Towards qualitative approaches to multi-stage decision making

AbstractIn this paper we propose a generalisation to multi-stage decision making of Dubois and Prade's qualitative decision theory. Our framework is a qualitative, possibilistic counterpart to Markov decision processes, and the computation of an optimal policy is done in a way similar to dynamic programming. We first study in detail the case where uncertainty about the results of actions is represented by possibility distributions and goals are described in a non-fuzzy way by a subset of the set of final states. Then we extend our framework to the case where goalness is defined fuzzily, by a qualitative utility function on the set of final states