Graphical Models for Utility Elicitation

This paper deals with preference representation and elicitation in the context of multiattribute utility theory under risk. Assuming the decision maker behaves ac-cording to the EU model, we investigate the elicitation of generalized additively decomposable utility func-tions on a product set (GAI-decomposable utilities). We propose a general elicitation procedure based on a new graphical model called a GAI-network. The latter is used to represent and manage independences between attributes, as junction graphs model independences be-tween random variables in Bayesian networks. It is used to design an elicitation questionnaire based on sim-ple lotteries involving completely specied outcomes. Our elicitation procedure is convenient for any GAI-decomposable utility function, thus enhancing the pos-sibilities offered by UCP-networks