Computational Decision Support Regret-Based Models for Optimization and Preference Elicitation
Craig E. Boutilier · Oxford University Press eBooks · 2013
The goal of decision support is to develop methods that assist decision makers. In this chapter computational methods are brought to bear on a multi-dimensional choice problem with the two-part challenge of efficiently determining the decision-maker’s preferences and then finding the best choice. Computer-aided decision support is growing rapidly. This chapter introduces some technical methods of broad applicability in this area, including specification of utility functions with uncertainty, Markov decision processes, and robust optimization. The author exploits the notion of minimax regret, where the goal is to find an option that minimizes the maximum regret relative to all possible manifestations of an uncertain utility function. This approach is robust to incomplete information and facilitates the important process of preference elicitation from the client. There are opportunities for using such approaches to investigate human decision making based on incomplete information; the impact of cognitive costs, biases, and heuristics; and choices made by groups.