ABSTRACT
This chapter describes two approaches to decision support: Influence Diagrams (IDs) and Symbolic Argumentation.
Bayesian Networks (BNs) and rule-based formalisms for hypothesis evaluation do not explicitly incorporate the concepts of action and utility, which are ubiquitous in decision-making contexts. When incorporating these concepts, BNs are converted to IDs, subsuming the functionality of the normative theory of decision-making under expected utility theory (EUT) and decision trees. Decision-making based on IDs means choosing, from a set of alternative actions, the most suitable action (or hypothesis) yielding the highest utility. For inferencing in IDs, we extend the junction tree algorithm for BNs. The extended algorithm presented in this chapter compiles an ID into a strong junction tree in which the computation of maximum expected utility can be done by local message-passing in the tree.
