Abstract:
Information modeling and management are fundamental to decision making, and
in this talk I will describe systems that support decision making in
environments where the information is uncertain, incomplete or liable to
change. The main idea is to model the plausibility of the information and
the preferences of the decision maker to improve the decision outcome when
ambiguities or inconsistencies arise. More specifically, I will briefly
outline the area of belief revision and describe some specific techniques
that are used to model and manage consumer preferences in Internet Marketing
and Customer Relationship Management applications. A wide range of problems
involve decision making in environments where uncertainty, incompleteness
and change are prevalent, and as a result techniques in belief
revision can be used in a wide variety of applications, e.g. information
filtering, intelligent payment systems, intelligent interfaces, intelligent
tutoring systems, and integrating information from multiple sources.
It turns out that integration is a generalised form of belief revision;
instead of incorporating a new piece of information into a knowledge base
by computing the "most important" information from the knowledge base that
is compatible, we build a new knowledge base by fusing the "most important"
compatible information from each knowledge base.
A web-based belief revision demo system is available at
http://ebusiness.newcastle.edu.au/saten It is written in Java and is
currently being incorporated into the agent-oriented system JADE
to be evaluated for ebusiness applications. Information about JADE can be
found at http://sharon.cselt.it/projects/jade/