15th European Conference on Artificial Intelligence
|July 21-26 2002 Lyon France|
Salem Benferhat, Didier Dubois, Souhila Kaci, Henri Prade
The preferences of an agent can be expressed in various ways. The agent may indicate goals having different levels of priority for him, or provides classes of choices with their level of satisfaction for him. The first type of specification can be captured in possibilistic logic under the form of constraints on a necessity measure. It is shown in this paper that the second manner for expressing preferences can be encoded as constraints on a so-called "guaranteed possibility" measure (a min-decomposable function with respect to disjunction). The paper shows how each representation is semantically associated with a possibility distribution (which plays the role of a value function), and how necessity-based possibilistic logic representations can be translated directly into a guaranteed possibility-based representations and vice-versa. In logical terms, it corresponds to the transformation of a generalized DNF into a generalized CNF. Reasoning in guaranteed possibility-based logic is also discussed. Moreover, the two types of representations can also be shown to be equivalent to sets of conditional preference statements. Thus, different basic modes of preference expression can be captured in the same framework.
Keywords: Reasoning under Uncertainty, Decision Theory
Citation: Salem Benferhat, Didier Dubois, Souhila Kaci, Henri Prade: Possibilistic logic representation of preferences: relating prioritized goals and satisfaction levels expressions. In F. van Harmelen (ed.): ECAI2002, Proceedings of the 15th European Conference on Artificial Intelligence, IOS Press, Amsterdam, 2002, pp.685-689.