| Intelligent agents performing a task on behalf of a user are often faced with decisions in which one of a set of possible actions must be chosen. These decisions are typically made by considering the agent's (often incomplete) information about the possible consequences of those actions and the probability and utility (desirability) of each of these consequences of those actions and the probability and utility (desireability) of each of these consequences. The objective of this course is to introduce students to decision theory and its applications in the design of intelligent agents for achieving a wide range of task. Topics include decision analysis (payoff tables, decision trees, minimax regret analysis), preferences and utility theory, multiattribute utility functions, preference elicitation techniques, Markov Decision processes, automated negotiation, and auctions. |
