From newshub.ccs.yorku.ca!ists!helios.physics.utoronto.ca!news-server.csri.toronto.edu!rutgers!cs.utexas.edu!uunet!mcsun!cernvax!chx400!cnedcu51.bitnet!inmuller Wed Feb  5 11:56:32 EST 1992
Article 3432 of comp.ai.philosophy:
Xref: newshub.ccs.yorku.ca comp.ai.philosophy:3432 comp.ai:2033
Path: newshub.ccs.yorku.ca!ists!helios.physics.utoronto.ca!news-server.csri.toronto.edu!rutgers!cs.utexas.edu!uunet!mcsun!cernvax!chx400!cnedcu51.bitnet!inmuller
>From: inmuller@cnedcu51.bitnet
Newsgroups: comp.ai.philosophy,comp.ai,ch.si.sgaico
Subject: CFP: Epistemology and Artificial Intelligence
Message-ID: <1992Feb3.115431.414@cnedcu51.bitnet>
Date: 3 Feb 92 11:54:31 GMT
Organization: University of Neuchatel, Switzerland
Lines: 102


           Colloqium: Artificial Intelligence and Epistemology
	   ---------------------------------------------------

    Centre interregional de perfectionnement, Tramelan - Switzerland
         wednesday september 30 to friday october 2, 1992

                             CALL FOR PAPERS


Colloqium Theme
---------------

Artificial Intelligence has witnessed a great number of schemes for 
representing knowledge and drawing legitimate inferences therefrom. 
Principles currently in use range from the classical, first order predicate 
calculus to procedural systems with, in between, such alternative systems of 
inference as multi-valued logic, nonmonotonic logic, fuzzy-set theory, and 
uncertain logic. In addition, a wide variety of inference systems have been 
developed for inductive, abductive, analogical, and revisable reasoning of 
many sorts.

This great diversity of schemes for knowledge representation and inference 
raises a number of issues that need clarification from the points of view of 
both Artificial Intelligence and Epistemology. Such is the purpose of the 
present conference. Among other questions to be addressed are the following.

- To what extent can knowledge be abstracted from the use to which it will 
be put? The representations fullfilling the purpose of simultaneously 
modelling a domain (referencial aspect), the perception we have of it 
(epistemic adequacy) and its evolution (genetic dimension) and solving a 
problem (orientedness) as best as possible (heuristic adequacy), this 
question has several aspects:
- the referential aspect: To which extent is each of these representations 
more adequate to model a static or dynamic domain? What sort of 
knowledge representation is best suited for modeling dynamic 
situations? Is contemporary temporal logic suited to this task?
- the epistemic aspect: What is the logic of uncertain and 
nondeterministic events, and how should (partial) knowledge of such 
events be represented? Are uncertainty and nondeterminacy even the 
same concept, and can they be treated identically?
- the genetic aspect: How should knowledge be structured so as to best 
facilitate learning from experience? What are the scope and limits of 
contemporary schemes for updating a knowledge base in the light of 
new information?
- the finalized aspect: Is it necessary to adapt knowledge representation 
to the specific kind of problem to be solved, or can knowledge be 
structured in a universal way, adequate for problems of any character? 
Are several different and complementary approaches requested to 
resolve some problems? In which case, how to integrate these 
approaches?
- the heuristic aspect: To what extent can the use of knowledge be part 
of the knowledge itself (e.g. the logical methods isolate knowledge and 
control, the procedural approach does not make the distinction)?
The present colloquium will allow practitioners in Artificial Intelligence to 
reflect about the assumptions of their models, and it will allow 
epistemologists to discover practical consequences of their theories. The 
discussion will revolve around the concrete problems faced by computer 
scientists attempting to implement a knowledge-based system capable of 
useful inference.

Format and deadline
-------------------

Authors are requested to submit a document at one of the addresses 
mentioned below comprised of (a) a title page with author names and 
addresses and (b) a paper that indicates some connection to the foregoing 
points. Length should not exceed 2000 words. The deadline is April 1, 1992. 
Notification of acceptance will be sent on May 15, 1992. Full versions of 
papers for distribution at the colloquium are due on July 31, 1992. A final 
version incorporating comments and criticisms encountered at the colloquium 
will be due subsequently, in view of publication.

Language
--------

The papers can be written in either french or english.

Program Committee
Francois Bonsack		(Institut de la Methode - CH)
Jacques Jacot			(Automelec SA - CH)
Jean-Pierre Muller		(Uni. Neuchatel - CH)
Daniel Osherson			(IDIAP - CH)
Nicolas Peguiron		(Institut de la Methode - CH)
Jean-Claude Prelaz		(Automelec SA - CH)
Werner Soerensen		(Uni. Neuchatel - CH)

Depending on the number of contributions, the program committee will be 
expanded. In consequence, the previous list is not definitive.

For information
---------------

Jean-Pierre Muller	or	Nicolas Peguiron
Universite de Neuchatel
Institut d'Informatique		Institut de la methode
rue de Monruz, 36		P.O. Box 1081
CH-2000 Neuch	tel		CH-2501 Bienne
Tel.: +41 (0)38 21 30 36	Tel: +41 (0)32 23 83 20
Fax.: +41 (0)38 24 26 95	Fax : +41 (0)32 23 46 57
Internet: Jean-Pierre.Muller@info.unine.ch
X400: C=ch;ADMD=arcom;PRMD=switch;O=unine;OU=info;S=Muller;G=Jean-Pierre


