Preference change (Q302212)

Standard decision and game theory presuppose that in any model, the sets of all possible agents, states, actions, player-types and beliefs of agents are pre-defined, and that preferences can change only as a result of change in belief. However, the literature contains examples of decision problems that cannot be handled within these confines, and also contains a number of proposals to relax the constraints -- but generally one or two at a time. The paper under review proposes a formal model in which they are loosened together. It builds on the idea of representing preferences as nets, rather analogous to Bayesian nets [\textit{C. Boutilier} et al., J. Artif. Intell. Res. (JAIR) 21, 135--191 (2004; Zbl 1080.68685)]. It also follows \textit{S. O. Hansson} [Theory Decis. 38, No. 1, 1--28 (1995; Zbl 0835.90001)] in adapting to the context of preference change certain ideas of the AGM account of belief change. But, importantly, it also allows agents to reason on the basis of merely partial descriptions of preferences, even of their own, so that preference change becomes a default process.