Ed, Pat, Leo, and Chris, (01)
Thanks for the comments. I'd like to make some brief
comments on your comments and then discuss one of the
paragraphs from Bundy's paper. (02)
JS>> A single, unified upper ontology is impossible to
>> achieve, and it's not necessary for interoperability. (03)
EH> Amen to both parts of the final sentence!
>
> ... As stated in a paper I wrote last year, I think there
> are five major methodologies: (04)
http://www.isi.edu/natural-language/people/hovy/papers/05ICCS-ontol-methods-hovy.pdf (05)
I'm glad you support the main points, and I like your five
methodologies plus the sixth one you added. My only quibble
is that I suspect there are a lot more methodologies either
not mentioned or still to be discovered. (06)
PC> I would argue that every system that successfully interoperates
> at any level shares some upper ontology, whether explicitly or
> implicitly. (07)
If you delete the word "upper", I would agree. The point Bundy was
making is that interoperability on shared data or results is always
on a problem or task-oriented basis. That is also the point that
Lenat learned after the 21+ years of working on Cyc: the axioms
at the upper level are not widely used, the middle-level axioms are
more important, and for any particular application, the microtheories
(i.e., the lowest-level axioms) are the most important. (08)
LO> John's and Alan's statements are not without controversy and
> should not be accepted without disputation however we might be
> disposed to accept them because they align with our desires. (09)
Even more important than disputation are facts. (010)
LO> I think we are both arguing that we need to embrace more a
> scientific (and computational) approach to these issues rather
> than a philosophical approach per se (although of course we are
> best informed scientifically by being best informed philosophically
> and not falling into old, well-worn but flawed, ensnaring
> philosophical arguments). (011)
I agree 100%. As Wittgenstein said, philosophy is primarily
a systematic presentation of reminders of what we already know.
Among the facts are the evidence and experience derived from
21+ years of Cyc, the 50 years of AI, and the past several
centuries of reasoning in science and engineering. (012)
For the latter, I'd like to emphasize the following paragraph,
which I quoted in my first excerpt from Bundy's paper: (013)
As another example, consider the experiment conducted
by Andreas diSessa on first-year MIT physics students.
The students were asked to imagine a situation in which
a ball is dropped from a height onto the floor. Initially,
the ball has potential but not kinetic energy. Just before
it hits the floor it has kinetic but not potential energy.
As it hits the floor it has neither. Where did the energy go?
The students had trouble answering this question because they
had idealised the ball as a particle with mass but no extent.
To solve the problem they had to refine their representation
to give the ball extent, so that the `missing' energy could be
stored in the deformation of the ball. Note that this requires
a change in the representation of the ball, not just a change
of belief about it. (014)
This is not a philosophical argument. It is a concrete example
of a typical physics problem, and it illustrates issues that
scientists and engineers face in every experiment or application: (015)
1. The formulas and equations for the "upper level" ontology of
physics -- i.e., the most general formulations of relativity
and quantum electrodynamics -- are almost impossible to solve
for any practical problem, and every application has to make
some approximation based on simplifying assumptions. (016)
2. Furthermore, the approximations made for one problem involve
assumptions that are usually inconsistent with the assumptions
required for some other problem. For example, the ball as a
rigid body while it is falling, but as a deformable body at
the instant of impact. (017)
3. When any problem becomes complex, it nearly always breaks down
into multiple subproblems that require different assumptions.
The case of the bouncing ball with just two subproblems is
extremely simple compared to the enormous number of conflicting
subproblems involved in designing a bridge or an airplane. (018)
4. The view of science as a unified body of theory is totally
unrealistic. Even physics, the hardest of the hard sciences,
is a hodge-podge of thousands of mutually inconsistent models
for each area of application. As Bundy said, there is no such
thing as a unified model of everything. (019)
5. If physics cannot give us a unified ontology, then how can we
expect any kind of unity at all when it comes to the squishier
aspects of life: economics, medicine, business, social sciences,
etc. In every one of these areas, the best we have are problem
or task-oriented theories, each of which is based on simplifying
assumptions that are inconsistent with those of any other areas. (020)
LO> ... stepwise testing of hypotheses against actual reality in
> the determination and refinement of theory. That's my own form of
> scientific pragmatism, and one I no doubt imperfectly follow. (021)
I think we all pay lip service to that version of scientific
pragmatism. The imperfections arise from picking and choosing
various facts to emphasize as important or to abstract away. (022)
CP> ... a counter-example to the claim that having a common upper
> ontology across a federation of inter-operating systems is
> "impossible". (023)
To quote another philosopher, I believe that Peirce shows how global
commonality can be achieved: (024)
It is easy to speak with precision upon a general theme. Only,
one must commonly surrender all ambition to be certain. It is
equally easy to be certain. One has only to be sufficiently vague.
It is not so difficult to be pretty precise and fairly certain at
once about a very narrow subject. (CP 4.237) (025)
This quotation summarizes the futility of any attempt to develop a
precisely defined ontology of everything, but it offers two useful
alternatives: an informal classification, such as a thesaurus or
terminology designed for human readers; and an open-ended collection
of formal theories about narrowly delimited subjects. (026)
In other words, it is possible to have an upper ontology, but only
if it is "sufficiently vague" -- i.e., denuded of any detailed axioms. (027)
In fact, I would claim that the upper levels should not make *any*
empirical claims -- the only axioms at the upper levels should be
true by the definition of the terms involved. (028)
John Sowa (029)
PS: In my previous two notes, I quoted the introduction and the
conclusion to Bundy's paper. The material in the middle part of
the paper is covered (with more detail) on the following web site: (030)
http://dream.inf.ed.ac.uk/projects/dor/
Dynamic Ontology Refinement (031)
_________________________________________________________________
Message Archives: http://colab.cim3.net/forum/ontac-forum/
To Post: mailto:ontac-forum@xxxxxxxxxxxxxx
Subscribe/Unsubscribe/Config:
http://colab.cim3.net/mailman/listinfo/ontac-forum/
Shared Files: http://colab.cim3.net/file/work/SICoP/ontac/
Community Wiki:
http://colab.cim3.net/cgi-bin/wiki.pl?SICoP/OntologyTaxonomyCoordinatingWG (032)
|