UNSW talk
Australian Graduate School of Management
20 Sept 2001
------------


		Bayesian Intelligent Decision Support


			    Dr. Kevin Korb
		     School of Computer Science &
			 Software Engineering
			  Monash University
			Clayton, Victoria 3800
			      Australia

		    email: korb@csse.monash.edu.au


Intelligent decision support is central for applying Artificial
Intelligence to decision making in industry and science.  An important
class of computational decision support tools incorporate decision
trees for calculating the expected value of different choices and
offer substantial support for eliciting probabilities and utilities
and for performing sensitivity analysis.  If the decision making
process concerns a dynamic system, where changing conditions may
require probabilities to be reassessed or where intermediate choices
may impact future conditions, such tools require that the decision
trees be explicitly altered and the expected values (or sensitivity
analysis) be recomputed.  Bayesian networks are a relatively recent
technology for reasoning under uncertainty, which explicitly model
causal relationships as well as supporting decision analysis.  This
technology is reaching maturity, with a number of products becoming
available on the market in recent years.  Furthermore, computational
statistical packages are becoming available for the automated learning
of Bayesian nets from sample data, relieving some of the burden of
eliciting probabilities from experts.  What's lacking thus far is the
provision of elicitation, validation and sensitivity analysis packaged
in a usable human-computer interface on top of the underlying Bayesian
net technology.  The Bayesian Intelligent Decision Support (BIDS)
project at Monash proposes to address these deficiencies.  In this
talk I will introduce the ideas behind Bayesian net technology and
their proposed use for intelligent decision support.