At 04:37 AM 12/11/2009, P. Def wrote:Of course it can. And should. Literary works and melodies exist just as much as do journal publications or protein databases or library catalogs. The issue is one of aboutness: do the former stand in aboutness relations to entities in reality, as the latter do?
Thank you for the clarification,
I am however not sure whether I correctly understand the difference between an idea or a formula, and a literary or musical work:
if an idea can be considered to have been concretized into the real world by the mere fact of residing in the brain of a person,
then why couldn't a literary work or a melody also be regarded as an actual entity by virtue of its existence in the brain of someone?
Much rather the literary work is, in large degree, at least, a combination of signs. It does not refer to a combination of signs. It is not about itself.
I understand that a literary or a musical work does not as such refer to any actual entity that subsists in the real world
but rather to that particular combination of signs and symbols, or to that particular arrangement of notes that constitutes the work.
Literary works and melodies are indeed GDCs in the sense you describe. But are they information artifacts. If 'information artifact' is defined in terms of aboutness, then no. This is in part a definitional matter, and we should not argue over definitions. However, the assumption we have been following thus far is that, when questions like:
yet I don't understand what would prevent it from being regarded as a particular GDC that may inhere into an independent continuant (e.g. a book)
and thereby lead to the creation of a SDC that would qualify that particular medium of expression as an information bearer of the GDC in the real world.
how much information is in that melody?
how much information is in that novel?
are given negative answers, this is because they are not (generally) about anything in the real world.
Yes, I had a very similar question. I'm interested in curating
misunderstandings, ambiguity, version skew, etc. with web standards
and biodiversity informatics as use cases. Bjoern mentioned a previous
agreement to limit IAO to situations where meaning was
uncontroversial, but Alan has been encouraging me to bring my own
issues to the group, so this dissonance leaves me unsure as to whether
I'm being disruptive, or just joining in.
Maybe we need tracks: [OBI] [meaning] [culture] in the subject line,
or something like that... or multiple lists, or multiple ontologies. I
don't know.
I take a different point of view on this matter, based perhaps on a broader reading of what the about relation about means, I expect. To be fair I think it is a minority opinion.
>>and if so, would an (imaginary) novel that is based on actual facts
>>also qualify as an ICE? or would only certain parts of it qualify as an ICE?
In the IAO sense, an information entity can be about something by mentioning it (mentions is a subproperty of is_about).  In that sense, a novel can be about trees, even if it is a fictional story about wizards. The acts that the are ascribed to the wizards in the story have no direct relation to anything that exists, taken as a whole, but their parts do  - waving of hands, smoke, the thing that appears in the smoke.
Similarly, I think music has elements that refer - sometimes to actual sounds - deliberate mimicry of natural rhythms or the sounds of actual processes, and sometimes to other elements of other pieces of music. Certainly music scores are bona fide ICEs - they are at least about the music and process of performing it.
In order to support OBI, we have focused first on those cases where the larger elements refer, in their entirety, to actual things, for the most part, but I, at least, consider the scope of IAO to be broader, and that taking the approach that IAO takes and applying it to the larger scope would be profitable.
On Fri, Dec 4, 2009 at 1:59 PM, Jonathan Rees <j...@creativecommons.org> wrote:Yes, I had a very similar question. I'm interested in curating
misunderstandings, ambiguity, version skew, etc. with web standards
and biodiversity informatics as use cases. Bjoern mentioned a previous
agreement to limit IAO to situations where meaning was
uncontroversial, but Alan has been encouraging me to bring my own
issues to the group, so this dissonance leaves me unsure as to whether
I'm being disruptive, or just joining in.As before, I consider these in scope and of interest. While there may be a need to partition our efforts so that competing interests can be addressed in the time we have, I would consider it a loss for IAO if these were not considered in scope. I think the area that would first need to be addressed are communication processes, since these are the source of misunderstanding, ambiguity,etc. In developing IAO we've first focused on what are the elements that are available for communication, and to the extend that this is a necessary prerequisite for discussing communication, I agree with the sentiment about getting the basics down before the more complicated cases. But I think we are making good progress on getting the basics down and I'm game to engage on the issues you raise.ÂMaybe we need tracks: [OBI] [meaning] [culture] in the subject line,
or something like that... or multiple lists, or multiple ontologies. I
don't know.I think the we shouldn't waste too much time on arguing about scope and instead work towards a framework in which all these elements can live. People ought to be able to exercise judgement as to which threads they participate in, and how they participate so that we can support multiple activities.My main concern is that we have a coherent framework in which all of these fit. We lose big, IMO, if these other interests need to be developed in separate forum and lead to incompatible foundations.One guy's opinion.-AlanÂ
--Â
Here I disagree. Mimicry is not aboutness, any more than similarity is aboutness.Similarly, I think music has elements that refer - sometimes to actual sounds - deliberate mimicry of natural rhythms or the sounds of actual processes, and sometimes to other elements of other pieces of music.
Music scores represent certain patterns, but they are not about those patterns, it seems to me. Again, the relevant intentional context is missing. As contrasted with the case when a portion of a score is cited e.g. in a work of music criticism.Certainly music scores are bona fide ICEs - they are at least about the music and process of performing it.
Later, yes. But let's do the easy cases first.In order to support OBI, we have focused first on those cases where the larger elements refer, in their entirety, to actual things, for the most part, but I, at least, consider the scope of IAO to be broader, and that taking the approach that IAO takes and applying it to the larger scope would be profitable.
> I take a different point of view on this matter, based perhaps on a
> broader reading of what the about relation about means, I expect. To
> be fair I think it is a minority opinion.
For the record, I agree with you.
>
> In the IAO sense, an information entity can be about something by
> mentioning it (mentions is a subproperty of is_about). In that
> sense, a novel can be about trees, even if it is a fictional story
> about wizards. The acts that the are ascribed to the wizards in the
> story have no direct relation to anything that exists, taken as a
> whole, but their parts do - waving of hands, smoke, the thing that
> appears in the smoke.
>
> Similarly, I think music has elements that refer - sometimes to
> actual sounds - deliberate mimicry of natural rhythms or the sounds
> of actual processes, and sometimes to other elements of other pieces
> of music. Certainly music scores are bona fide ICEs - they are at
> least about the music and process of performing it.
A thought here that may be relevant. It seems clear (as anything here
is clear) that a musical score is an ICE and refers to the actual
music, the sound produced in a performance. (I realize there are many
complexities here to be unravelled, but...) It is however notable that
we do not call the score "music", and we do not call its author an
"artist": this term is reserved for the musician who reads the score
and, as it were, un-references it by creating the sound for it to
refer to. A musical "artist" is a musician, not a composer. Similarly,
the "artist" in dance is the dancer, not the choreographer who writes
the labanotation. If we were to transfer the same linguistic treatment
to the plastic arts, then the "artist" for sculpture would seem to be
the foundryman rather than the sculptor who makes the clay original;
for architecture, the building contractor rather than the architect.
Why the extreme disparity, I wonder? It cannot be simply a matter of
skill: some casting or construction operations require extraordinary,
rare, skills. Is it precisely because the connection in the first two
cases between the originator and the final performance goes through an
ICE, via symbols which need to be interpreted by the final 'artist' ?
Whereas in contrast, a foundryman or a building contractor, even
though they may have to perform extraordinary feats of technical skill
to realize the sculptor/architect's vision, never have to *interpret*,
but only reproduce?
Not sure where this goes, but I found it a striking thought.
Pat
------------------------------------------------------------
IHMC (850)434 8903 or (650)494 3973
40 South Alcaniz St. (850)202 4416 office
Pensacola (850)202 4440 fax
FL 32502 (850)291 0667 mobile
phayesAT-SIGNihmc.us http://www.ihmc.us/users/phayes
> At 10:52 AM 12/15/2009, Michel Dumontier wrote:
>
>>>> and if so, would an (imaginary) novel that is based on actual facts
>>>> also qualify as an ICE? or would only certain parts of it qualify
>> as an ICE?
>>
>> I take a different point of view on this matter, based perhaps on a
>> broader reading of what the about relation about means, I expect. To
>> be fair I think it is a minority opinion.
>>
>> In the IAO sense, an information entity can be about something by
>> mentioning it (mentions is a subproperty of is_about). In that
>> sense, a novel can be about trees, even if it is a fictional story
>> about wizards. The acts that the are ascribed to the wizards in the
>> story have no direct relation to anything that exists, taken as a
>> whole, but their parts do - waving of hands, smoke, the thing that
>> appears in the smoke.
>
> I think one can go either way on this. Ingarden would say that the
> trees, and cities (including Paris, Moscow) in the story are as
> fictional (and thus not existent) as are the wizards. Others might
> see War and Peace as being about, say, trees and Moscow.
> In defense of Ingarden one might say that, in the canonical cases of
> novel reading and writing the relevant intentional context is missing.
I have not read Ingarden, but this whole debate seems to me to be off
key. Surely an ICE can be about something nonexistent? We have to
allow this, and not as a mere edge case or exception. Planning systems
routinely consider future scenarios for the express purpose of making
sure that they will not exist, for example. In any case, the actual
existence of a referent is almost irrelevant to the notion of
aboutness in all formal semantic theories. I submit that we will
actually make our work harder, not easier, if we start by ruling out
references to imaginary or otherwise nonexistent entities.
>> Similarly, I think music has elements that refer - sometimes to
>> actual sounds - deliberate mimicry of natural rhythms or the sounds
>> of actual processes, and sometimes to other elements of other
>> pieces of music.
>
> Here I disagree. Mimicry is not aboutness, any more than similarity
> is aboutness.
It is not necessarily, but it can be. Spoken language often uses
mimicry in a referential way, eg by imitating a way of speaking to
refer to the speaker. And in any case, it is not obvious that the
musical cases are most accurately described as mimicry. The 'bird
calls' in Beethoven's pastoral seem more like the musical equivalent
of a rebus in early writing.
>
>> Certainly music scores are bona fide ICEs - they are at least about
>> the music and process of performing it.
>
> Music scores represent certain patterns, but they are not about those
> patterns, it seems to me.
? Surely the score is a description of - and hence (?) is about - the
sound of the music when the score is performed, no? There is some kind
of close relationship between the symbols and the sound, and if it is
other than descriptive, then I am puzzled by what to say about it.
> Again, the relevant intentional context is
> missing.
I do not follow this.
Pat
> --
> informatio...@googlegroups.com
> To change settings, visit
> http://groups.google.com/group/information-ontology
>
------------------------------------------------------------
Hi,
PH> I have not read Ingarden, but this whole debate seems to me to
PH> be off key. Surely an ICE can be about something nonexistent?
It seems to me as if this is not possible according to the axioms in
OWL:
ICE subclassOf about some Entity
I have a related question: how can an ICE be about a class at all? If
I wanted to define a "methodology article" in IAO, I would assume it
to be about a class of investigations, about some (oneOf
{investigation}) or something, i.e., not about
an instance. If I understand BFO's reading of "Entity" correctly, it
has only ontological individuals as instances (no "universals"), so
being about some Entity should mean being about an individual.
So the problem is, I think, not only whether an ICE can be about a
non-existing individual (which seems a strange notion to me, but maybe
I should read more Meinong), but whether it can be about a class (and
then we can discuss if it can be about classes which have no instances
in this world, or whether these classes exist).
Would it be possible to define an investigation-GDC, that is dependent
on the existence of investigations of a specific kind, and say that a
methodology-article is about this newly introduced investigation-GDC?
Then, could we just completely reflect "universals" as GDCs which
depend on instances of the "universal", and treat them as (logical)
individuals?
Rob.
What is a scientific protocol about when it mentions e.g. 'mouse' in a sentence like 'inject mouse with PBS at base of tail'. My reading was that it is about the universal mouse (read 'class' if you don't like 'universal'). Alan's reading was (I hope I reproduce it accurately) that it is about existing instances of mice the author of the protocol has seen (or knowledge of?). This difference in opinion was based on different understandings of the next point that Robert raised, namely if a universal is considered an entity in BFO or not. I thought yes, Alan originally thought no (but I think may have latter revised this).
This should be resolvable, and needs to be documented.
- Bjoern
----- "Barry Smith" <phis...@buffalo.edu> wrote:
--
Hi,
BP> author of the protocol has seen (or knowledge of?). This
BP> difference in opinion was based on different understandings of
BP> the next point that Robert raised, namely if a universal is
BP> considered an entity in BFO or not. I thought yes, Alan
BP> originally thought no (but I think may have latter revised
BP> this).
At least in OBO Relationship Ontology, Universal vs. Particular seems
to be the top-level distinction, and at least in the RO, both
universals and particulars are used as arguments for relations (not in
the OWL implementation, though). So I think this should be made
explicit /somewhere/ (maybe not in BFO, but in something closely
related to BFO, such as the RO).
But even if no "Universal" class should be introduced in BFO, would it
be possible to provide reflection classes under GDC, GDCs which are
defined as things concretized in qualities of instances of U for every
universal/class U? It could be these things that GDCs about classes
are about, if they are not supposed to be about universals themselves.
This, I think, is only an option if we worry about breaking the
OWL-class/BFO-universal identity (or correlation) that seems to be the
basis for every implemented BFO-ontology, as we would get decidability
problems then.
Rob.
>>>>>> "PH" == Pat Hayes <pha...@ihmc.us> writes:
>
> Hi,
>
> PH> I have not read Ingarden, but this whole debate seems to me to
> PH> be off key. Surely an ICE can be about something nonexistent?
>
> It seems to me as if this is not possible according to the axioms in
> OWL:
> ICE subclassOf about some Entity
We would have to allow nonexistent entities in the universe of
discourse, of course. With the consequence that the logical
existential quantifier no longer means 'exists in the actual world',
but more like 'could possibly exist' or maybe 'can be mentioned as a
possibility'. I have no problem with this, in spite of Kant's
objections to treating existence as a predicate; and OWL takes no
stance on the matter. People do it all the time, and the logic works
in exactly the same way. It is more or less required when one takes a
modal logic with possibilia and renders the result back to a single FO
theory, using the Kripke possible-worlds modal semantics (which is
pretty much universally accepted). When discussing physical reality it
would be overkill, but when discussing the denotations of signs it is
almost impossible to proceed without accepting this.
> I have a related question: how can an ICE be about a class at all? If
> I wanted to define a "methodology article" in IAO, I would assume it
> to be about a class of investigations, about some (oneOf
> {investigation}) or something, i.e., not about
> an instance.
I wonder. A similar issue came up early in the OWL deployment, in how
to talk about species, as in "The elephant is the largest land
mammal." which seems to be about elephants in general, in spite of its
definite singular subject. One idea is that this is about 'the
typical elephant', a mythical creature which has the ontic status of
an individual, and possesses all the typical properties of elephants
in general, and may not be in the class Elephant. Another idea is
that, as you suggest, it is about the class of elephants. In spite of
the obvious problems, the first seems more realistic.
> If I understand BFO's reading of "Entity" correctly, it
> has only ontological individuals as instances (no "universals"), so
> being about some Entity should mean being about an individual.
>
> So the problem is, I think, not only whether an ICE can be about a
> non-existing individual (which seems a strange notion to me, but maybe
> I should read more Meinong),
No, just read a basic textbook about planning methods, and consider
the use of an ontology in such a planner.
It is rarely a good idea, when thinking about ontology engineering, to
start by reading philosophy. As a general rule (there are of course
individual exceptions) philosophers know squat about ontology.
> but whether it can be about a class (and
> then we can discuss if it can be about classes which have no instances
> in this world, or whether these classes exist).
>
> Would it be possible to define an investigation-GDC, that is dependent
> on the existence of investigations of a specific kind, and say that a
> methodology-article is about this newly introduced investigation-GDC?
Sounds like the typical elephant.
>
> Then, could we just completely reflect "universals" as GDCs which
> depend on instances of the "universal", and treat them as (logical)
> individuals?
If y'all were to adopt the ISO Common Logic version of FOL, all
universals would already be individuals anyway, since everything in CL
is an individual. It makes everything a lot easier if one simply
adopts this as a logical principle. It amounts to saying that anything
(that can be named in any way at all) can stand in relationships to
other entities. Which seems to me to be correct: after all, if I can
refer to it, what can possibly prevent me assigning it properties, or
defining relations for it to stand in?
Pat Hayes
>
> Rob.
> No, just read a basic textbook about planning methods, and consider
> the use of an ontology in such a planner.
Pat, my experience in this community so far is that the ontologists really don't think about the uses their ontologies will be put to, particularly in representing and reasoning about knowledge. Those of us who want to use these ontologies to represent knowledge are, of course, free to (ab)use them in any way we see fit. So, for example, in a paper that is currently under review (although I know Barry has read), we talked about the challenges of using ontology to semantically annotate biomedical journal articles. One example: although the ontological classes and relations are defined as canonical (e.g. every cell-membrane part-of some cell), many mentions in the literature are not (e.g. cell-membrane that is no long part of some cell). Or another that you will like: it is often hard to tell whether a text mention is a continuent (the function of being able to phosphorylate) or an occurent (a particular phosphorylating) -- it seems likely in many of these cases the author doesn't care about the distinction. Nevertheless, the ontologies produced by the BFO process have tended to be far better for our (perhaps viewed as nefarious) purposes than idiosyncratic ones we or others havet come up with.
There is, of course, nothing preventing us from taking the ontologies produced by this community, and using them to plan to maintain the non-existance states or entities we would rather avoid, or to assert that this particular cell membrane is not part of a cell. So far, I have not found the rather cramped interpretations that the BFO crowd imposes on themselves to be problematic for my uses. I will keep you posted as our attempts to make more complex and dicey inferences progresses.
Larry
Hi,
PH> I have not read Ingarden, but this whole debate seems to me to
PH> be off key. Surely an ICE can be about something nonexistent?
>> It seems to me as if this is not possible according to the
>> axioms in OWL: ICE subclassOf about some Entity
PH> We would have to allow nonexistent entities in the universe of
PH> discourse, of course. With the consequence that the logical
PH> existential quantifier no longer means 'exists in the actual
PH> world', but more like 'could possibly exist' or maybe 'can be
PH> mentioned as a possibility'. I have no problem with this, in
PH> spite of Kant's objections to treating existence as a
PH> predicate; and OWL takes no stance on the matter. People do it
I have yet to see a convincing existence-predicate. The theories I
have seen which admit non-existing objects in the universe use
self-identity as criterion for existence, i.e., exists(x) <-> x=x,
which does not convince me.
PH> all the time, and the logic works in exactly the same way. It
PH> is more or less required when one takes a modal logic with
PH> possibilia and renders the result back to a single FO theory,
PH> using the Kripke possible-worlds modal semantics (which is
PH> pretty much universally accepted). When discussing physical
I have no objection to possible worlds, either, and possible existence
as existence in some possible world. But how do you admit non-existing
objects in one FOL theory? It would be something like:
exists x (not ex(x))
with ex(x) being some kind of existence predicate. I would understand
a world-indexed existence predicate, i.e., ex(c,w_1), not ex(c,w_2),
and so on, but in a single FOL theory?
PH> reality it would be overkill, but when discussing the
PH> denotations of signs it is almost impossible to proceed
PH> without accepting this.
I am not sure I agree (but I am not sure I understand, either). A sign
for the unicorn in my kitchen and another sign for the unicorn in
yours would have the same referent (nothing, I presume), and be
intensionally different: there are possible worlds where they refer,
and refer to different entities. Possible worlds I accept, but
non-existing objects within one world seem still strange to me.
>> would assume it to be about a class of investigations, about
>> some (oneOf {investigation}) or something, i.e., not about an
>> instance.
PH> I wonder. A similar issue came up early in the OWL deployment,
PH> in how to talk about species, as in "The elephant is the
PH> largest land mammal." which seems to be about elephants in
PH> general, in spite of its definite singular subject. One idea
PH> is that this is about 'the typical elephant', a mythical
PH> creature which has the ontic status of an individual, and
PH> possesses all the typical properties of elephants in general,
PH> and may not be in the class Elephant. Another idea is that,
PH> as you suggest, it is about the class of elephants. In spite
PH> of the obvious problems, the first seems more realistic.
Why "realistic"? The property of the class itself does say something
about all the instances of this class, which seems intuitive to me.
And then, what are instances of "Species"? Idealized objects? It seems
more natural to me to say that Elephant(Dumbo) and Species(Elephant),
and this does not introduce unnecessary (?) additional entities.
>> So the problem is, I think, not only whether an ICE can be
>> about a non-existing individual (which seems a strange notion
>> to me, but maybe I should read more Meinong),
PH> No, just read a basic textbook about planning methods, and
PH> consider the use of an ontology in such a planner.
PH> It is rarely a good idea, when thinking about ontology
PH> engineering, to start by reading philosophy. As a general rule
PH> (there are of course individual exceptions) philosophers know
PH> squat about ontology.
>> Would it be possible to define an investigation-GDC, that is
>> dependent on the existence of investigations of a specific
>> kind, and say that a methodology-article is about this newly
>> introduced investigation-GDC?
PH> Sounds like the typical elephant.
>> Then, could we just completely reflect "universals" as GDCs
>> which depend on instances of the "universal", and treat them as
>> (logical) individuals?
PH> If y'all were to adopt the ISO Common Logic version of FOL,
PH> all universals would already be individuals anyway, since
PH> everything in CL is an individual. It makes everything a lot
PH> easier if one simply adopts this as a logical principle. It
PH> amounts to saying that anything (that can be named in any way
PH> at all) can stand in relationships to other entities. Which
PH> seems to me to be correct: after all, if I can refer to it,
PH> what can possibly prevent me assigning it properties, or
PH> defining relations for it to stand in?
Ok, the CL standard is certainly on my reading-list for Christmas.
I only fear that this property of CL will be something like the OWL2
metamodeling/punning feature, which has no impact on semantics at
all (in CL, when I quantify over a predicate symbol as in
\forall R \forall x ( R(x) )
does the quantifier over R range over the powerset of the universe?)
Ok, maybe off topic, and I need to read the CL standard before making
an informed comment.
Yet how would we implement this in OWL, not in CL?
Rob.
>
> On Dec 17, 2009, at 4:03 PM, Pat Hayes wrote:
>
>> No, just read a basic textbook about planning methods, and consider
>> the use of an ontology in such a planner.
>
> Pat, my experience in this community so far is that the ontologists
> really don't think about the uses their ontologies will be put to,
> particularly in representing and reasoning about knowledge. Those
> of us who want to use these ontologies to represent knowledge are,
> of course, free to (ab)use them in any way we see fit. So, for
> example, in a paper that is currently under review (although I know
> Barry has read), we talked about the challenges of using ontology to
> semantically annotate biomedical journal articles. One example:
> although the ontological classes and relations are defined as
> canonical (e.g. every cell-membrane part-of some cell), many
> mentions in the literature are not (e.g. cell-membrane that is no
> long part of some cell). Or another that you will like: it is often
> hard to tell whether a text mention is a continuent (the function of
> being able to phosphorylate) or an occurent (a particular
> phosphorylating) -- it seems likely in many of these cases the
> author doesn't care about the distinction.
I should hope not, as this particular distinction is completely
artificial and incomprehensible outside of a narrow philosophical
perspective. It is also not recognized by several standard ontologies
in widespread use, which are constructed in such a way that to
introduce 'continuants' into them would result in immediate
contradictions.
> Nevertheless, the ontologies produced by the BFO process have tended
> to be far better for our (perhaps viewed as nefarious) purposes than
> idiosyncratic ones we or others havet come up with.
I keep reading assertions like this. I wonder how you make these
judgements of the utility or effectiveness of BFO (or any other
ontology) when by your own admission, if I follow you, you have
neither used them to make inferences or found them terribly helpful in
formalizing knowledge. What exactly *are* your purposes, nefarious or
otherwise?
Pat
>
> There is, of course, nothing preventing us from taking the
> ontologies produced by this community, and using them to plan to
> maintain the non-existance states or entities we would rather avoid,
> or to assert that this particular cell membrane is not part of a
> cell. So far, I have not found the rather cramped interpretations
> that the BFO crowd imposes on themselves to be problematic for my
> uses. I will keep you posted as our attempts to make more complex
> and dicey inferences progresses.
>
> Larry
>
>>>>>> "PH" == Pat Hayes <pha...@ihmc.us> writes:
>
> Hi,
>
> PH> I have not read Ingarden, but this whole debate seems to me to
> PH> be off key. Surely an ICE can be about something nonexistent?
>>> It seems to me as if this is not possible according to the
>>> axioms in OWL: ICE subclassOf about some Entity
> PH> We would have to allow nonexistent entities in the universe of
> PH> discourse, of course. With the consequence that the logical
> PH> existential quantifier no longer means 'exists in the actual
> PH> world', but more like 'could possibly exist' or maybe 'can be
> PH> mentioned as a possibility'. I have no problem with this, in
> PH> spite of Kant's objections to treating existence as a
> PH> predicate; and OWL takes no stance on the matter. People do it
>
> I have yet to see a convincing existence-predicate. The theories I
> have seen which admit non-existing objects in the universe use
> self-identity as criterion for existence, i.e., exists(x) <-> x=x,
> which does not convince me.
Nor me either. I don't suggest having a 'criterion' for existence, or
even a theory of existence. Rather, I would simply assert real
existence, using a predicate such as Actual. Hence for example,
(Actual Eiffel_Tower), (not (Actual Sherlock_Holmes)). But most of the
time, when stating generalities and 'laws' involving what Barry calls
universals, there is no need to even introduce the distinction: the
laws apply to all things, real or unreal, so one simply ignores the
issue.
BTW, in other work, we developed a technique for constructing terms to
denote 'imaginary' entities such as 'the person who Louis believes is
called "Superman"', which provides a much more robust (though
admittedly rebarbative) technique for describing what are usually
called 'opaque' naming situations. There are some slides on the
general topic, with examples, at
http://www.slideshare.net/PatHayes/ikl-presentation-for-ontolog
(slides 15-20) As you will see, these only work if one allows terms to
refer to entities which do not exist (but would, if people's false
beliefs were correct.)
>
> PH> all the time, and the logic works in exactly the same way. It
> PH> is more or less required when one takes a modal logic with
> PH> possibilia and renders the result back to a single FO theory,
> PH> using the Kripke possible-worlds modal semantics (which is
> PH> pretty much universally accepted). When discussing physical
>
> I have no objection to possible worlds, either, and possible existence
> as existence in some possible world. But how do you admit non-existing
> objects in one FOL theory? It would be something like:
> exists x (not ex(x))
> with ex(x) being some kind of existence predicate. I would understand
> a world-indexed existence predicate, i.e., ex(c,w_1), not ex(c,w_2),
> and so on, but in a single FOL theory?
I gather you have been trained in the School of Quine. I learned logic
there myself. But here is the cure. If you are happy with possible
worlds containing possibilia, then just imagine this entire Kripke
structure described in FOL, in what is surely now the standard way, so
that possible worlds are actual entities, and the truth of (R a) in
world w is written (R a w), and the modalities amount to
quantification over worlds, so that NEC (R a) is (forall (w)(if (alt
w0 w)(R a w))) and so on. Since the possible worlds may differ in
their local universes, existence in the original modal theory has to
be limited to the things in the the origin possible world w0, ie the
'real' world, so that (exist (x)(R x)) becomes in the Kripkje-ized
theory, (exist (y)(and (Member y w0) (R y w0))). In other words, one
simply takes the Kripke structure to be a first-order structure with
an appropriately extended vocabulary. Now, the universe of this FO
theory is the set of all possibilia in all possible worlds of the
original theory; and existence in the original theory is membership in
the local universe of the origin world. It is a straightforward
mechanizable translation.
>
> PH> reality it would be overkill, but when discussing the
> PH> denotations of signs it is almost impossible to proceed
> PH> without accepting this.
>
> I am not sure I agree (but I am not sure I understand, either). A sign
> for the unicorn in my kitchen and another sign for the unicorn in
> yours would have the same referent (nothing, I presume),
No, its much more likely that they would refer to different unicorns,
unless of course they were reproductions of the same unicorn picture,
say this one: http://www.chainleader.com/articles/blog/180000418/20090507/unicorn.jpg
, in which case it would seem more reasonable to say that they were
both images of the same unicorn.
> and be
> intensionally different: there are possible worlds where they refer,
> and refer to different entities. Possible worlds I accept, but
> non-existing objects within one world seem still strange to me.
But the difference is only one of scope. If we are willing to talk of
non-actual possible *worlds*, surely there can be less mental trauma
involved in speaking of mere possible *individuals*. After all, what
is a world other than a large collection of individuals related
together in a sufficiently complex way?
>
>>> would assume it to be about a class of investigations, about
>>> some (oneOf {investigation}) or something, i.e., not about an
>>> instance.
> PH> I wonder. A similar issue came up early in the OWL deployment,
> PH> in how to talk about species, as in "The elephant is the
> PH> largest land mammal." which seems to be about elephants in
> PH> general, in spite of its definite singular subject. One idea
> PH> is that this is about 'the typical elephant', a mythical
> PH> creature which has the ontic status of an individual, and
> PH> possesses all the typical properties of elephants in general,
> PH> and may not be in the class Elephant. Another idea is that,
> PH> as you suggest, it is about the class of elephants. In spite
> PH> of the obvious problems, the first seems more realistic.
>
> Why "realistic"? The property of the class itself does say something
> about all the instances of this class, which seems intuitive to me.
To me that is exactly what it fails to do. Properties of a class are
not usually properties of its instances. A class may for example have
a cardinality, but none of its instances will inherit this class
property. Or it may be a subclass of another class: again, its
instances will not be subclasses.
> And then, what are instances of "Species"? Idealized objects? It seems
> more natural to me to say that Elephant(Dumbo) and Species(Elephant),
> and this does not introduce unnecessary (?) additional entities.
Well, indeed, that is a commonly accepted convention, and it does have
the advantage you cite. But it seems flawed to me, for the reasons
just given. At the very least, one would need to carefully distinguish
properties of such a class that were simply class properties from
those that were supposed to be somehow be, or imply, properties of its
members or instances.
The ISO standard document is not very readable, I am afraid. You might
find the IKL guide
http://www.ihmc.us/users/phayes/ikl/guide/guide.html
easier to digest as an introduction. (IKL is common logic with a
couple of extensions, notably a convention for naming propositions
expressed by sentences. But if you ignore this, the rest of the
document applies to Common Logic almost without change.) There are
also some slides on the topic on slideshare
http://www.slideshare.net/PatHayes/translating-into-common-
logic-459009 (A 'tutorial' lecture given to SemWebTech 2008)
and even a recording of a longish presentation, with accompanying
slides, downloadable from
http://ontolog.cim3.net/cgi-bin/wiki.pl?ConferenceCall_2006_10_26
(scroll down to find it, quite a long way down.)
>
> I only fear that this property of CL will be something like the OWL2
> metamodeling/punning feature, which has no impact on semantics at
> all (in CL, when I quantify over a predicate symbol as in
> \forall R \forall x ( R(x) )
> does the quantifier over R range over the powerset of the universe?
No, it does not. If it did, the logic would be higher-order. CL is
strictly first-order. Here is a test case:
given
1. (and (P a)(Q b))
does it follow that
2. (exists (r)(and (r a)(r b)))
? that is, do a and b have a property in common, if they have
properties at all?
In HOL, this does follow. It also follows in the weakened version of
HOL described by Henkin. (It does in any logic which supports lambda-
abstraction.) It does not follow in CL. However, one can add axioms
which assert that the required properties exist, for example in this
case:
3. (forall (p q x)(iff ((OR p q) x)(or (p x)(q x)) ))
introduces a disjunction function on predicates, and then 1. and 3.
together entail 2.:
(and (P a)(Q b))
(P a) and-elim
(or (P a)(Q a)) or-intro
((OR P Q) a) from 3
(Q b) from 1, and-elim
(or (Q a)(Q b)) or-intro
((OR P Q) b) from 3
(and ((OR P Q) a)((OR P Q) b)) and-intro
(exists (r)(and (r a)(r b))) exist-gen
Basically, CL follows the standard first-order notion that things are
required to exist only if you say they do. There are no 'comprehension
principles'.
> )
> Ok, maybe off topic, and I need to read the CL standard before making
> an informed comment.
> Yet how would we implement this in OWL, not in CL?
OWL 2 is very close to a subset of CL (subset because it is a
restricted logic; very close because the OWL 'punning' does not quite
handle equality correctly.) An OWL 2 version would probably work for
most cases, though one would need to check the details and perhaps
reformulate some of the axioms.
Pat
>>
>> Nevertheless, the ontologies produced by the BFO process have tended
>> to be far better for our (perhaps viewed as nefarious) purposes than
>> idiosyncratic ones we or others havet come up with.
>
> I keep reading assertions like this. I wonder how you make these
> judgements of the utility or effectiveness of BFO (or any other
> ontology) when by your own admission, if I follow you, you have
> neither used them to make inferences or found them terribly helpful in
> formalizing knowledge. What exactly *are* your purposes, nefarious or
> otherwise?
Pat,
Concretely, we use OBO ontologies as the targets for NLP (e.g. http://www.biomedcentral.com/1471-2105/9/78 ) and as the basis for a knowledge-based system for analyzing genome-scale datasets (e.g. http://www.ploscompbiol.org/article/info:doi%2F10.1371%2Fjournal.pcbi.1000215 ). There are quite modest, but non-negligible uses to make inferences (e.g. http://www.ncbi.nlm.nih.gov/pmc/articles/PMC2084396 ) and they are helpful (but not sufficient) for formalizing knowledge.
Our focus is on building knowledge-based systems (for helping people understand complex molecular biology, and generate biomedically significant hypotheses) whose ontological foundations are broadly accepted. The main desiderata is that the terms, definitions and relations expressed in the ontology be both (a) broadly accepted in the biomedical research community as "good", and (b) not create insurmountable obstacles to their use within computational systems. The OBO ontologies (particularly those in the OBO Foundry) meet these criteria. The first is met by the fact that most of the content development was done in the biomedical research community, primarily in the model organism databases, but also including broader committees of experts. That's extremely important to us -- from a high level perspective, by far the most important use of the ontology is to ensure that communications across biomedical disciplines that do not by and large share vocabulary is clearly understood by all. The second desideratum is met by the demands of Barry, et al., who have largely prevented any horrible logical mistakes (cf, say, MeSH). I agree with you that some of the distinctions insisted upon by the philosophers are of little use, but I find it entirely possible to "mend this stuff" for our needs. We are starting down the road of trying to get much more sophisticated inferences done (our collaborators want to use KM to start with, of all things!) and the translation of OBO representations to CLIB isn't looking that hard (although I reserve final judgment on that), so will have more to say about this in a few months.
I think it was Larry Birnbaum who first observed that it is much easier to fix an odd formal framework with a lot of good content than it is to take an empty formal framework and fill it with lots of good content. That's the nature of our support for the utility and effectiveness of OBO (including BFO).
Larry
In all the BFO documentation we have made it clear that universals
are full-fledged entities, and thus that universals exist; thus that
universals can be the targets of the canonical aboutness relation, as
for example in statements of scientific laws.
Each node in BFO, as in ontologies such as the GO or the PRO,
represents exactly one universal, and thereby represents all the
instances of that universal. Nodes are allowed into such ontologies
only if they represent universals. Hence there is no BFO node
labelled 'universal'. For what would be the individual spatiotemporal
entities which could be its instances?
How, now, to address Robert's point as concerns defining the
(canonical) aboutness relation?
I still believe that the most appropriate strategy is to define
'aboutness' in terms of 'entity', meaning by the latter (informally):
anything that exists. 'Entity', like 'existence', would then be primitive.
Instead of attempting to define 'entity' we could provide instead
axioms giving sufficient conditions for existence, for example along
the lines of:
for some b, a part_of b -> a exists
for some b, a is_a b -> a exists
Incidentally, in earlier formulations of BFO we had two distinct
ontologies, one for occurrents, one for continuants. In more recent
formulations of BFO we combined these together under the entity node.
I am now tempted to propose reverting to the original formulation (in
reflection of the distinct is_a relations operating on the continuant
and occurrent side).
BS