AI Seminar
May 27th, 2008, 14.15, glasburen

Prof. Witold Lukaszewicz, University of Warmia and Mazury and College
of Informatics and Economics, Olsztyn, Poland

Title: Extending Knowledge Base Update into First-Order Knowledge Bases.

Abstract: 

Belief update is the task of modifying a knowledge base when an action
is performed. More specifically, given a knowledge base KB,
representing the world under consideration, and a formula alpha,
representing an effect of an action, the task is to specify a new
knowledge base, KB*alpha, representing the world after performing the
action.

A number of different knowledge base update operators have been
proposed in the literature. Ten of them are examined in a paper of
Herzig and Rifi ( Propositional Belief Base Update and Minimal
Change. Artificial Intelligence Journal, 115(1):107-138, 1999). The
conclusion they come to is that only two of the considered operators
are satisfactory. One of them is MPMA, introduced by P. Doherty,
W. Lukaszewicz and E. Madalinska-Bugaj in (The PMA and Relativizing
Minimal Change for Action Update.  Proceedings of the 6th
International Conference on Principles of Knowledge Representation and
Reasoning, KR-98, 158-169, 1998.)

All approaches to belief update are based on propositional logic what
significantly restricts the use of these formalisms in practical
applications. In this talk I generalize the MPMA update operator by
admitting first-order knowledge bases and restricted first-order
update formulae. This allows me to consider scenarios which cannot be
properly formalized in existing approaches to belief update.