> should
> Alan> diploid be considered GDC? Some other sort of thing? Again,
> this
> Alan> discussion continues.
>
> Alan> Finally there is the issue of entities that seem to be
> qualities
> Alan> of GDCs. The compressibility of a file is one example. This
> Alan> would seem to be a "quality" of the file.
>
> Compressibility is an quality of the pattern of the file -- it's
> essentially the inverse of the degree of randomness.

Out of the BFO box, it isn't a quality, as quality is a dependent
continuant and those inhere in independent continuant. As information
is a dependent, it can't have a quality, by the current BFO definition.

> It's also independent of the interpretation; so a PDF is a number
> interpreted in a
> certain way; the contents of a PDF depend on this interpretation. The
> compressibility does not.

It's not so easy as that. Even in the case of lossless compression,
there is the issue of the program that does the compression being part
of the size of the compressed file (Kolmogorov complexity). So the
measure of compressibility, in the commonly used way, depends on more
than just the "number". (Otherwise what would explain the fact that
different compression programs yield different compressibility's).
Then consider the fact that PDF file is a program with instructions on
where to put marks on a page. The same instructions issued 100 times
give you the same result. So a PDF, considered a conveyer of
information about where to put marks on page (as intended), can
sometimes be substantially more compressed, losslessly, when one knows
how to interpret it.

> Alan> However qualities are specifically dependent on independent
> Alan> continuants, and a file is dependent continuant. You can't
> have,
> Alan> according to the current definitions, dependent continuants
> Alan> dependent on other dependent continuants.
>
> Oh dear.
>
> Alan> So the bottom line is that certainly decisions are not
> justified
> Alan> by "cause this is how reality is".
>
> No. The opposite is often true, however, where good models are denied
> because the "entities don't exist in reality".

Let's look at a specific case and review it, as I'm not sure what you
are referring to. More often than not what I see are either poor
models, or incomprehensible ones.

> I'd prefer a clear set of compentency questions and a justification of
> how the end model fulfils them. Reality or otherwise, I don't
> understand.

I've said before that I'm interested representation that is as
independent of purpose as possible, so as to make it more likely that
information can be used for unplanned (scientific) purposes.
Competency is achieved when independent groups can represent the same
phenomena and the representations can trivially combined and the
aggregate queried, with the query results being as accurate as if one
group had represented the whole lot. I think that's an appropriate
goal for sharing scientific results, to maximum effect, on the web.

If you've got an alternative formulation and different competency
questions that you think ought to be the goal of collaborative
ontology development then put them forward for consideration.

-Alan

>
>
> Phil
>
> >

[Alan Ruttenberg]

On Jul 15, 2008, at 5:55 AM, Phillip Lord wrote:

>
>>>>>> "Alan" == Alan Ruttenberg <alanrut...@gmail.com> writes:
>
> Alan> On Jul 11, 2008, at 7:03 AM, Cristian Cocos wrote:
>>> (Note that exact "copyability" is precisely the reason behind GDCs.
>>> Exact copyability *is* something that *can* be found in nature, and
>>> not just among "informational objects.")
>
> Alan> At our workshop, there was some argumentation around this, and
> Alan> we decided to more narrowly define the scope of our ontology
> Alan> work to be information artifacts. Some argued that the natural
> Alan> entities that exhibited perfect copyability, such as the
> Alan> sequence of amino acids that defines the primary structure of a
> Alan> protein, carried information. Others disagreed that such
> Alan> entities should be considered information.
>
> DNA clearly carries information. If I send you a clone and you
> sequence
> it, then I have communicated something to you, that you did not have
> before.

Shirts clearly carry information. If I send you a shirt and you sew
another one that looks like it then I have communicated something to
you that you did not have before.

And now I will ask: By your definition, what doesn't carry
information? And if you answer everything does, then I will ask what
makes DNA special, or, why is not "carries information" a content free
statement?

> Alan> I don't. But, as I've said, such issues do arise with qualities
> Alan> like diploid.
>
>>> then you've moved away from taking just color into consideration:
>>> what you're doing is (illegitimately) moving the focus from color
>>> (wavelength?)
>
> Alan> Most objective would be something like spectrum of light
> under a
> Alan> standard spectrum of illumination.
>
> A (reflection) spectrum is a not the same thing as colour. Reflection
> spectra admits to a standard definition, while colour depends on the
> perception of the spectra. This is why colour models reflect
> trichromatic human perception rather 14-chromatic shrimp preception.
> PATO, for example, blesses HSV for this reason.

Are we disagreeing about anything here?

>
>
> Phil
>
>
> >

[Alan Ruttenberg]

On Jul 15, 2008, at 7:10 AM, Phillip Lord wrote:

> Why not just talk about measurements.

If we only talked about measurements, how would we communicate.
Doesn't the practice of science checking that measurements that others
make can be repeated? How can you talk about a measurement being
repeated if you have nothing to anchor the measurement on, i.e. if you
can't talk about what the measurement was of?

-Alan

[Phillip Lord]

>>>>> "Alan" == Alan Ruttenberg <alanrut...@gmail.com> writes:

>> If I have not lost track of this thread, I don't think that this is
>> analogous. To make this equivalent, you would have to define a
>> binding property and a property which describes the measurement of
>> binding. Then you would say "well, we may never know whether two
>> proteins bind in reality, but every time we measure it, it comes
>> out at this value".

Alan> The comparison here was to motivate why an ontology need refer
Alan> to instances, even though one might never make a statement about
Alan> a particular protein.

Alan> But to pop a level, back to the question of measurement, while I
Alan> see the analogy you are making, binding is not the kind of
Alan> quality measurement of which might yield an huge number of
Alan> distinct possibilities. It is more like the case, considered as
Alan> a measurement, of diploid.

I think that this is very wrong. Protein-protein interaction data is
very noisy at best. Quantative measures of binding, also, is open to a
significant error.

>> The determination of which thread is longest is, however, a
>> property of the aggregate.

Alan> A determination is the result of a process, not a property of
Alan> the thing being studied.

My apologies, poor language. The determination of which thread is
longest requires knowledge of all the individuals in the aggregate.

Alan> Now, is the average height of the people in the population the
Alan> sort of entity that is closer to a temperature, or to the
Alan> average thread length. And to that question, the answer isn't
Alan> too clear. So I continue to thing about it and I say maybe.
>>
>> Well, I go with the statisticians on this. This is a population
>> average. If the height of any member of the population changes, the
>> average changes. I see no way in which this could not be considered
>> to be a quality of the population which would be an aggregate.

Alan> Why can't it be an information artifact that is the product of a
Alan> process that statisticians do?

I can ask that question about anything. Whether it is or not, I don't.
Regardless, it's refers to the population, the aggregate as a whole.

>> Error bars are a representation of error. An error is an intrinsic
>> part of every measurement. I think I would argue that error exists
>> in every measurement, whether you have know that error or not.

Alan> I would agree that error exists in many measurements (don't know
Alan> about all). However the error bar doesn't represent the error of
Alan> any particular measurement.

All measurements do have an error -- it's not really a measurement
otherwise. Error bars represent an estimate of the error for any
particular measure, yes.

>> Again, I have to question the use of qualitities like height, if we
>> can't attach measurements of their value to them. And to attach
>> values, we have to be able to talk about errors.

Alan> Who said that we can't measure them and talk about error
Alan> (estimates)?

You seem to be implying that measurements are information artifacts. You
said "we may never be able to instantiate them".

>> My worry, here, that I have expressed before, is that we are going
>> to find a large part of science stuck under GDC, or worse,
>> duplicated under GDC.

Alan> Don't know. Either it does or doesn't. What to worry about? This
Alan> reminds me of the anxiety about too much of OBI landing up under
Alan> bfo:Role. But virtually everything about biomedical
Alan> investigations that isn't dealing with the "natural" which is
Alan> the scope of other ontologies is about social processes, and
Alan> social processes involve roles. So this is perfectly reasonable,
Alan> if not one might first expect.

I asked earlier about having a clear understanding of what questions you
could or could not ask. If most of science comes under GDC, then there
is a problem; if the answer to my question "what kind of entity is this"
is always answered "it's a GDC", then I haven't actually done much.
Essentially, the terms in BFO cease to be discriminatory between the
things that I care. In this sense, the BFO concepts carry no
information.

>>
>> Ah, okay. Well, the domain and range wouldn't be incorrect then,
>> just not as specific as they could be. I'll have a think about
>> this. In practice, I'm not 100% sure what you are trying to achieve
>> with "is_proxy_for", so I may not have much luck.

Alan> is_proxy_for is to mark theory based connections between the
Alan> entities that we intend to measure and the things we actually
Alan> measure directly. A typical example is where we might measure
Alan> time, but in order to do so we count the number of oscillations
Alan> of a crystal, and use a formula to convert count to time.

Can you give me an example of where you would measure time directly
rather than through a proxy?

Phil

[Phillip Lord]

>>>>> "Alan" == Alan Ruttenberg <alanrut...@gmail.com> writes:
>>

>> Well, aside from being very unclear about what "for real" means,

Alan> I don't believe you are unclear about this.

Alan, why would I lie to you. I am unclear what it means. The only
answers I have recieved are: science is hard; I feel it in my bones; we
can't define all terms on pain of infinite regress.

Alan> OTOH, Pierre and perhaps Cristian (can't exactly remember)
Alan> insisted that despite the distastefulness, the conclusion that
Alan> the PDF was a class was the only reasonable one. Sameness that
Alan> manifests at different locations = Universal. End of story.
>>
>> I think that this is where the "audit trail to reality" breaks
>> down.

Alan> I don't see how. In both cases the trail leads to the bearer of
Alan> the PDF. There isn't any question of whether they are carried by
Alan> bits of charge, magnetic fields, etc.

We need an abstraction which is information.

>> A PDF file, like all digitial entities is a basically a big number
>> that we choose to intepret in a given way. So the question is the
>> same as whether there is more than one instance of 3.

Alan> Given this view every data file (and every ebook) is basically a
Alan> big number. I don't find that view to be particularly
Alan> illuminating or useful.

Interpreted in a certain way, yes.

>> Personally, I think realism has caused problems because we ask
>> pointless questions like "does 3 really exist". I don't care
>> because I know that numbers are useful.

Alan> I haven't been devoting very much time to this question. Who do
Alan> you see has?

The question has been raised here; it seems to be open to some
unclarity. I think Ingvar said that 0 doesn't exist because there are no
instances (by definition) and that only 1 is real. Others may differ.

>> Ultimately, the distinction is minimal. Personally, I would have
>> preferred that BFO had not conflated entities like a PDF with the
>> paper is is printed on (or the disk it's written on),

Alan> It precisely hasn't! There is the paper and disk. These are one
Alan> sort of thing. Then there is the information that they bear.
Alan> That is a different sort of thing. There is a connection. This
Alan> is the opposite of conflation. The question above is whether the
Alan> thing that is the information is a class or an instance.

There is a conflation. GDC just conflates the two and then says, well
it's a generic piece of paper. This is aside from the issue that
information can be born by things which are not continuants.

The point is that the least interesting, the least important thing about
a piece of information is that it's written onto a disk, piece of paper,
etc.

>> Compressibility is an quality of the pattern of the file -- it's
>> essentially the inverse of the degree of randomness.

Alan> Out of the BFO box, it isn't a quality, as quality is a
Alan> dependent continuant and those inhere in independent continuant.
Alan> As information is a dependent, it can't have a quality, by the
Alan> current BFO definition.

Not according to the defintions on the web. Qualities "inhere in
entities". DependentContinuant -- A continuant [snap:Continuant] that is
either dependent on one or other independent continuant
[snap:IndependentContinuant] bearers or inheres in or is borne by other
entities.

Other entities. Qualites can inhere in anything then.

>> It's also independent of the interpretation; so a PDF is a number
>> interpreted in a certain way; the contents of a PDF depend on this
>> interpretation. The compressibility does not.

Alan> (Otherwise what would explain the fact that different
Alan> compression programs yield different compressibility's).

There is a gain and a loss between compute time and compression. There
is a theoretical maximum compressibility.

Alan> Then consider the fact that PDF file is a program with
Alan> instructions on where to put marks on a page. The same
Alan> instructions issued 100 times give you the same result. So a
Alan> PDF, considered a conveyer of information about where to put
Alan> marks on page (as intended), can sometimes be substantially more
Alan> compressed, losslessly, when one knows how to interpret it.

I think that this statement is wrong.

Alan> If you've got an alternative formulation and different
Alan> competency questions that you think ought to be the goal of
Alan> collaborative ontology development then put them forward for
Alan> consideration.

It depends on the ontology that you are developing, is the short answer.

Phil

[Phillip Lord]

>>>>> "Alan" == Alan Ruttenberg <alanrut...@gmail.com> writes:

>> DNA clearly carries information. If I send you a clone and you
>> sequence it, then I have communicated something to you, that you
>> did not have before.

Alan> Shirts clearly carry information. If I send you a shirt and you
Alan> sew another one that looks like it then I have communicated
Alan> something to you that you did not have before.

Yes.

Alan> And now I will ask: By your definition, what doesn't carry
Alan> information?

A random stream of numbers. The patterns created by waves on a wave
guide. The same shirt that you sent me yesterday (unless you've written
something on it). An evenly distributed stream of monotones.

Alan> And if you answer everything does, then I will ask what makes
Alan> DNA special, or, why is not "carries information" a content free
Alan> statement?

If I send you a DNA sequence file with the sequence of pUC19, then the
file contains information. If a tube of pUC19, it doesn't. I find this
untenable.

Alan> Most objective would be something like spectrum of light
>> under a
Alan> standard spectrum of illumination.
>>
>> A (reflection) spectrum is a not the same thing as colour.
>> Reflection spectra admits to a standard definition, while colour
>> depends on the perception of the spectra. This is why colour models
>> reflect trichromatic human perception rather 14-chromatic shrimp
>> preception. PATO, for example, blesses HSV for this reason.

Alan> Are we disagreeing about anything here?

Colour is not "objective" in the sense that some arbitrary decisions
have to be made.

Phil

[Phillip Lord]

>>>>> "Alan" == Alan Ruttenberg <alanrut...@gmail.com> writes:

Alan> On Jul 15, 2008, at 7:10 AM, Phillip Lord wrote:

>> Why not just talk about measurements.

Alan> If we only talked about measurements, how would we communicate.
Alan> Doesn't the practice of science checking that measurements that
Alan> others make can be repeated? How can you talk about a
Alan> measurement being repeated if you have nothing to anchor the
Alan> measurement on, i.e. if you can't talk about what the
Alan> measurement was of?

Measurements of the same kind.

Phil

[lampucerka]

Phillip Lord wrote:
>
> If I send you a DNA sequence file with the sequence of pUC19, then the
> file contains information. If a tube of pUC19, it doesn't. I find this
> untenable.
>

what doesn't? doesn't the tube contain information, or doesn't the DNA
in the tube contain information?

(btw. the 'contains' here makes me a bit uncomfortable.)

vQ

[Alan Ruttenberg]

Not enough. Often confidence is built because one can do different
kinds of measurements of the same thing.

-Alan

[Alan Ruttenberg]

On Jul 17, 2008, at 7:20 AM, Phillip Lord wrote:

>
>>>>>> "Alan" == Alan Ruttenberg <alanrut...@gmail.com> writes:
>
>>> If I have not lost track of this thread, I don't think that this is
>>> analogous. To make this equivalent, you would have to define a
>>> binding property and a property which describes the measurement of
>>> binding. Then you would say "well, we may never know whether two
>>> proteins bind in reality, but every time we measure it, it comes
>>> out at this value".
>
> Alan> The comparison here was to motivate why an ontology need refer
> Alan> to instances, even though one might never make a statement
> about
> Alan> a particular protein.
>
> Alan> But to pop a level, back to the question of measurement,
> while I
> Alan> see the analogy you are making, binding is not the kind of
> Alan> quality measurement of which might yield an huge number of
> Alan> distinct possibilities. It is more like the case, considered as
> Alan> a measurement, of diploid.
>
> I think that this is very wrong. Protein-protein interaction data is
> very noisy at best.

That's a different issue, I think. The noise is usually of the sort
that measures .7 of a binding. It's the sort that yields false
positive and negatives as to whether there is a binding or not.

> Quantative measures of binding, also, is open to a significant error.

There are certainly measures associated with binding that are not
categorical. Like how long two proteins were bound, or the energy
needed to form the bond.

> Alan> Now, is the average height of the people in the population the
> Alan> sort of entity that is closer to a temperature, or to the
> Alan> average thread length. And to that question, the answer isn't
> Alan> too clear. So I continue to thing about it and I say maybe.
>>>
>>> Well, I go with the statisticians on this. This is a population
>>> average. If the height of any member of the population changes, the
>>> average changes. I see no way in which this could not be considered
>>> to be a quality of the population which would be an aggregate.
>
> Alan> Why can't it be an information artifact that is the product
> of a
> Alan> process that statisticians do?
>
> I can ask that question about anything. Whether it is or not, I don't.
> Regardless, it's refers to the population, the aggregate as a whole.

We agree on this. So do many other things - e.g the ratio of people
who've held a apple sometime today to the number who tripped over an
untied shoelace.

>>> Error bars are a representation of error. An error is an intrinsic
>>> part of every measurement. I think I would argue that error exists
>>> in every measurement, whether you have know that error or not.
>
> Alan> I would agree that error exists in many measurements (don't
> know
> Alan> about all). However the error bar doesn't represent the error
> of
> Alan> any particular measurement.
>
> All measurements do have an error -- it's not really a measurement
> otherwise.

I disagree. Whether there is an error or not is a matter of fact, and
depending on the quality, knowing the fact may be possible. If I ask
you the number of nucleated cells in a sample (a measurement) and you
give me an answer, if the answer is correct there is no error. So I
think that the measurements that always have errors are the ones that
attempt to measure quality values from a dense enough space to be
considered continuous.

>>> Again, I have to question the use of qualitities like height, if we
>>> can't attach measurements of their value to them. And to attach
>>> values, we have to be able to talk about errors.
>
> Alan> Who said that we can't measure them and talk about error
> Alan> (estimates)?
>
> You seem to be implying that measurements are information artifacts.
> You
> said "we may never be able to instantiate them".

Yes, but information artifacts are about things. In this case the
measurement is about the quality. The representation of information
artifacts includes this relation. So we talk about some qualities by
way of the information objects that are the results of attempts to
measure them.

>>> My worry, here, that I have expressed before, is that we are going
>>> to find a large part of science stuck under GDC, or worse,
>>> duplicated under GDC.
>
> Alan> Don't know. Either it does or doesn't. What to worry about?
> This
> Alan> reminds me of the anxiety about too much of OBI landing up
> under
> Alan> bfo:Role. But virtually everything about biomedical
> Alan> investigations that isn't dealing with the "natural" which is
> Alan> the scope of other ontologies is about social processes, and
> Alan> social processes involve roles. So this is perfectly
> reasonable,
> Alan> if not one might first expect.
>
> I asked earlier about having a clear understanding of what questions
> you
> could or could not ask. If most of science comes under GDC, then there
> is a problem; if the answer to my question "what kind of entity is
> this"
> is always answered "it's a GDC", then I haven't actually done much.

You have if you've said what the information artifact is about. In
order to do that you need the full kinds of ontologies we already
have. However, by talking about the GDCs and what they are about, when
appropriate, we avoid ascribing to entities properties which don't
make sense for them. Error estimates are a good example. The error
estimate associated with a measurement of a height isn't something
that is a quality of the person whose height it is. Nor is it a
quality of a population. It is something that relates to measurement
methods and our means of estimating errors (not trivial, and with
continuing controversy and new ideas).

> Essentially, the terms in BFO cease to be discriminatory between the
> things that I care. In this sense, the BFO concepts carry no
> information.

I hope you see now that this is a mistaken impression.

>>> Ah, okay. Well, the domain and range wouldn't be incorrect then,
>>> just not as specific as they could be. I'll have a think about
>>> this. In practice, I'm not 100% sure what you are trying to achieve
>>> with "is_proxy_for", so I may not have much luck.
>
> Alan> is_proxy_for is to mark theory based connections between the
> Alan> entities that we intend to measure and the things we actually
> Alan> measure directly. A typical example is where we might measure
> Alan> time, but in order to do so we count the number of oscillations
> Alan> of a crystal, and use a formula to convert count to time.
>
> Can you give me an example of where you would measure time directly
> rather than through a proxy?

No :-) !

But I can list a whole bunch of proxy chains involved with different
methods of measuring time and discuss the sort of systematic and
statistical errors associated with each.

Which is nice, IMO.

[Alan Ruttenberg]

On Jul 17, 2008, at 7:36 AM, Phillip Lord wrote:

>
>>>>>> "Alan" == Alan Ruttenberg <alanrut...@gmail.com> writes:
>>>
>>> Well, aside from being very unclear about what "for real" means,
>
> Alan> I don't believe you are unclear about this.
>
> Alan, why would I lie to you. I am unclear what it means. The only
> answers I have recieved are: science is hard; I feel it in my bones;
> we
> can't define all terms on pain of infinite regress.

There is 1) Whether there is stuff out there that is real and 2) Our
ability to describe it.

One can, and most scientists do, I would claim, understand and believe
(1) and work hard towards (2) - their life's work.

I think you and others often argue that you don't believe 1, possibly
for good pedagogic reason, but I really doubt that you don't believe
it. I could say I know how I walk with a lot of confidence. I would
be wrong. There are more reasons to question the truth of statements
than to fish out lying.

> Alan> OTOH, Pierre and perhaps Cristian (can't exactly remember)
> Alan> insisted that despite the distastefulness, the conclusion that
> Alan> the PDF was a class was the only reasonable one. Sameness that
> Alan> manifests at different locations = Universal. End of story.
>>>
>>> I think that this is where the "audit trail to reality" breaks
>>> down.
>
> Alan> I don't see how. In both cases the trail leads to the bearer of
> Alan> the PDF. There isn't any question of whether they are carried
> by
> Alan> bits of charge, magnetic fields, etc.
>
> We need an abstraction which is information.

Umm.. Information artifact is_a generically dependent continuant ?

>
>>> A PDF file, like all digitial entities is a basically a big number
>>> that we choose to intepret in a given way. So the question is the
>>> same as whether there is more than one instance of 3.
>
> Alan> Given this view every data file (and every ebook) is
> basically a
> Alan> big number. I don't find that view to be particularly
> Alan> illuminating or useful.
>
> Interpreted in a certain way, yes.

Don't understand this remark.

>
>
>>> Personally, I think realism has caused problems because we ask
>>> pointless questions like "does 3 really exist". I don't care
>>> because I know that numbers are useful.
>
> Alan> I haven't been devoting very much time to this question. Who do
> Alan> you see has?
>
> The question has been raised here; it seems to be open to some
> unclarity. I think Ingvar said that 0 doesn't exist because there
> are no
> instances (by definition) and that only 1 is real. Others may differ.

Perhaps. I haven't spend a lot of time on it, and Ingvar isn't here
very often. I don't see this as much affecting our work, though I do
know it's a good way to get a few people out there riled up ;-)

>>> Ultimately, the distinction is minimal. Personally, I would have
>>> preferred that BFO had not conflated entities like a PDF with the
>>> paper is is printed on (or the disk it's written on),
>
> Alan> It precisely hasn't! There is the paper and disk. These are one
> Alan> sort of thing. Then there is the information that they bear.
> Alan> That is a different sort of thing. There is a connection. This
> Alan> is the opposite of conflation. The question above is whether
> the
> Alan> thing that is the information is a class or an instance.
>
> There is a conflation. GDC just conflates the two and then says, well
> it's a generic piece of paper.

No it doesn't. Who said anything about a generic piece of paper.

> This is aside from the issue that information can be born by things
> which are not continuants.

That there processes that convey information is one of the issues to
figure out. I don't know if this is the same sense of borne as the
inverse of inheres - I suspect not. But I agree this are has to be
developed.

> The point is that the least interesting, the least important thing
> about a piece of information is that it's written onto a disk, piece
> of paper,
> etc.

Yes, so we mostly don't have to make statements about the bearer.
Doesn't mean that it's not there (and essential) and I've argued why I
think ontologies ought to account for the things that are there.

>>> Compressibility is an quality of the pattern of the file -- it's
>>> essentially the inverse of the degree of randomness.
>
> Alan> Out of the BFO box, it isn't a quality, as quality is a
> Alan> dependent continuant and those inhere in independent
> continuant.
> Alan> As information is a dependent, it can't have a quality, by the
> Alan> current BFO definition.
>
> Not according to the defintions on the web. Qualities "inhere in
> entities". DependentContinuant -- A continuant [snap:Continuant]
> that is
> either dependent on one or other independent continuant
> [snap:IndependentContinuant] bearers or inheres in or is borne by
> other
> entities.
>
> Other entities. Qualites can inhere in anything then.

I will inquire further. Perhaps my understanding of qualities is
defective.

>>> It's also independent of the interpretation; so a PDF is a number
>>> interpreted in a certain way; the contents of a PDF depend on this
>>> interpretation. The compressibility does not.
>
> Alan> (Otherwise what would explain the fact that different
> Alan> compression programs yield different compressibility's).
>
> There is a gain and a loss between compute time and compression. There
> is a theoretical maximum compressibility.
>
> Alan> Then consider the fact that PDF file is a program with
> Alan> instructions on where to put marks on a page. The same
> Alan> instructions issued 100 times give you the same result. So a
> Alan> PDF, considered a conveyer of information about where to put
> Alan> marks on page (as intended), can sometimes be substantially
> more
> Alan> compressed, losslessly, when one knows how to interpret it.
>
> I think that this statement is wrong.

That's not helpful. Care to say why?

>
> Alan> If you've got an alternative formulation and different
> Alan> competency questions that you think ought to be the goal of
> Alan> collaborative ontology development then put them forward for
> Alan> consideration.
>
> It depends on the ontology that you are developing, is the short
> answer.

Circular. Please elaborate.
>
>
> Phil
>
> >

[Alan Ruttenberg]

On Jul 17, 2008, at 7:43 AM, Phillip Lord wrote:

>
>>>>>> "Alan" == Alan Ruttenberg <alanrut...@gmail.com> writes:
>
>
>>> DNA clearly carries information. If I send you a clone and you
>>> sequence it, then I have communicated something to you, that you
>>> did not have before.
>
> Alan> Shirts clearly carry information. If I send you a shirt and you
> Alan> sew another one that looks like it then I have communicated
> Alan> something to you that you did not have before.
>
> Yes.
>
> Alan> And now I will ask: By your definition, what doesn't carry
> Alan> information?
>
> A random stream of numbers. The patterns created by waves on a wave
> guide. The same shirt that you sent me yesterday (unless you've
> written
> something on it). An evenly distributed stream of monotones.

Each of the things you mention could carry some information in the
sense you mean.
The "same shirt" corresponds to Jonathan Reeses definition
"Information is that which informs". We haven't explored that one
enough yet.

However, I don't think it will solve the problem. It's a kind of
surprising conclusion that information that some entity carries
disappears as soon as you have it. What about other people who see the
same shirt. Is there something carried by the shirt or not?

> Alan> And if you answer everything does, then I will ask what makes
> Alan> DNA special, or, why is not "carries information" a content
> free
> Alan> statement?
>
> If I send you a DNA sequence file with the sequence of pUC19, then the
> file contains information. If a tube of pUC19, it doesn't. I find this
> untenable.
>
> Alan> Most objective would be something like spectrum of light
>>> under a
> Alan> standard spectrum of illumination.
>>>
>>> A (reflection) spectrum is a not the same thing as colour.
>>> Reflection spectra admits to a standard definition, while colour
>>> depends on the perception of the spectra. This is why colour models
>>> reflect trichromatic human perception rather 14-chromatic shrimp
>>> preception. PATO, for example, blesses HSV for this reason.
>
> Alan> Are we disagreeing about anything here?
>
> Colour is not "objective" in the sense that some arbitrary decisions
> have to be made.

Then we agree. It is also not objective in that different people and
animals perceive color differently arising from different qualities of
their sensory organs. Color blindness is just one example. More than
trichromal color is another.

>
>
> Phil
>
> >

[Phillip Lord]

>>>>> "Alan" == Alan Ruttenberg <alanrut...@gmail.com> writes:

>> Alan, why would I lie to you. I am unclear what it means. The only
>> answers I have recieved are: science is hard; I feel it in my
>> bones; we can't define all terms on pain of infinite regress.

Alan> There is 1) Whether there is stuff out there that is real and 2)
Alan> Our ability to describe it.

Alan> One can, and most scientists do, I would claim, understand and
Alan> believe
Alan> (1) and work hard towards (2) - their life's work.

Alan> I think you and others often argue that you don't believe 1,
Alan> possibly for good pedagogic reason, but I really doubt that you
Alan> don't believe it.

So, you saying that I am deliberately lying?

I have no criteria as to how to judge whether something is real
according to the realist principles. Function is real, apparently. Role
is real. However, objectives for a process are not.

As a scientist, I stopped asking whether it's real about the time that I
learnt about genes (as units of inheritence) and the last 150 years of
physics. Do genes really exist? Are electrons really a wave and a
particle. I don't know.

I do, however, understand a model which is descriptive or predicitive
of experimental data. This is what I think science is about. This I
understand. Is this what "real" in the "realism" sense means.

Alan> I don't see how. In both cases the trail leads to the bearer of
Alan> the PDF. There isn't any question of whether they are carried
>> by
Alan> bits of charge, magnetic fields, etc.
>>
>> We need an abstraction which is information.

Alan> Umm.. Information artifact is_a generically dependent continuant
Alan> ?

As I have said, this conflates the book with the paper it's written on.
It's wrong, because information does not need a continuant to exist.

Alan> Given this view every data file (and every ebook) is
>> basically a
Alan> big number. I don't find that view to be particularly
Alan> illuminating or useful.
>>
>> Interpreted in a certain way, yes.

Alan> Don't understand this remark.

For a PDF file, we interpret a number that is the file contents, and
take parts of it to mean certain things so that we can read the file.

63 for example can be one less than 64. Or it can be "a" in ascii.

>> The question has been raised here; it seems to be open to some
>> unclarity. I think Ingvar said that 0 doesn't exist because there
>> are no instances (by definition) and that only 1 is real. Others
>> may differ.

Alan> Perhaps. I haven't spend a lot of time on it, and Ingvar isn't
Alan> here very often. I don't see this as much affecting our work,
Alan> though I do know it's a good way to get a few people out there
Alan> riled up ;-)

I agree, it's not worth spending time on. My point is that asking
whether things really exist, is not worth spending time on. 3 is useful,
therefore it should be in the ontology.

>> The point is that the least interesting, the least important thing
>> about a piece of information is that it's written onto a disk,
>> piece of paper, etc.

Alan> Yes, so we mostly don't have to make statements about the
Alan> bearer. Doesn't mean that it's not there (and essential)

There may be a bearer, but that bearer need not be a continuant.
Information can (literally) exist in a vacuum. Ask an astronomer if you
don't believe me.

>> Not according to the defintions on the web. Qualities "inhere in
>> entities". DependentContinuant -- A continuant [snap:Continuant]
>> that is either dependent on one or other independent continuant
>> [snap:IndependentContinuant] bearers or inheres in or is borne by
>> other entities.
>>
>> Other entities. Qualites can inhere in anything then.

Alan> I will inquire further. Perhaps my understanding of qualities is
Alan> defective.

It's the documentation on the web that's defective. All the examples
inhere in continuants, but the definition does not distinguish between
entity as a word and the BFO:entity term. Hence the confusion.

Alan> Then consider the fact that PDF file is a program with
Alan> instructions on where to put marks on a page. The same
Alan> instructions issued 100 times give you the same result. So a
Alan> PDF, considered a conveyer of information about where to put
Alan> marks on page (as intended), can sometimes be substantially
>> more
Alan> compressed, losslessly, when one knows how to interpret it.
>>
>> I think that this statement is wrong.

Alan> That's not helpful. Care to say why?

Well, this is my understanding of information science. Compression
depends only on the patterns; all the systems that I know of which use
knowledge of the interpretation, such as mp3 for example, are lossy, or
use the fact that information with a particular interpretation often are
non-random with respect to each other; i.e. certain patterns occur more
frequently in mp3 than might happen randomly. These compression
algorithms still don't need to understand the intepretation, just what
part of the potential pattern space than mp3's exist in.

Alan> If you've got an alternative formulation and different
Alan> competency questions that you think ought to be the goal of
Alan> collaborative ontology development then put them forward for
Alan> consideration.
>>
>> It depends on the ontology that you are developing, is the short
>> answer.

Alan> Circular. Please elaborate.

If we take too sets of knowledge in RDF, then we can always integration
with no effort -- that is, we can put your RDF and my RDF into the same
bucket. Likewise, I can integration Macbeth and Romeo and Juliet by
concatening the files.

However, this is a shallow integration. To define a better, deeper
integration needs knowledge of the things we are integration. So two
shakespeare plays might need a unified concordence. Or, we make rewrite
the stage directions so that one company can perform both back to back.

So whether integration has succeeded or not, depends on the questions
that you wish to ask of the integrated data set. Compromise is
inevitable; an integration good for asking one set of questions may be
bad for asking another.

Phil

[Barry Smith]

At 11:24 AM 7/17/2008, Phillip Lord wrote:

> >>>>> "Alan" == Alan Ruttenberg <alanrut...@gmail.com> writes:
>
> >> Alan, why would I lie to you. I am unclear what it means. The only
> >> answers I have recieved are: science is hard; I feel it in my
> >> bones; we can't define all terms on pain of infinite regress.
>
> Alan> There is 1) Whether there is stuff out there that is real and 2)
> Alan> Our ability to describe it.
>
> Alan> One can, and most scientists do, I would claim, understand and
> Alan> believe
> Alan> (1) and work hard towards (2) - their life's work.
>
> Alan> I think you and others often argue that you don't believe 1,
> Alan> possibly for good pedagogic reason, but I really doubt that you
> Alan> don't believe it.
>
>So, you saying that I am deliberately lying?
>
>I have no criteria as to how to judge whether something is real
>according to the realist principles. Function is real, apparently. Role
>is real. However, objectives for a process are not.

Again: I have nowhere stated that objectives are not real.
Merely that we do not yet have a good strategy for fitting any type
deserving this name 'objective' into a single inheritance ontology.
Since we (OBI) do have a good strategy for fitting in a type called
'objective specification' we are testing the consequences of doing so.

>As a scientist, I stopped asking whether it's real about the time that I
>learnt about genes (as units of inheritence) and the last 150 years of
>physics. Do genes really exist? Are electrons really a wave and a
>particle. I don't know.

Scientists do not indeed need to ask this question. Rather they ask
whether given theories have been refuted, and the like.

>I do, however, understand a model which is descriptive or predicitive
>of experimental data. This is what I think science is about. This I
>understand. Is this what "real" in the "realism" sense means.
>
> Alan> I don't see how. In both cases the trail leads to the bearer of
> Alan> the PDF. There isn't any question of whether they are carried
> >> by
> Alan> bits of charge, magnetic fields, etc.
> >>
> >> We need an abstraction which is information.
>
> Alan> Umm.. Information artifact is_a generically dependent continuant
> Alan> ?
>
>As I have said, this conflates the book with the paper it's written on.
>It's wrong, because information does not need a continuant to exist.

We are focusing our energies on the ontology of information artifact
precisely in order to have a narrower domain within which we can get
clear about some very complex issues.
We will need to deal with associated occurrents at a later stage.
We are nowhere stating that information only exists where there are
information artifacts.
BS

[Alan Ruttenberg]

On Jul 17, 2008, at 11:24 AM, Phillip Lord wrote:

>
>>>>>> "Alan" == Alan Ruttenberg <alanrut...@gmail.com> writes:
>
>>> Alan, why would I lie to you. I am unclear what it means. The only
>>> answers I have recieved are: science is hard; I feel it in my
>>> bones; we can't define all terms on pain of infinite regress.
>
> Alan> There is 1) Whether there is stuff out there that is real
> and 2)
> Alan> Our ability to describe it.
>
> Alan> One can, and most scientists do, I would claim, understand and
> Alan> believe
> Alan> (1) and work hard towards (2) - their life's work.
>
> Alan> I think you and others often argue that you don't believe 1,
> Alan> possibly for good pedagogic reason, but I really doubt that
> you
> Alan> don't believe it.
>
> So, you saying that I am deliberately lying?

No. You snipped out where I said "There are more reasons to question

the truth of statements than to fish out lying."

> I have no criteria as to how to judge whether something is real

> according to the realist principles. Function is real, apparently.
> Role
> is real. However, objectives for a process are not.

Barry answered this one.

>
> As a scientist, I stopped asking whether it's real about the time
> that I
> learnt about genes (as units of inheritence) and the last 150 years of
> physics. Do genes really exist? Are electrons really a wave and a
> particle. I don't know.

My answers: The word "Gene" is used in too many senses to have a
coherent definition. Elections are really something that has wave-
like and a particle-like properties. I don't think electrons or the
things people use

>
> I do, however, understand a model which is descriptive or predicitive
> of experimental data. This is what I think science is about. This I
> understand. Is this what "real" in the "realism" sense means.

I have a pretty broad view of what is real. For me, the question is,
how does one sensibly arrange things so that similar kinds of real
are organized in vicinity of each other, so that specification of
inferences one can and should make can be consolidated.

I think a focus on "real" as in "realism" isn't terribly helpful.
(Yes, I know that this has been used on this list - sorry) But do you
think a mouse is real? Rather than saying you don't know what is real
and pointing to the most difficult cases (e.g. quantum) why not start
by us agreeing on the things that we both agree are real, and
describing what they are.

>
> Alan> I don't see how. In both cases the trail leads to the
> bearer of
> Alan> the PDF. There isn't any question of whether they are
> carried by
> Alan> bits of charge, magnetic fields, etc.
>>>
>>> We need an abstraction which is information.
>
> Alan> Umm.. Information artifact is_a generically dependent

> continuant?

>
> As I have said, this conflates the book with the paper it's written
> on.

And as I have said, you simply have this wrong.

> It's wrong, because information does not need a continuant to exist.

Barry responds appropriately.

> Alan> Given this view every data file (and every ebook) is
> basically a
> Alan> big number. I don't find that view to be particularly
> Alan> illuminating or useful.
>>>
>>> Interpreted in a certain way, yes.
> Alan> Don't understand this remark.
>
> For a PDF file, we interpret a number that is the file contents, and
> take parts of it to mean certain things so that we can read the file.
>
> 63 for example can be one less than 64. Or it can be "a" in ascii.

Seems to me like you are sanctioning numerology. Look, if we don't
interpret that PDF file as a PDF file, we are making a mistake. You
are confusing encoding with content. Viewing a PDF file as a number
gives the completely wrong set of inferences. It is a sensible thing
to add one to a number. It is nonsense to add 1 to a PDF file.

>>> The point is that the least interesting, the least important thing
>>> about a piece of information is that it's written onto a disk,
>>> piece of paper, etc.
>
> Alan> Yes, so we mostly don't have to make statements about the
> Alan> bearer. Doesn't mean that it's not there (and essential)
>
> There may be a bearer, but that bearer need not be a continuant.
> Information can (literally) exist in a vacuum. Ask an astronomer if
> you
> don't believe me.

I know the arguments regarding this.

But see http://neurocommons.org/page/
How_much_are_we_anthropomorphizing_when_discussing_what_information_is

There a question here to be considered. But, as Barry points out, we
are currently working on ontology of information artifacts, which is
a deliberately narrow subset of things that people label information.

>>> Not according to the defintions on the web. Qualities "inhere in
>>> entities". DependentContinuant -- A continuant [snap:Continuant]
>>> that is either dependent on one or other independent continuant
>>> [snap:IndependentContinuant] bearers or inheres in or is borne by
>>> other entities.
>>>
>>> Other entities. Qualites can inhere in anything then.
>
> Alan> I will inquire further. Perhaps my understanding of
> qualities is
> Alan> defective.
>
> It's the documentation on the web that's defective. All the examples
> inhere in continuants, but the definition does not distinguish between
> entity as a word and the BFO:entity term. Hence the confusion.

OK. Bug that should be fixed.

>
> Alan> Then consider the fact that PDF file is a program with
> Alan> instructions on where to put marks on a page. The same
> Alan> instructions issued 100 times give you the same result. So a
> Alan> PDF, considered a conveyer of information about where to put
> Alan> marks on page (as intended), can sometimes be substantially
> more
> Alan> compressed, losslessly, when one knows how to interpret it.
>>>
>>> I think that this statement is wrong.
> Alan> That's not helpful. Care to say why?
>
> Well, this is my understanding of information science. Compression
> depends only on the patterns; all the systems that I know of which use
> knowledge of the interpretation, such as mp3 for example, are
> lossy, or
> use the fact that information with a particular interpretation
> often are
> non-random with respect to each other; i.e. certain patterns occur
> more
> frequently in mp3 than might happen randomly. These compression
> algorithms still don't need to understand the intepretation, just what
> part of the potential pattern space than mp3's exist in.

Some compression algorithms are not lossy and do depend on the
interpretation. It is possible to ignore the interpretation, but not
necessary. Compare a compressed version of the ascii text of a bunch
of floating point numbers with that of the binary representation of
same.

But fundamentally, Shannon information is about sending a message -
the amount of information is measured in message space.

Sure you can say that the "message" that any file conveys is the
bits. But that would be wrong.

See also http://prize.hutter1.net/

> Alan> If you've got an alternative formulation and different
> Alan> competency questions that you think ought to be the goal of
> Alan> collaborative ontology development then put them forward for
> Alan> consideration.
>>>
>>> It depends on the ontology that you are developing, is the short
>>> answer.
> Alan> Circular. Please elaborate.
>
> If we take too sets of knowledge in RDF, then we can always
> integration
> with no effort -- that is, we can put your RDF and my RDF into the
> same
> bucket. Likewise, I can integration Macbeth and Romeo and Juliet by
> concatening the files.

I don't call his integration, personally, I call it aggregation.
http://www.golovchenko.org/cgi-bin/wnsearch?q=aggregation#2n

> However, this is a shallow integration. To define a better, deeper
> integration needs knowledge of the things we are integration. So two
> shakespeare plays might need a unified concordence. Or, we make
> rewrite
> the stage directions so that one company can perform both back to
> back.
>
> So whether integration has succeeded or not, depends on the questions
> that you wish to ask of the integrated data set. Compromise is
> inevitable; an integration good for asking one set of questions may be
> bad for asking another.

By this view it is hopeless to build ontologies for general use in
science, because the point of science is asking new questions, ones
that couldn't have been known in advance of writing the ontology.

I'm more optimistic (and more ambitious, I suppose).

-Alan

[Phillip Lord]

>>>>> "Alan" == Alan Ruttenberg <alanrut...@gmail.com> writes:

Alan> Why can't it be an information artifact that is the product
>> of a
Alan> process that statisticians do?
>>
>> I can ask that question about anything. Whether it is or not, I
>> don't. Regardless, it's refers to the population, the aggregate as
>> a whole.

Alan> We agree on this. So do many other things - e.g the ratio of
Alan> people who've held a apple sometime today to the number who
Alan> tripped over an untied shoelace.

Okay.

>> All measurements do have an error -- it's not really a measurement
>> otherwise.

Alan> I disagree. Whether there is an error or not is a matter of
Alan> fact, and depending on the quality, knowing the fact may be
Alan> possible. If I ask you the number of nucleated cells in a sample
Alan> (a measurement) and you give me an answer, if the answer is
Alan> correct there is no error.

Clearly, if the answer is correct there is no error; this is true of
anything. In this case, I think what you are saying is that the number
of nucleated cells is a integer measurement. Integer values are still
open to error -- you don't have to do too much image analysis to realise
that "the number of cells" or worse "the number of nucleated cells" is
definately open to error.

Alan> So I think that the measurements that always have errors are the
Alan> ones that attempt to measure quality values from a dense enough
Alan> space to be considered continuous.

I think this is wrong.

Alan> Who said that we can't measure them and talk about error
Alan> (estimates)?
>>
>> You seem to be implying that measurements are information
>> artifacts. You said "we may never be able to instantiate them".

Alan> Yes, but information artifacts are about things. In this case
Alan> the measurement is about the quality. The representation of
Alan> information artifacts includes this relation. So we talk about
Alan> some qualities by way of the information objects that are the
Alan> results of attempts to measure them.

Alan> You have if you've said what the information artifact is about.
Alan> In order to do that you need the full kinds of ontologies we
Alan> already have. However, by talking about the GDCs and what they
Alan> are about, when appropriate, we avoid ascribing to entities
Alan> properties which don't make sense for them. Error estimates are
Alan> a good example. The error estimate associated with a measurement
Alan> of a height isn't something that is a quality of the person
Alan> whose height it is. Nor is it a quality of a population. It is
Alan> something that relates to measurement methods and our means of
Alan> estimating errors (not trivial, and with continuing controversy
Alan> and new ideas).

This doesn't help if the relation is to another GDC. The error is a GDC
about a GDC for example.

>>>> Ah, okay. Well, the domain and range wouldn't be incorrect then,
>>>> just not as specific as they could be. I'll have a think about
>>>> this. In practice, I'm not 100% sure what you are trying to
>>>> achieve with "is_proxy_for", so I may not have much luck.
>>
Alan> is_proxy_for is to mark theory based connections between the
Alan> entities that we intend to measure and the things we actually
Alan> measure directly. A typical example is where we might measure
Alan> time, but in order to do so we count the number of oscillations
Alan> of a crystal, and use a formula to convert count to time.
>>
>> Can you give me an example of where you would measure time directly
>> rather than through a proxy?

Alan> No :-) !

Actually there is one. If you are counting the waves from Caesium
emittence, then you are measuring time directly.

Alan> But I can list a whole bunch of proxy chains involved with
Alan> different methods of measuring time and discuss the sort of
Alan> systematic and statistical errors associated with each.

Yes, this is handy, although in many cases you need to determine the
errors -- I doubt that they will be inferrable from the steps in the
chain.

Phil

[Alan Ruttenberg]

On Jul 17, 2008, at 4:10 PM, Phillip Lord wrote:

>>> All measurements do have an error -- it's not really a measurement
>>> otherwise.
>
> Alan> I disagree. Whether there is an error or not is a matter of
> Alan> fact, and depending on the quality, knowing the fact may be
> Alan> possible. If I ask you the number of nucleated cells in a
> sample
> Alan> (a measurement) and you give me an answer, if the answer is
> Alan> correct there is no error.
>
> Clearly, if the answer is correct there is no error; this is true of
> anything. In this case, I think what you are saying is that the number
> of nucleated cells is a integer measurement. Integer values are still
> open to error -- you don't have to do too much image analysis to
> realise
> that "the number of cells" or worse "the number of nucleated cells" is
> definately open to error.

Didn't mean to imply not, simply to point out that the statement "All
measurements do have an error" didn't seem to be true.

> Alan> So I think that the measurements that always have errors
> are the
> Alan> ones that attempt to measure quality values from a dense
> enough
> Alan> space to be considered continuous.
>
> I think this is wrong.

Because you think that *some* measurements that attempt to measure
quality values from a dense enough space to be considered continuous
*don't* have errors? Or because you think I am saying that *only*
measurements that attempt to measure quality values from a dense
enough space to be considered continuous have errors? (I am not
saying that)

Why doesn't it help? About is a pretty loose relation (I'd like to
get more specific subproperties defined). I think (though haven't
thought it all the way through) that a GDC G1 about a GDC G2 about X
is also about X (about is transitive). But G1 is also about the error
estimation process./

>>>>> Ah, okay. Well, the domain and range wouldn't be incorrect then,
>>>>> just not as specific as they could be. I'll have a think about
>>>>> this. In practice, I'm not 100% sure what you are trying to
>>>>> achieve with "is_proxy_for", so I may not have much luck.
>>>
> Alan> is_proxy_for is to mark theory based connections between the
> Alan> entities that we intend to measure and the things we actually
> Alan> measure directly. A typical example is where we might measure
> Alan> time, but in order to do so we count the number of
> oscillations
> Alan> of a crystal, and use a formula to convert count to time.
>>>
>>> Can you give me an example of where you would measure time directly
>>> rather than through a proxy?
>
> Alan> No :-) !
>
> Actually there is one. If you are counting the waves from Caesium
> emittence, then you are measuring time directly.

You mean this stuff? http://tycho.usno.navy.mil/cesium.html

[Phillip Lord]

>>>>> "Alan" == Alan Ruttenberg <alanrut...@gmail.com> writes:

>> Alan, why would I lie to you. I am unclear what it means. The only
>> answers I have recieved are: science is hard; I feel it in my
>> bones; we can't define all terms on pain of infinite regress.

Alan> There is 1) Whether there is stuff out there that is real and 2)

Alan> Our ability to describe it.

Alan> One can, and most scientists do, I would claim, understand and

Alan> believe (1) and work hard towards (2) - their life's work.

Alan

If you ask people, "do you believe that things exist", it's obviously
the case that people will say "yes". The question is close enough to be
tautological that it has no meaning.

When I hear statements like "we should only represent things that
exist", I do not need to deny existence to say that I don't understand
this. I am simply stating the without clear criteria for determining
whether something exists, then the statement does not help me get
forward.

Yes, fine, I think that you exist. You seem to me to do so; you could be
a figment of my imagination, of course. But it's easier to assume that
you exist because the alternative leads me to endless solipsism.

Do I think function exists? Do I think that there are units of reality
which are roles. I don't know. I don't really care. I do know that these
things are useful abstractions, useful simplifications over the primary
experimental data which enable me to build descriptive or predictive
models of that experimental data. So I would model them. As a scientist
building predictive and descriptive models is what I do.

As it happens, I don't think that saying ontologies should only contain
entities in reality should be modelled is right either. There are good
practical reasons why you might wish to divert from this at times.

Alan> Given this view every data file (and every ebook) is basically a
Alan> big number. I don't find that view to be particularly
Alan> illuminating or useful.
>>
>> Interpreted in a certain way, yes.

Alan> Don't understand this remark.

This email is a single number -- don't know what it is, although I could
work it out.

There is an agreement between the software that you use and that I use
to interpret parts of that number in certain ways. The whole thing is
encoded in ascii, for example. Parts of it are header, parts of it
message body, and so on.

>> The question has been raised here; it seems to be open to some
>> unclarity. I think Ingvar said that 0 doesn't exist because there
>> are no instances (by definition) and that only 1 is real. Others
>> may differ.

Alan> Perhaps. I haven't spend a lot of time on it, and Ingvar isn't

Alan> here very often. I don't see this as much affecting our work,

Yes, I agree. Whether 1, 2 or 3 are parts of reality don't affect our
work. They ARE part of science. As I say, being part of reality isn't
really a good criteria.

Alan> It precisely hasn't! There is the paper and disk. These are one
Alan> sort of thing. Then there is the information that they bear.
Alan> That is a different sort of thing. There is a connection. This
Alan> is the opposite of conflation. The question above is whether the
Alan> thing that is the information is a class or an instance.
>>
>> There is a conflation. GDC just conflates the two and then says,
>> well it's a generic piece of paper.

Alan> No it doesn't. Who said anything about a generic piece of paper.

Or generic file or any continuant. Even though information can be born
by things which are not continuant.

>> The point is that the least interesting, the least important thing
>> about a piece of information is that it's written onto a disk,
>> piece of paper, etc.

Alan> Yes, so we mostly don't have to make statements about the
Alan> bearer. Doesn't mean that it's not there (and essential) and
Alan> I've argued why I think ontologies ought to account for the
Alan> things that are there.

I don't think you have accounted for it. We don't care what the bearer
is, if the bearer is "generic".

Look at is this way. I could go through any ontology and add a property
to every class, every individual saying "Entity is_part_of Universe".
It's true. It's pointless.

There's no point telling me that information inheres in something,
whatever that thing is. The only thing that GDC says is that GDC's don't
inhere in occurrents. And that's wrong.

I haven't replied to the rest of the mail cause it gave me a strange
sense of deja vu -- did I make a wierd cut and paste error somewhere?

Phil

[Phillip Lord]

>>>>> "w" == waku <lampu...@gmail.com> writes:

w> Phillip Lord wrote:
>>
>> If I send you a DNA sequence file with the sequence of pUC19, then
>> the file contains information. If a tube of pUC19, it doesn't. I
>> find this untenable.
>>

w> what doesn't? doesn't the tube contain information, or doesn't the
w> DNA in the tube contain information?

I've answered this in another email. Not all things can carry
information.

w> (btw. the 'contains' here makes me a bit uncomfortable.)

I'd tend to agree -- it makes sense for the second example, but isn't
great for the first.

Phil

[Phillip Lord]

>>>>> "Alan" == Alan Ruttenberg <alanrut...@gmail.com> writes:

Alan> And now I will ask: By your definition, what doesn't carry
Alan> information?
>>
>> A random stream of numbers. The patterns created by waves on a wave
>> guide. The same shirt that you sent me yesterday (unless you've
>> written something on it). An evenly distributed stream of
>> monotones.

Alan> Each of the things you mention could carry some information in
Alan> the sense you mean.

This isn't right. Random numbers cannot carry information by definition.
If they can, then they are not random.

A pattern on a wave guide can't carry information, although changes to
the pattern can. If you think about a water wave hitting a wall and
creating a spout -- this spout travels down the wall. The spout can
progress at any speed. If it hits the wall exactly perpendicularly, then
it travels infiniately fast. Information can't be transmitted that fast
as it violates relativity. So the water spout can't be carrying
information.

An evenly distributed set of monotones can't carry information -- at
least not while they continue. You have to change the tone, or change
the spacing to convey information.

Alan> The "same shirt" corresponds to Jonathan Reeses definition
Alan> "Information is that which informs". We haven't explored that
Alan> one enough yet.

Alan> However, I don't think it will solve the problem. It's a kind of
Alan> surprising conclusion that information that some entity carries
Alan> disappears as soon as you have it. What about other people who
Alan> see the same shirt. Is there something carried by the shirt or
Alan> not?

Yes, there is, but there is no additional information carried by a
steady stream of identical shirts -- this is the same case as the
monotones. Likewise, you gain the same information (or lack of, if I am
talking gibberish) from this email however many times you read it.

Alan> Are we disagreeing about anything here?
>>
>> Colour is not "objective" in the sense that some arbitrary
>> decisions have to be made.

Alan> Then we agree. It is also not objective in that different people
Alan> and animals perceive color differently arising from different
Alan> qualities of their sensory organs. Color blindness is just one
Alan> example. More than trichromal color is another.

Yes, we agree on this.

Phil

[Alan Ruttenberg]

On Jul 17, 2008, at 11:31 PM, Phillip Lord wrote:

>
>>>>>> "Alan" == Alan Ruttenberg <alanrut...@gmail.com> writes:
>
> Alan> And now I will ask: By your definition, what doesn't carry
> Alan> information?
>>>
>>> A random stream of numbers. The patterns created by waves on a wave
>>> guide. The same shirt that you sent me yesterday (unless you've
>>> written something on it). An evenly distributed stream of
>>> monotones.
>
> Alan> Each of the things you mention could carry some information in
> Alan> the sense you mean.
>
> This isn't right. Random numbers cannot carry information by
> definition.
> If they can, then they are not random.

You used the sense of information as that which informs. Now, I don't
usually spew random numbers, but if all of a sudden I started sending
out emails with streams of random numbers, then people would
interpret that as something wrong with me. That is, the stream would
become information bearing in the sense you used it.

> A pattern on a wave guide can't carry information, although changes to
> the pattern can. If you think about a water wave hitting a wall and
> creating a spout -- this spout travels down the wall. The spout can
> progress at any speed. If it hits the wall exactly perpendicularly,
> then
> it travels infiniately fast. Information can't be transmitted that
> fast
> as it violates relativity. So the water spout can't be carrying
> information.

You're referring to properties of group velocity here, yes? (http://
en.wikipedia.org/wiki/Group_velocity) Seems to me that's rather like
the perceived backwards motion of a wheel one sees sometimes in
movies - caused by the interaction of the periodicity at which the
wheel is imaged and the rotational velocity of the wheel. Sometimes
it looks like the wheel is rotating in the opposite direction to the
one it actually is.

>
> An evenly distributed set of monotones can't carry information -- at
> least not while they continue. You have to change the tone, or change
> the spacing to convey information.

If I don't usually play an evenly distributed set of monotones and
now I start to, then you will know something is up. John cage showed
that silence can be informative.

> Alan> The "same shirt" corresponds to Jonathan Reeses definition
> Alan> "Information is that which informs". We haven't explored that
> Alan> one enough yet.
>
> Alan> However, I don't think it will solve the problem. It's a
> kind of
> Alan> surprising conclusion that information that some entity
> carries
> Alan> disappears as soon as you have it. What about other people who
> Alan> see the same shirt. Is there something carried by the shirt or
> Alan> not?
>
> Yes, there is, but there is no additional information carried by a
> steady stream of identical shirts -- this is the same case as the
> monotones. Likewise, you gain the same information (or lack of, if
> I am
> talking gibberish) from this email however many times you read it.

So the first shirt (A) has something that the second (B) doesn't. And
if I reversed the order of presentation of the shirts, then (B) would
have it and (A) not? Are you sure this information is carried by the
shirts? Seems to be an odd sort of thing that can hop around
depending on what which one is picked up first.

Or is it possible the thing you are calling information carried by
the shirt, is, in fact, the result of properties of your brain.
Frankly, given the phenomena you are describing, the real locus
strikes me as the brain, not the shirt.

In the discussion of information artifact the consensus was that
information was defined by the originator of it, not the receiver.
There is no guarantee that receivers get the intended message and the
range of possible effects of a communication is large. If one viewed
information as being the thing the originator intended it seemed that
there could be a coherent entity. Start with the receiver and not.

That's not to say there isn't something important going on in the
receiver's mind. However the connection between what goes on in their
minds and what is intended by the originator is all over the place,
and trying to base an objective account of information on the
receiving side of things seemed to be a losing proposition.

Once intention is brought into the equation, it becomes harder to
view DNA as carrying information. After all, what are the intentions
of some piece of DNA? Anyways, because we weren't allowed to rule out
DNA as carrying information (otherwise some in our meeting were going
to be upset ;-) we agreed that we should rename the ontology from
Information Entity O to Information Artifact O, that we had a
reasonable scope of within which we could do productive work, and
that as DNA was now out of scope we didn't need to commit to whether
it was information bearing.

The question of occurrents bearing information is not ruled out, as
you would suggest when you say that GDCs are wrong. It's an issue
we're trying to figure out.

-Alan

[Alan Ruttenberg]

On Jul 17, 2008, at 11:14 PM, Phillip Lord wrote:

>>>>>> "Alan" == Alan Ruttenberg <alanrut...@gmail.com> writes:
>
>>> Alan, why would I lie to you. I am unclear what it means. The only
>>> answers I have recieved are: science is hard; I feel it in my
>>> bones; we can't define all terms on pain of infinite regress.
>
> Alan> There is 1) Whether there is stuff out there that is real
> and 2)
> Alan> Our ability to describe it.
>
> Alan> One can, and most scientists do, I would claim, understand and
> Alan> believe (1) and work hard towards (2) - their life's work.
>
> Alan
>
> If you ask people, "do you believe that things exist", it's obviously
> the case that people will say "yes". The question is close enough
> to be
> tautological that it has no meaning.
>
> When I hear statements like "we should only represent things that
> exist", I do not need to deny existence to say that I don't understand
> this. I am simply stating the without clear criteria for determining
> whether something exists, then the statement does not help me get
> forward.

FWIW, the way I understand this is not to try to define a predicate
"is_real" and then apply it to entities to decide whether they belong
in or out of BFO. Rather I try to focus on what I call the "audit
trail to reality" - the way in which the things we represent are
related to things we consider to be uncontroversialy real.

> Yes, fine, I think that you exist. You seem to me to do so; you
> could be
> a figment of my imagination, of course. But it's easier to assume that
> you exist because the alternative leads me to endless solipsism.
>
> Do I think function exists? Do I think that there are units of reality
> which are roles. I don't know. I don't really care. I do know that
> these
> things are useful abstractions, useful simplifications over the
> primary
> experimental data which enable me to build descriptive or predictive
> models of that experimental data. So I would model them. As a
> scientist
> building predictive and descriptive models is what I do.

building predictive and descriptive models of something.

> As it happens, I don't think that saying ontologies should only
> contain
> entities in reality should be modelled is right either. There are good
> practical reasons why you might wish to divert from this at times.

For example? (recalling that in my view the question isn't so much
which things are real, but what part of reality are they?)

> Alan> Given this view every data file (and every ebook) is
> basically a
> Alan> big number. I don't find that view to be particularly
> Alan> illuminating or useful.
>>>
>>> Interpreted in a certain way, yes.
> Alan> Don't understand this remark.
>
> This email is a single number -- don't know what it is, although I
> could
> work it out.

Oh come on! If there is any interpretation going on it is your
interpreting this email as a number.

> There is an agreement between the software that you use and that I use
> to interpret parts of that number in certain ways. The whole thing is
> encoded in ascii, for example. Parts of it are header, parts of it
> message body, and so on.

But "it" is a number? Seems to me you're out on a limb here.

>>> The question has been raised here; it seems to be open to some
>>> unclarity. I think Ingvar said that 0 doesn't exist because there
>>> are no instances (by definition) and that only 1 is real. Others
>>> may differ.
>
> Alan> Perhaps. I haven't spend a lot of time on it, and Ingvar isn't
> Alan> here very often. I don't see this as much affecting our work,
>
> Yes, I agree. Whether 1, 2 or 3 are parts of reality don't affect our
> work. They ARE part of science. As I say, being part of reality isn't
> really a good criteria.

The issue of 1, 2, 3 is that unlike the rest of BFO, there isn't a
clear relation of them to particulars, that audit trail that leads to
something uncontroversially real. That the rest of the entities can
be traced this way is the strongest feature of the BFO and realist
approach. Numbers don't fit (or at least there isn't a theory of the
audit trail). Rather than letting them in, and giving license to
representing things that can't be so traced, they are left out. You
might say: Well, who needs this tracing thing anyways. But this is
one of the *very few* principles that one can use as some objective
measure of elements of an ontology. Without it people propose all
sorts of silly things. Such as that this email is a number ;-)

The thing about saying this email is a number is that it *explains
absolutely nothing*. It is a thought stopper. We need more thought
about what terms mean, not less. The meat is in the "it has a header,
a footer" and further elaboration as to their content and intent. "
What about number speaks to any of that?

> Alan> It precisely hasn't! There is the paper and disk. These are
> one
> Alan> sort of thing. Then there is the information that they bear.
> Alan> That is a different sort of thing. There is a connection. This
> Alan> is the opposite of conflation. The question above is
> whether the
> Alan> thing that is the information is a class or an instance.
>>>
>>> There is a conflation. GDC just conflates the two and then says,
>>> well it's a generic piece of paper.
>
> Alan> No it doesn't. Who said anything about a generic piece of
> paper.
>
> Or generic file or any continuant.

No generic file mentioned either.

> Even though information can be born by things which are not
> continuant.

Think I mentioned in an earlier mail that this is being thought
about. Sorry we haven't caught up to your understanding yet ;-)

>>> The point is that the least interesting, the least important thing
>>> about a piece of information is that it's written onto a disk,
>>> piece of paper, etc.
>
> Alan> Yes, so we mostly don't have to make statements about the
> Alan> bearer. Doesn't mean that it's not there (and essential) and
> Alan> I've argued why I think ontologies ought to account for the
> Alan> things that are there.
>
> I don't think you have accounted for it. We don't care what the bearer
> is, if the bearer is "generic".
>
> Look at is this way. I could go through any ontology and add a
> property
> to every class, every individual saying "Entity is_part_of Universe".
> It's true. It's pointless.
>
> There's no point telling me that information inheres in something,
> whatever that thing is. The only thing that GDC says is that GDC's
> don't
> inhere in occurrents. And that's wrong.

Not every GGC can inhere in every sort of continuant. There will be
further elaboration on this as the ontology is developed. Those
constraints will tighten up what each GDC means, and add the ability
to detect nonsense statements. So there is more than what you suggest.

Regarding information and occurrents - something is happening. It
will need to be accounted for. Whether or not it's sensible to do
that using GDCs (an assumption you are making) is an open question.

-Alan

[Phillip Lord]

>>>>> "Alan" == Alan Ruttenberg <alanrut...@gmail.com> writes:

>> Measurements of the same kind.

Alan> Not enough. Often confidence is built because one can do
Alan> different kinds of measurements of the same thing.

True.

Is it the case, that value of all qualities is never directly statable,
but only through a measurement of some kind?

Phil

[Phillip Lord]

>>>>> "Barry" == Barry Smith <phis...@buffalo.edu> writes:

>> So, you saying that I am deliberately lying?
>>
>> I have no criteria as to how to judge whether something is real
>> according to the realist principles. Function is real, apparently.
>> Role is real. However, objectives for a process are not.

Barry> Again: I have nowhere stated that objectives are not real.

You haven't stated that they are real either. The consequences seem much
the same.

Barry> Merely that we do not yet have a good strategy for fitting any
Barry> type deserving this name 'objective' into a single inheritance
Barry> ontology.

Well, single inheritence doesn't seem to be a big issue. With terms like
FiatObjectPart, BFO requires the use of multiple inheritance in more
specific ontologies.

Barry> Since we (OBI) do have a good strategy for fitting in
Barry> a type called 'objective specification' we are testing the
Barry> consequences of doing so.

Yes, I agree. The issue of reality is not important; the modelling
consequence of design decisions are.

>> As a scientist, I stopped asking whether it's real about the time
>> that I learnt about genes (as units of inheritence) and the last
>> 150 years of physics. Do genes really exist? Are electrons really a
>> wave and a particle. I don't know.

Barry> Scientists do not indeed need to ask this question. Rather they
Barry> ask whether given theories have been refuted, and the like.

Scientists don't need to ask about reality, but ontology builders do?

>> As I have said, this conflates the book with the paper it's written
>> on. It's wrong, because information does not need a continuant to
>> exist.

Barry> We are focusing our energies on the ontology of information
Barry> artifact precisely in order to have a narrower domain within
Barry> which we can get clear about some very complex issues. We will
Barry> need to deal with associated occurrents at a later stage.

If you remove the statement that an piece of information is necessarily
associated with a continuant, then you wouldn't need to. Occurrents
would come for free.

Barry> We are nowhere stating that information only exists where there
Barry> are information artifacts.

This is good. In which case you need to talk about information
independently of these information artifacts.

Phil

[Phillip Lord]

>>>>> "Alan" == Alan Ruttenberg <alanrut...@gmail.com> writes:

Alan> I think you and others often argue that you don't believe 1,
Alan> possibly for good pedagogic reason, but I really doubt that you
Alan> don't believe it.
>>
>> So, you saying that I am deliberately lying?

Alan> No. You snipped out where I said "There are more reasons to
Alan> question the truth of statements than to fish out lying."

I have told you, and am telling you that I don't understand what "for
real" means. This is my position, and this is an accurate statement of
my belief. Either you accept this, you think I am lying or you think I
am confused that I don't know what I believe.

>> I have no criteria as to how to judge whether something is real
>> according to the realist principles. Function is real, apparently.
>> Role is real. However, objectives for a process are not.

Alan> Barry answered this one.

>>
>> As a scientist, I stopped asking whether it's real about the time
>> that I learnt about genes (as units of inheritence) and the last
>> 150 years of physics. Do genes really exist? Are electrons really a
>> wave and a particle. I don't know.

Alan> My answers: The word "Gene" is used in too many senses to have a
Alan> coherent definition.

Is why I qualified it with "unit of inheritence". Call it allelle in the
sense Mendel used it if you will.

Alan> Elections are really something that has wave- like and a
Alan> particle-like properties. I don't think electrons or the things
Alan> people use

Think something got snipped here.

>> I do, however, understand a model which is descriptive or
>> predicitive of experimental data. This is what I think science is
>> about. This I understand. Is this what "real" in the "realism"
>> sense means.

Alan> I have a pretty broad view of what is real. For me, the question
Alan> is, how does one sensibly arrange things so that similar kinds
Alan> of real are organized in vicinity of each other, so that
Alan> specification of inferences one can and should make can be
Alan> consolidated.

Alan> I think a focus on "real" as in "realism" isn't terribly
Alan> helpful.

Yes, I agree. That's what I've been saying; "real" doesn't help.

Alan>(Yes, I know that this has been used on this list -
Alan> sorry) But do you think a mouse is real?

I am happy to agree that anything that I can bump into (or tread on) is
real.

Alan> Rather than saying you don't know what is real and pointing to
Alan> the most difficult cases (e.g. quantum) why not start by us
Alan> agreeing on the things that we both agree are real, and
Alan> describing what they are.

Because "difficult cases" seem to cover most of the things I care out.
Function, gene (allele) and so on. Lets deal

>> As I have said, this conflates the book with the paper it's written
>> on.

Alan> And as I have said, you simply have this wrong.

Okay, we agree to disagree.

>> It's wrong, because information does not need a continuant to
>> exist.

Alan> Barry responds appropriately.

I've answered.

>>
>> For a PDF file, we interpret a number that is the file contents,
>> and take parts of it to mean certain things so that we can read the
>> file.
>>
>> 63 for example can be one less than 64. Or it can be "a" in ascii.

Alan> Seems to me like you are sanctioning numerology. Look, if we
Alan> don't interpret that PDF file as a PDF file, we are making a
Alan> mistake. You are confusing encoding with content. Viewing a PDF
Alan> file as a number gives the completely wrong set of inferences.
Alan> It is a sensible thing to add one to a number. It is nonsense to
Alan> add 1 to a PDF file.

Well, for this we need abstract data types. I agree that "integer" is
not enough infomration to understand the sensible operations that you
can perform on a PDF. My point remains -- many sane operations on a PDF
do not depend the interpretation of the PDF as anything other than a
number.

>>>> The point is that the least interesting, the least important
>>>> thing about a piece of information is that it's written onto a
>>>> disk, piece of paper, etc.
>>
Alan> Yes, so we mostly don't have to make statements about the
Alan> bearer. Doesn't mean that it's not there (and essential)
>>
>> There may be a bearer, but that bearer need not be a continuant.
>> Information can (literally) exist in a vacuum. Ask an astronomer if
>> you don't believe me.

Alan> I know the arguments regarding this.

Alan> But see http://neurocommons.org/page/
Alan> How_much_are_we_anthropomorphizing_when_discussing_what_information_is

It's interesting, but Barry is wrong. DNA does carry information, and
that information is transfered into a protein. It's not a metaphor.

Alan> Some compression algorithms are not lossy and do depend on the
Alan> interpretation.

We agree here -- if you can reduce the size of the number space which
algorithm works on, then you can do better compression.

Alan> But fundamentally, Shannon information is about sending a
Alan> message - the amount of information is measured in message
Alan> space.

Alan> Sure you can say that the "message" that any file conveys is the
Alan> bits. But that would be wrong.

Alan> See also http://prize.hutter1.net/

Well, in this case, the prize seems to have reduced the space to 1; you
know what input you are getting.

>> If we take too sets of knowledge in RDF, then we can always
>> integration with no effort -- that is, we can put your RDF and my
>> RDF into the same bucket. Likewise, I can integration Macbeth and
>> Romeo and Juliet by concatening the files.

Alan> I don't call his integration, personally, I call it aggregation.
Alan> http://www.golovchenko.org/cgi-bin/wnsearch?q=aggregation#2n

Sure; but there is a thin line. If you fix the page numbers into one
sequence then at least the page numbers are integrated.

>> So whether integration has succeeded or not, depends on the
>> questions that you wish to ask of the integrated data set.
>> Compromise is inevitable; an integration good for asking one set of
>> questions may be bad for asking another.

Alan> By this view it is hopeless to build ontologies for general use
Alan> in science, because the point of science is asking new
Alan> questions, ones that couldn't have been known in advance of
Alan> writing the ontology.

I don't think that this is correct. BLAST can search over DNA sequences
which were not in existence when BLAST was created. So you can ask new
questions. You just can't do microarray normalisation with it.

Alan> I'm more optimistic (and more ambitious, I suppose).

I like optimism; it's a good thing!

Phil

[Alan Ruttenberg]

I don't believe so. Diploid, from our discussion. Nucleated. But
there is a notable set of them that seem to have this property - the
ones whose determinate instances are taken from sets that are dense
enough that, for all practical purposes, they resemble a continuous
space. However there are exceptions here and there. Exact
dimensionless numbers related to pi, or other notable transcendental
numbers, or rational numbers, for example, as qualities of certain
things, might be able to be inferred from other facts.

-Alan

[Alan Ruttenberg]

On Jul 18, 2008, at 9:49 AM, Phillip Lord wrote:

>>>>>> "Alan" == Alan Ruttenberg <alanrut...@gmail.com> writes:
>>>
>>> As a scientist, I stopped asking whether it's real about the time
>>> that I learnt about genes (as units of inheritence) and the last
>>> 150 years of physics. Do genes really exist? Are electrons really a
>>> wave and a particle. I don't know.
>
> Alan> My answers: The word "Gene" is used in too many senses to
> have a
> Alan> coherent definition.
>
> Is why I qualified it with "unit of inheritence". Call it allelle in
> the
> sense Mendel used it if you will.

I'm less interested in names, and more interested in definitions. The
way I would assess such a question is to try to get as clear as
possible on what you mean, and then think about it. I'm not at the
point that I know what you mean yet.

> Alan> Elections are really something that has wave- like and a
> Alan> particle-like properties. I don't think electrons or the things
> Alan> people use
>
> Think something got snipped here.

Yup. Not sure what anymore :)

>>> I do, however, understand a model which is descriptive or
>>> predicitive of experimental data. This is what I think science is
>>> about. This I understand. Is this what "real" in the "realism"
>>> sense means.
>
> Alan> I have a pretty broad view of what is real. For me, the
> question
> Alan> is, how does one sensibly arrange things so that similar kinds
> Alan> of real are organized in vicinity of each other, so that
> Alan> specification of inferences one can and should make can be
> Alan> consolidated.
>
> Alan> I think a focus on "real" as in "realism" isn't terribly
> Alan> helpful.
>
> Yes, I agree. That's what I've been saying; "real" doesn't help.

It helps me :) Though I tend to try to think about what is
"real" (period) rather than what is "real as in realism".

> Alan>(Yes, I know that this has been used on this list -
> Alan> sorry) But do you think a mouse is real?
>
> I am happy to agree that anything that I can bump into (or tread on)
> is
> real.

Hurray! Common ground!

> Alan> Rather than saying you don't know what is real and pointing to
> Alan> the most difficult cases (e.g. quantum) why not start by us
> Alan> agreeing on the things that we both agree are real, and
> Alan> describing what they are.
>
> Because "difficult cases" seem to cover most of the things I care out.
> Function, gene (allele) and so on. Lets deal

Your turn to lose a snip...

The strategy towards understanding what to represent is, in the two
cases, similar, but there are differences. General strategy is to
first get clear about what you mean when you use those terms. At least
identify things that that and things that are not, and some criteria
for telling the difference. Test whether those criteria can be
communicated - can others apply them and arrive at the same
conclusions? Then try to do similar exercises with others, not
assuming that they mean the same thing when they use the word.

Having collected a number of these, look for what is common about
them. Is there clearly a central thing of which the different uses
are simple variants? Are there two things that are closely related but
of different sorts. e.g. with function the senses tend to cluster
around the realizable entity sort of thing and the process it is
realized in. That's good because one implies the other. Sometimes
there are clearly different sorts of things being talked about. Start
numbering senses and giving each of them clear definitions. In the
definitions try to ensure that there is at least some set of relations
that eventually relate entity that is being define to something you
can bump into.

In the case of function, as I've mentioned, there tend to be two
closely related senses - the process view (where the function bearing
thing functions) and the "what is it about the thing that lets it act
in that process that way" view. Fortunately these are easily related
and the work is towards consensus on what each of these entities
should be named. What I see as the open question with Function is, in
some sense, how is it that the thing that has a function got it. This
is closely related to the question of the difference between what a
function and disposition is. There have been various attempts in the
direction of helping make this distinction - Barry has his "functions
are beneficial" point of view (which I find difficult to work with).
For manufactured objects, a function of them is to do what they were
designed to do. For biological entities "purpose" doesn't work, though
my working assumption is that we will be able to make the distinction
clear by thinking in terms of selection and evolution.

In the case of gene, there isn't really as clear a set of senses. I
think we need to break it down into the set of senses, and then do the
"give each a name" thing, but no one ever seems willing to part with
their own sense of "gene". So I've generally stayed away from trying
to figure out whether it's "real" in some sense, because I can't pin
down what it is I would be saying that about.

>>> For a PDF file, we interpret a number that is the file contents,
>>> and take parts of it to mean certain things so that we can read the
>>> file.
>>>
>>> 63 for example can be one less than 64. Or it can be "a" in ascii.
>
> Alan> Seems to me like you are sanctioning numerology. Look, if we
> Alan> don't interpret that PDF file as a PDF file, we are making a
> Alan> mistake. You are confusing encoding with content. Viewing a PDF
> Alan> file as a number gives the completely wrong set of inferences.
> Alan> It is a sensible thing to add one to a number. It is nonsense
> to
> Alan> add 1 to a PDF file.
>
> Well, for this we need abstract data types.

Your abstract data types, are, I suspect, related to what we call
information artifacts and their parts. Maybe.

> I agree that "integer" is not enough infomration to understand the
> sensible operations that you
> can perform on a PDF. My point remains -- many sane operations on a
> PDF do not depend the interpretation of the PDF as anything other
> than a number.

Sure, and I can carry many different kinds of things in a bag, without
knowing exactly what they are. But I'm in the business of
communicating what the the things in the bag are, and knowing that
they can be carried in a bag only gives me the smallest bit of clue
about that.

>>>>> The point is that the least interesting, the least important
>>>>> thing about a piece of information is that it's written onto a
>>>>> disk, piece of paper, etc.
>>>
> Alan> Yes, so we mostly don't have to make statements about the
> Alan> bearer. Doesn't mean that it's not there (and essential)
>>>
>>> There may be a bearer, but that bearer need not be a continuant.
>>> Information can (literally) exist in a vacuum. Ask an astronomer if
>>> you don't believe me.
>
> Alan> I know the arguments regarding this.
>
> Alan> But see http://neurocommons.org/page/
> Alan>
> How_much_are_we_anthropomorphizing_when_discussing_what_information_is
>
> It's interesting, but Barry is wrong. DNA does carry information, and
> that information is transfered into a protein. It's not a metaphor.

I'm amazed at your certainty :)
How about turning some of that intuition into the beginnings of a
definition, so that there's a chance someone else might be able to
reproduce it.

> Alan> But fundamentally, Shannon information is about sending a
> Alan> message - the amount of information is measured in message
> Alan> space.
>
> Alan> Sure you can say that the "message" that any file conveys is
> the
> Alan> bits. But that would be wrong.
>
> Alan> See also http://prize.hutter1.net/
>
> Well, in this case, the prize seems to have reduced the space to 1;
> you
> know what input you are getting.

That doesn't quite seem to help understand the problem they are trying
to address in the context. p log p is 0 if you have 1 token with
probability 1, which sends away the Shannon information theoretic
analysis mechanisms.

How does what they are doing there fit into your idea about what
information is?

>>> If we take too sets of knowledge in RDF, then we can always
>>> integration with no effort -- that is, we can put your RDF and my
>>> RDF into the same bucket. Likewise, I can integration Macbeth and
>>> Romeo and Juliet by concatening the files.
>
> Alan> I don't call his integration, personally, I call it
> aggregation.
> Alan> http://www.golovchenko.org/cgi-bin/wnsearch?q=aggregation#2n
>
> Sure; but there is a thin line. If you fix the page numbers into one
> sequence then at least the page numbers are integrated.

I'll be honest - I really don't know what you are talking about when
you talk about integrating Macbeth and Romeo. If you think this is a
useful example to pursue, I'm going to need to need some more
information about what you mean by integration.

>>> So whether integration has succeeded or not, depends on the
>>> questions that you wish to ask of the integrated data set.
>>> Compromise is inevitable; an integration good for asking one set of
>>> questions may be bad for asking another.
>
> Alan> By this view it is hopeless to build ontologies for general use
> Alan> in science, because the point of science is asking new
> Alan> questions, ones that couldn't have been known in advance of
> Alan> writing the ontology.
>
> I don't think that this is correct. BLAST can search over DNA
> sequences
> which were not in existence when BLAST was created. So you can ask new
> questions. You just can't do microarray normalisation with it.

There's something wrong with this analogy, but I'm running out of
steam. Will try to respond later.

[Phillip Lord]

>>>>> "Alan" == Alan Ruttenberg <alanrut...@gmail.com> writes:

Alan> Didn't mean to imply not, simply to point out that the statement
Alan> "All measurements do have an error" didn't seem to be true.

We could argue it both ways. It's possible that some measures have
errors where the error is zero.

>> I think this is wrong.

Alan> Because you think that *some* measurements that attempt to
Alan> measure quality values from a dense enough space to be
Alan> considered continuous *don't* have errors? Or because you think
Alan> I am saying that *only* measurements that attempt to measure
Alan> quality values from a dense enough space to be considered
Alan> continuous have errors? (I am not saying that)

I thought the latter.

>>
>> This doesn't help if the relation is to another GDC. The error is a
>> GDC about a GDC for example.

Alan> Why doesn't it help? About is a pretty loose relation (I'd like
Alan> to get more specific subproperties defined). I think (though
Alan> haven't thought it all the way through) that a GDC G1 about a
Alan> GDC G2 about X is also about X (about is transitive). But G1 is
Alan> also about the error estimation process./

I think I was suggesting that you get back to everything being a GDC.

>>>> Can you give me an example of where you would measure time
>>>> directly rather than through a proxy?
>>
Alan> No :-) !
>>
>> Actually there is one. If you are counting the waves from Caesium
>> emittence, then you are measuring time directly.

Alan> You mean this stuff? http://tycho.usno.navy.mil/cesium.html

Time, in SI, is defined in terms of the frequency of emittence from
caesium. If you are counting waves of this form, then you are measuring
time (in numbers of seconds) by definition. The same is true for all
other SI units; you are only measuing mass directly if you are comparing
to the kilo prototype.

Phil

[Phillip Lord]

>>>>> "Alan" == Alan Ruttenberg <alanrut...@gmail.com> writes:

>>
>> When I hear statements like "we should only represent things that
>> exist", I do not need to deny existence to say that I don't
>> understand this. I am simply stating the without clear criteria for
>> determining whether something exists, then the statement does not
>> help me get forward.

Alan> FWIW, the way I understand this is not to try to define a
Alan> predicate "is_real" and then apply it to entities to decide
Alan> whether they belong in or out of BFO. Rather I try to focus on
Alan> what I call the "audit trail to reality" - the way in which the
Alan> things we represent are related to things we consider to be
Alan> uncontroversialy real.

Well, it's good that we have backed away from representing entities in
reality to "audit trail to reality". But then, we have concepts. Fiat
boundaries are things that really exist therefore fiat boundaries exist.
PRoteins exist, therefore protein functions can exist.

>> Do I think function exists? Do I think that there are units of
>> reality which are roles. I don't know. I don't really care. I do
>> know that these things are useful abstractions, useful
>> simplifications over the primary experimental data which enable me
>> to build descriptive or predictive models of that experimental
>> data. So I would model them. As a scientist building predictive and
>> descriptive models is what I do.

Alan> building predictive and descriptive models of something.

Experimental data; or in the case of predictive models experimental data
that we hope that we will get.

>> As it happens, I don't think that saying ontologies should only
>> contain entities in reality should be modelled is right either.
>> There are good practical reasons why you might wish to divert from
>> this at times.

Alan> For example? (recalling that in my view the question isn't so
Alan> much which things are real, but what part of reality are they?)

Well, numbers for example. Or, indeed, other parts of continuous
mathematics.

>> This email is a single number -- don't know what it is, although I
>> could work it out.

Alan> Oh come on! If there is any interpretation going on it is your
Alan> interpreting this email as a number.

It's a set of binary digits, right. It's a number.

>> There is an agreement between the software that you use and that I
>> use to interpret parts of that number in certain ways. The whole
>> thing is encoded in ascii, for example. Parts of it are header,
>> parts of it message body, and so on.

Alan> But "it" is a number? Seems to me you're out on a limb here.

It's a number. We can interpret it as more, I agree.

Alan> The issue of 1, 2, 3 is that unlike the rest of BFO, there isn't
Alan> a clear relation of them to particulars, that audit trail that
Alan> leads to something uncontroversially real. That the rest of the
Alan> entities can be traced this way is the strongest feature of the
Alan> BFO and realist approach. Numbers don't fit (or at least there
Alan> isn't a theory of the audit trail). Rather than letting them in,
Alan> and giving license to representing things that can't be so
Alan> traced, they are left out.

Thereby removing relationships to one of the most powerful features of
science. I don't think that is a good design principle. Without numbers,
we have no science.

Ironically, the current new foundary principles suggest:

"terms in an ontology should correspond to instances in reality
(including both physical and not, e.g. mathematical descriptions)"

So, mathematical descriptions do have an audit trail to reality, it
appears. So $dx/dt$ has a clear relationship between particulars and
universals. But numbers don't. Or maybe they do.

Perhaps you can see why I get confused here; moreover, the confusion is
unnecessary. We need numbers. I don't care whether they have an audit
trail.

Alan> You might say: Well, who needs this tracing thing anyways. But
Alan> this is one of the *very few* principles that one can use as
Alan> some objective measure of elements of an ontology.

There is no objective measure here, not until you give me some criteria,
which I have asked for.

Alan> Without it people propose all sorts of silly things.

I agree. Having some methodology is better than having none at all.

Alan> The thing about saying this email is a number is that it
Alan> *explains absolutely nothing*. It is a thought stopper.

I don't think this is true. It tells you where information goes in an
ontology; that is, it goes in the same part as numbers, patterns and so
on.

Alan> We need more thought about what terms mean, not less. The meat
Alan> is in the "it has a header, a footer" and further elaboration as
Alan> to their content and intent. " What about number speaks to any
Alan> of that?

Nothing. You need to state that as well.

Alan> Not every GGC can inhere in every sort of continuant. There will
Alan> be further elaboration on this as the ontology is developed.
Alan> Those constraints will tighten up what each GDC means, and add
Alan> the ability to detect nonsense statements. So there is more than
Alan> what you suggest.

Alan> Regarding information and occurrents - something is happening.
Alan> It will need to be accounted for. Whether or not it's sensible
Alan> to do that using GDCs (an assumption you are making) is an open
Alan> question.

The assumption that I am making is that it's not sensible to do that
with using GDCs; as I have said, GDC is wrong, precisely because the
minor issue --- that information is carried by something -- has been
taken to me the major one.

Phil

[Phillip Lord]

>>>>> "Alan" == Alan Ruttenberg <alanrut...@gmail.com> writes:

>> This isn't right. Random numbers cannot carry information by
>> definition. If they can, then they are not random.

Alan> You used the sense of information as that which informs. Now, I
Alan> don't usually spew random numbers, but if all of a sudden I
Alan> started sending out emails with streams of random numbers, then
Alan> people would interpret that as something wrong with me. That is,
Alan> the stream would become information bearing in the sense you
Alan> used it.

This doesn't work. They might interpret the start or stop in this way --
if I send you short bursts of white (random) noise in morse code, it's
not the random noise which carries information. Even if you just send
out a burst of random numbers they might just happen to be a congent
email.

Really, you can't get around this. Randomness is, by definition,
meaningless.

>> A pattern on a wave guide can't carry information, although changes
>> to the pattern can. If you think about a water wave hitting a wall
>> and creating a spout -- this spout travels down the wall. The spout
>> can progress at any speed. If it hits the wall exactly
>> perpendicularly, then it travels infiniately fast. Information
>> can't be transmitted that fast as it violates relativity. So the
>> water spout can't be carrying information.

Alan> You're referring to properties of group velocity here, yes?
Alan> (http:// en.wikipedia.org/wiki/Group_velocity) Seems to me
Alan> that's rather like the perceived backwards motion of a wheel one
Alan> sees sometimes in movies - caused by the interaction of the
Alan> periodicity at which the wheel is imaged and the rotational
Alan> velocity of the wheel. Sometimes it looks like the wheel is
Alan> rotating in the opposite direction to the one it actually is.

Here is a simpler example which is much the same thing. If I shine a
laser at the moon from earth and wiggle it backward and forward, the
spot can easily move faster than light. But the spot can't carry
information -- any change you make to the pattern of movement will
propagate to the moon at the speed of light.

http://math.ucr.edu/home/baez/physics/Relativity/SpeedOfLight/FTL.html#3

>> An evenly distributed set of monotones can't carry information --
>> at least not while they continue. You have to change the tone, or
>> change the spacing to convey information.

Alan> If I don't usually play an evenly distributed set of monotones
Alan> and now I start to, then you will know something is up. John
Alan> cage showed that silence can be informative.

The length of the gap can be informative. But if the monotones start or
stop, then they are not evenly spaced at either end.

>> Yes, there is, but there is no additional information carried by a
>> steady stream of identical shirts -- this is the same case as the
>> monotones. Likewise, you gain the same information (or lack of, if
>> I am talking gibberish) from this email however many times you read
>> it.

Alan> So the first shirt (A) has something that the second (B)
Alan> doesn't. And if I reversed the order of presentation of the
Alan> shirts, then (B) would have it and (A) not? Are you sure this
Alan> information is carried by the shirts? Seems to be an odd sort of
Alan> thing that can hop around depending on what which one is picked
Alan> up first.

I said ther is no additional information carried. Reversing the order
has no effect. The total amount of information you have recieved remains
the same.

Information isn't really instantiable. If I tell you "10" and then later
"10" again, I've only told you about one number.

Alan> In the discussion of information artifact the consensus was that
Alan> information was defined by the originator of it, not the
Alan> receiver.

I don't think it matters; that there could be a reciever is enough.

Alan> Once intention is brought into the equation, it becomes harder
Alan> to view DNA as carrying information. After all, what are the
Alan> intentions of some piece of DNA?

If I send you a file with the DNA sequence of pUC19, I have an
intention. If I send you a piece of pUC19, I have an intention.

This does not mean that DNA necessarily carries information. A random
sequence of DNA carries no information. If I send you a piece of DNA,
with the same sequence as yesterday, I send you no new information.

Alan> The question of occurrents bearing information is not ruled out,
Alan> as you would suggest when you say that GDCs are wrong. It's an
Alan> issue we're trying to figure out.

Good.

Phil

[Alan Ruttenberg]

On Jul 18, 2008, at 3:30 PM, Phillip Lord wrote:

> Really, you can't get around this. Randomness is, by definition,
> meaningless.

Funny you should mention definition. I'm still looking for some of
those ;-)

But let's try again. You are listening to music on the radio. All of a
sudden the sound becomes white noise. You at least know something has
happened, and that it probably isn't good. Was information involved in
this determination?

You are trying to repair the radio. From experience and knowledge of
the circuitry you're able to narrow down where the problem is
depending on what the noise sounds like (after all, there are
different frequency space distributions for time domain noise). Did
the noise have information?

> Here is a simpler example which is much the same thing. If I shine a
> laser at the moon from earth and wiggle it backward and forward, the
> spot can easily move faster than light. But the spot can't carry
> information -- any change you make to the pattern of movement will
> propagate to the moon at the speed of light.
>
> http://math.ucr.edu/home/baez/physics/Relativity/SpeedOfLight/FTL.html#3

What this says is that whatever information is, it has to obey the
laws of physics. I'm not sure what else you mean it to suggest.

>>> An evenly distributed set of monotones can't carry information --
>>> at least not while they continue. You have to change the tone, or
>>> change the spacing to convey information.
>
> Alan> If I don't usually play an evenly distributed set of monotones
> Alan> and now I start to, then you will know something is up. John
> Alan> cage showed that silence can be informative.
>
> The length of the gap can be informative. But if the monotones start
> or
> stop, then they are not evenly spaced at either end.

Then you are talking about fiction, since there is no evenly spaced
series of monotones that never started and will never end.

>>> Yes, there is, but there is no additional information carried by a
>>> steady stream of identical shirts -- this is the same case as the
>>> monotones. Likewise, you gain the same information (or lack of, if
>>> I am talking gibberish) from this email however many times you read
>>> it.
>
> Alan> So the first shirt (A) has something that the second (B)
> Alan> doesn't. And if I reversed the order of presentation of the
> Alan> shirts, then (B) would have it and (A) not? Are you sure this
> Alan> information is carried by the shirts? Seems to be an odd sort
> of
> Alan> thing that can hop around depending on what which one is picked
> Alan> up first.
>
> I said ther is no additional information carried. Reversing the order
> has no effect. The total amount of information you have recieved
> remains
> the same.

That's not the point I was trying to get at. I'm doing my "audit trail
to reality" and trying to figure out where the trail that starts at
the information ends. I'm still trying to divine what you think the
nature of information is, and would hope that that includes some idea
of where that trail ends. It can't be in a specific shirt, since
apparently switching the shirts around doesn't make a difference? What
is the source of the information?

> Information isn't really instantiable. If I tell you "10" and then
> later
> "10" again, I've only told you about one number.

I don't have any idea what you mean when you say "Information isn't
really instantiable"

> Alan> In the discussion of information artifact the consensus was
> that
> Alan> information was defined by the originator of it, not the
> Alan> receiver.
>
> I don't think it matters; that there could be a reciever is enough.

Aside for "I don't think it matters", I believe that your statement
implies the same thing. If the only condition on it being information
is that there could be a receiver, then there will be cases where the
receiver doesn't happen to exist. Does that mean the information
didn't exist? If you believe it did, then the only participant in
process involving the information is the originator, and this makes
the auditing easy - the information exists by virtue of the originator.

Once that is settled, a second assessment needs to be made. Does there
have to be something special about the originator. For example, does
the originator have to have intention to created information? If the
answer is that there need not be anything special about the
originator, then we conclude that every entity in the world is always
emanating information. If that's the property of information then I
don't find the story very convincing as there is no differentiation
among information generating processes. In fact there does seem to be
some differentiation.

> Alan> Once intention is brought into the equation, it becomes harder
> Alan> to view DNA as carrying information. After all, what are the
> Alan> intentions of some piece of DNA?
>
> If I send you a file with the DNA sequence of pUC19, I have an
> intention. If I send you a piece of pUC19, I have an intention.
>
> This does not mean that DNA necessarily carries information. A random
> sequence of DNA carries no information.

We don't have good ways to assess what "random" sequence means for
DNA. Regardless of that, certainly if my intention was to have you
synthesize some DNA for me, even if the sequence is "random",
information is carried by it. How do I know? I get back the
synthesized DNA and it has the same sequence as the one I sent. How
else, if the sent DNA didn't carry information, could the round trip
have happened.

> If I send you a piece of DNA,
> with the same sequence as yesterday, I send you no new information.

This is better that the way you put it earlier. It can still have
information, just not new (to you) information. The newness is a
property of you, not a property of the information carried by the DNA.

-Alan

[Alan Ruttenberg]

On Jul 18, 2008, at 2:36 PM, Phillip Lord wrote:
>>>>>> "Alan" == Alan Ruttenberg <alanrut...@gmail.com> writes:
>
> Well, it's good that we have backed away from representing entities in
> reality to "audit trail to reality".

We haven't backed down from reality being important, just offered
another way to understand the nature of entities.

> But then, we have concepts. Fiat boundaries are things that really
> exist therefore fiat boundaries exist.

Can't parse that.

> Proteins exist, therefore protein functions can exist.
Yes.

>>> Do I think function exists? Do I think that there are units of
>>> reality which are roles. I don't know. I don't really care. I do
>>> know that these things are useful abstractions, useful
>>> simplifications over the primary experimental data which enable me
>>> to build descriptive or predictive models of that experimental
>>> data. So I would model them. As a scientist building predictive and
>>> descriptive models is what I do.
>
> Alan> building predictive and descriptive models of something.
>
> Experimental data; or in the case of predictive models experimental
> data
> that we hope that we will get.

I thought we agreed that this wasn't enough - that the data was about
something.

>>> As it happens, I don't think that saying ontologies should only
>>> contain entities in reality should be modelled is right either.
>>> There are good practical reasons why you might wish to divert from
>>> this at times.
>
> Alan> For example? (recalling that in my view the question isn't so
> Alan> much which things are real, but what part of reality are they?)
>
> Well, numbers for example. Or, indeed, other parts of continuous
> mathematics.
>
>>> This email is a single number -- don't know what it is, although I
>>> could work it out.
>
> Alan> Oh come on! If there is any interpretation going on it is your
> Alan> interpreting this email as a number.
>
> It's a set of binary digits, right. It's a number.

For something to *be* a number it's nature should match that of a
number. Email doesn't. Once can interpret an email as a number (not
uniquely, btw - using a different character encoding doesn't change
the email but does change the number) and then play with it as such,
but it isn't an email message you are playing with.

>>> There is an agreement between the software that you use and that I
>>> use to interpret parts of that number in certain ways. The whole
>>> thing is encoded in ascii, for example. Parts of it are header,
>>> parts of it message body, and so on.
>
> Alan> But "it" is a number? Seems to me you're out on a limb here.
>
> It's a number. We can interpret it as more, I agree.

It isn't is_a number in any meaningful way. OK. We're just going to go
back and for on this. Let's stop.

> Alan> The issue of 1, 2, 3 is that unlike the rest of BFO, there
> isn't
> Alan> a clear relation of them to particulars, that audit trail that
> Alan> leads to something uncontroversially real. That the rest of the
> Alan> entities can be traced this way is the strongest feature of the
> Alan> BFO and realist approach. Numbers don't fit (or at least there
> Alan> isn't a theory of the audit trail). Rather than letting them
> in,
> Alan> and giving license to representing things that can't be so
> Alan> traced, they are left out.
>
> Thereby removing relationships to one of the most powerful features of
> science. I don't think that is a good design principle. Without
> numbers,
> we have no science.

No. Thereby not putting numbers into particular ontology documents
that serve certain purposes. No one is saying not to use numbers. And
numbers appear as the magnitude of measurements, so it's not like they
are not used in the context of ontologies in OBOF. It's just that we
don't mix them in with stuff of a different sort.

> Ironically, the current new foundary principles suggest:

Not sure what's ironic.

>
> "terms in an ontology should correspond to instances in reality
> (including both physical and not, e.g. mathematical descriptions)"
>
> So, mathematical descriptions do have an audit trail to reality, it
> appears. So $dx/dt$ has a clear relationship between particulars and
> universals. But numbers don't. Or maybe they do.

Mathematical descriptions are information artifacts. In fact, in a
separate thread I've proposed to add them to the IAO.

> Perhaps you can see why I get confused here; moreover, the confusion
> is
> unnecessary. We need numbers. I don't care whether they have an audit
> trail.

We differ on how you should resolve your confusion. As I've said,
we're trying to create quality ontologies and a fundamental check is
this audit trail. So *we* care whether they have an audit trail
because our experience tells us that if you get rid of this principal,
quality falls.

> Alan> You might say: Well, who needs this tracing thing anyways. But
> Alan> this is one of the *very few* principles that one can use as
> Alan> some objective measure of elements of an ontology.
>
> There is no objective measure here, not until you give me some
> criteria,
> which I have asked for.

I gave you one. Does an audit trail exist. Is the proposed trail
correct from a scientific perspective. There are more - I've tried to
give examples in our correspondence. But remember too that ontology
development is a collaborative process - within the OBO Foundry
participants have committed to try to develop more measures.

>
> Alan> Without it people propose all sorts of silly things.
>
> I agree. Having some methodology is better than having none at all.
>
>
> Alan> The thing about saying this email is a number is that it
> Alan> *explains absolutely nothing*. It is a thought stopper.
>
> I don't think this is true. It tells you where information goes in an
> ontology; that is, it goes in the same part as numbers, patterns and
> so
> on.

Numbers are one thing. They don't go with numbers. They go with things
that are used to communicate with, at least. Patterns are closer. If
you toss all digital media into number don't you think you will be
losing something? What exactly are the inferences that you think
follow from categorizing something as a number?

> Alan> We need more thought about what terms mean, not less. The meat
> Alan> is in the "it has a header, a footer" and further elaboration
> as
> Alan> to their content and intent. " What about number speaks to any
> Alan> of that?
>
> Nothing. You need to state that as well.

Sure. IAO work in progress. We already have parts of this - "Narrative
object", for example, from the OBI Denrie branch.

> Alan> Not every GGC can inhere in every sort of continuant. There
> will
> Alan> be further elaboration on this as the ontology is developed.
> Alan> Those constraints will tighten up what each GDC means, and add
> Alan> the ability to detect nonsense statements. So there is more
> than
> Alan> what you suggest.
>
> Alan> Regarding information and occurrents - something is happening.
> Alan> It will need to be accounted for. Whether or not it's sensible
> Alan> to do that using GDCs (an assumption you are making) is an open
> Alan> question.
>
>
> The assumption that I am making is that it's not sensible to do that
> with using GDCs; as I have said, GDC is wrong, precisely because the
> minor issue --- that information is carried by something -- has been
> taken to me the major one.

You are distracted by bookkeeping. That someone keeps their checkbook
balance up to date doesn't imply that they can't spend their money
sensibly.

-Alan

>
>
> Phil
>
> >