Newsgroups: comp.ai.philosophy
Path: utzoo!utgpu!news-server.csri.toronto.edu!rpi!think.com!zaphod.mps.ohio-state.edu!magnus.acs.ohio-state.edu!news
From: nehmer@hpuxa.acs.ohio-state.edu (Robert Nehmer)
Subject: Re: Logic & AI
Message-ID: <1991May18.162731.629@magnus.acs.ohio-state.edu>
Sender: news@magnus.acs.ohio-state.edu
Nntp-Posting-Host: hpuxa.acs.ohio-state.edu
Organization: The Ohio State University
References: <122653@tut.cis.ohio-state.edu>
Date: Sat, 18 May 1991 16:27:31 GMT
Lines: 46

In article <122653@tut.cis.ohio-state.edu>
 chandra@cis.ohio-state.edu (B Chandrasekaran) writes:
>[I recently posted this to comp.ai, but I got messages
 from several people
>that comp.ai.philosophy is a more appropriate newsgroup
 for this. So here
>it goes.  Replies to me or to the bboard as you see appropriate.]

[Deletions of thoughtful material]

>
>Where I have been
>impressed by the use of logic is when it is used for precision in the
>definition of tasks and knowledge.  Regarding logic for validation, I
>have already expressed my betting that once we move away from
>deduction, and go on to nonmonotonic and default inferences, we are no
>longer in the realm of a few universal justification rules, but rather
>a large variety of highly contextualized patterns, indexed by the type
>of goals and domains and so on.  There are still useful general
>patterns to be sure, but in application they take on numerous mutant
>forms.  A study of nonmonotonic reasoning then is then ipso facto a
>study of the multiplicity of tasks and methods and goals.  The use of
>these rules again are as validators rather than as generators.
>
>--
>B. Chandrasekaran
>Department of Computer & Information Science
>The Ohio State University
>Columbus, OH 43210  Phone: 614-292-0923

I think that one of the possible ways to conceive of the practical
difference between situations where first order logics prove valuable
as inference engines and those where nonmonotonic systems seem more
appropriate is in the construction of the conditional clauses. What
I mean is that if we have a rule A --> B yet the "real world"
situation seems to behave as though this were only the case 90% of
the time, it may be that our interpretation of A and/or B is too
broad in the first order system and the rule needs to be decomposed
into a finer set, perhaps A' --> B' and B'--> B''. I'm not saying
that this would always work but have noticed a tendency to use some
rather broad atomic classifications in many AI systems. At any rate,
it seems an interesting notion and something that I haven't seen
the literature address.

Rob Nehmer