Re: SEMIS Bulletin
Enrico,
I very much want to change the status quo. But I pointed out
certain "facts of life", which must be addressed. I am also
concerned with what you call "the current limitations that
led to the semantic web challenge." I believe it's a very
important challenge, and I am disappointed that the W3C has
done too little to meet that challenge.
And thanks for citing that debate on issues of ontology.
It's an excellent summary of a variety of topics that we have
been discussing. When you first mentioned it, I clicked on
the URL and got a "404 not found" message. But this time, I
discovered that the final part of the URL, "-04.pdf" had been
detached. For anyone who may be interested, I highly recommend
the following collection of position papers:
http://www.inf.unibz.it/%7Efranconi/papers/ieee-intelligent-systems-04.pdf
Knowledge Representation with Ontologies: The Present and Future
I'd like to discuss some excerpts from the introduction by Brewster
and O'Hara. The position papers they mention (including Enrico's)
can be found in the collection cited above.
> Sociologist and epistemologist Steve Fuller kicks of our debate by
> distinguishing between two views of ontology, which he calls the
> _Newtonian_ and the _Leibnizian_. The former refers to views of
> ontology as finding elegant simplifying principles; the latter hopes
> to do justice to the extreme complexity of experience. Clearly,
> in the context of the Semantic Web and other knowledge management
> contexts, the two approaches offer contrasting advantages. Elegant
> ontologies might be easier to manage, but scruffy ones might be
> easier to apply.
>
> Lining up for the scruffy, Leibnizian team is Yorick Wilks, who
> combines philosophy and linguistics to argue that you can't take
> a lofty unengaged view of what exists. Every ontological theory
> has a viewpoint and is associated with a set of interests, and
> through the terms used -- just like any dictionary or thesaurus
> -- involves covert ontological commitment. Newtonian abstractions
> will always be a chimera.
I've known Yorick Wilks for nearly 30 years, and we agree on some
issues and disagree on others. Although I am more inclined toward
a logic-based foundation, I very strongly agree with Yorick about
the need "to do justice to the extreme complexity of experience."
But I believe that it is a mistake to call Aristotle and Leibniz
scruffy. Aristotle founded formal logic, and Leibniz was the
first to propose a symbolic logic that used mathematics to
represent logical operations (in fact, his invention of the first
calculator that could multiply and divide was motivated, in part,
by his desire to automate his logic). Recognizing the complexity
of experience does not make people scruffy; it just means that
they understand the problem. I'd like to quote the following
remarks by Whitehead (from his book _Modes of Thought_, 1938):
"The conjunction of premises, from which logic proceeds,
presupposes that no difficulty will arise from the conjunction
of the various unexpressed presuppositions involved in those
premises. Both in science and in logic, you have only to develop
your argument sufficiently, and sooner or later you are bound
to arrive at a contradiction, either internally within the
argument, or externally in its reference to fact." (p. 14)
"It should be noticed that logical proof starts from premises,
and that premises are based upon evidence. Thus evidence is
presupposed by logic; at least, it is presupposed by the
assumption that logic has any importance." (p. 67)
"The premises are conceived in the simplicity of their individual
isolation. But there can be no logical test for the possibility
that deductive procedure, leading to the elaboration of
compositions, may introduce into relevance considerations from
which the primitive notions of the topic have been abstracted....
Thus deductive logic has not the coercive supremacy which is
conventionally conceded to it. When applied to concrete instances,
it is a tentative procedure, finally to be judged by the self-
evidence of its issues." (p. 144)
"The topic of every science is an abstraction from the full
concrete happenings of nature. But every abstraction neglects
the influx of the factors omitted into the factors retained."
(p. 196)
My favorite logicians, besides Aristotle and Leibniz, are people
like Charles Sanders Peirce and Alfred North Whitehead, who also
understood the overwhelming complexity of experience. The failure
of 20th-century analytic philosophy was the result of people like
Frege, Russell, and Carnap who tried to hide from the complexity
instead of addressing it head on. See the following paper:
http://www.jfsowa.com/pubs/signproc.htm
Signs, Processes, and Language Games
To continue with the paper by Brewster & O'Hara:
> Computer scientist Enrico Franconi represents the elegant
> Newtonians... "Ontologies with unambiguous meanings" are clear
> Platonic descendants as opposed to Wilks' scruffy relativistic,
> and task-dependent ontologies, which are more Aristotelian in
> their practicality. The question is whether the ontoloy used
> depends on what you want to do with it -- in other words, the
> task for which it is developed. Aristotle's followers answer
> "yes", and the Platonists say "no"...
That passage mischaracterizes Plato, Aristotle, Newton, and Leibniz.
If any of them were around today, they'd say something along the
lines of Karl Marx's famous quotation "Je ne suis pas un marxiste."
> If ontologies are irrevocably task-relative (that is, we
> characterize the world differently depending on what we're
> trying to do), several heterogeneous ontologies should exist.
> But then will the modeling overhead be too high? Mark Musen,
> from his experience in medical informatics, reminds us of when
> the primary focus in this area was not ontologies but problem-
> solving methods. This research taught us much about how
> information gets deployed to achieve tasks, and Musen is keen
> that we not forget this knowledge in the rush to address
> fashionable representational issues. Both knowledge types are
> valuable.
My only disagreement is with the word "several". I believe
that it should be replaced with "infinite". There is no limit
to the number of ontologies that must be supported. But that
approach is not scruffy -- on the contrary, it can be formalized
very nicely and elegantly. See my article cited above on Signs,
Processes, and Knowledge Games.
> Jeremy Ellman, CTO of Wordmap, discusses a fascinating survey
> on how ontologies are actually used; his research shifts the
> focus away from the domain properties, and even the task
> requirements, and toward the necessity of integration into
> existing systems. Given current research assumptions, it is
> significant that less than 10 percent of the ontologies his
> company has dealt with involved inferential requirements.
> Simon Buckingham Shum, from his perspective of research into
> knowledge media, endorses this view, arguing that over-
> engineered systems simply won't be applied...
I sympathize with these ideas, but I also believe that good
inferential systems can support much more than 10% of the systems.
However, I include the kind of reasoning done by SQL database
systems in the logic camp, and I also include analogical reasoning
as an important method of inference. See
http://www.jfsowa.com/pubs/analog.htm
Analogical Reasoning
As for the debate between Yorick and Enrico, I sympathize with
Enrico's desire for logical purity, but I believe that he is
spending too much effort in trying to defend technology that has
been around for 30 years without even begining to scratch the
surface of the issues that Yorick is addressing.
Rather than staying in the tiny playpen of subsets of logic
whose complexity and decidability issues are fully understood,
I believe that it is essential to push the envelope of what can
be done with logic. That's what Aristotle, Leibniz, Peirce,
and Whitehead were trying to do. I believe that it's important
for logicians to study the formal properties of logics, but I
also believe that the real action and the greatest benefit to
be gained from using logic is in addressing a much wider range
of applications and methods of reasoning.
John