Sunday, November 27, 2005
Center of Excellence Proposal
à
ONTAC stands for Ontology and Taxonomy Coordinating Working Group
It is a working group of
Semantic Interoperability Community of Practice (SICoP)
To John (Sowa)
I understand your point about use.
<quote from John>
Paul,
There are many versions of
rough sets, fuzzy sets, mereology, etc.
And there are many different and incompatible (at least for infinite sets)
axiomatizations of set theory. [1]
Take you pick. You can use any of them for any purpose
whatever. That's the beauty of
mathematics; it allows anybody to choose any set of axioms to define anything
they like.
You can also define any
version of fuzzy logic, quantum logic, nonmonotonic logic, etc.
But, and this is a very big
*BUT*, every version of mathematics since Euclid to the present, uses classical
FOL (usually stated in a highly stylized version of Greek, English, or other
natural language) as its metalanguage for stating the axioms and the proofs.
The point is that
*everybody* who has ever defined any version logic -- quantum, fuzzy, rough,
informal, formal, non monotonic, or whatever -- has always used good old fashioned
classical first-order logic, as embodied in every natural language as the
metalanguage for stating the definitions and the proofs.
They may use the syntax of
a natural language, but the semantics of every statement they make is
translatable to FOL (except when they make mistakes, as all good mathematicians
do from time to time).
John
<end quote>
But there is the position that inference creates its own appearance of truth, when in fact that measurement of facts has been (may have been) made erroneously. In many cases, the problem might be in the formal deduction itself, even if measurement of facts were perfect. (A prediction of the decision of an individual buyer. The actions of a population of people.)
Uncertainty over facts, and incomplete information are common, at least in many anticipated applications of semantic web (web services).
So there is an enhanced need to have a human who is knowledgeable in the loop..
Does the human "reason" best when the human employs first order predict logic? In many cases, yes. In all cases? The answer is no. You may disagree, but that disagreement does not make me an ignorant person who just does not know better.
Does human behavior follow first order logics. Not often. There are many things to say here about the phenomenon of category formation in natural language, and the importance of well formed and predicable categories to any system that wishes to control economic processes, or governmental processes. I will with-hold.
The loop being, in this case, an action-perception cycle driven ontology mediated sharing of reified community knowledge.
The promise of having knowledge bases that act as superior sources of knowledge is encumbered by the differences between something like geometry and normal human reasoning/behavior.
In geometry (if one avoids talking about axiomatic forcing) the set of theorems are "true" and "relevant" because of the consistency and completeness of the axiomatization. Godel's work is considered irrelevant. In human reasoning/behavior there is only a shadow of geometric forms.
I am saddened that the opposition to my presence is so strong that a campaign to discredit me was launched. You should know better than to think that my questioning of deductive inference was based on ignorance. In any case, my main argument was that Edelman's notion of response degeneracy presents a challenge to deductive logics IF the logics are not able to change underlying ontological commitments such as the meaning of specific data before and after the emergence of an event.
I know that you have been reading some of my work, in particular:
http://www.ontologystream.com/area2/KSF/KnowledgeScience.htm
I wish that you would represent me not as someone who does not know first order predicate logic but as someone who has worked for years to understand how to do other things other than (in addition to) deduction using predicate logic - in a way that WHEN IT IS PROPER TO DO SO one is able to produce if-then and case statements as part of categorization. Then the "program" ; "if categorization x, then take (computational) action of y" can occur and be used by informed professional in biomedical work, B-2-B transaction modeling, etc.
But the rub here is the non-performance of the semantic web technologies (up to this point). You and I agree that the Tim Berners-Lee layer cake is deeply and fundamentally flawed. I know my argument as to why this flaw is disabilitating. I am not sure what your argument is. I just know that you have expressed dis-satisfaction with it.
[1] Let me point out however, that the concept of an infinite set is one constructed (in a beautiful way). But there is a position that these formal constructions have specific problems when used to model the emergence phenomenon in natural world.