Preferences in AI and CP: Symbolic Approaches
Papers from the AAAI Workshop
Ulrich Junker, Chair
Preferences are gaining more and more attention in AI and CP. As described in the article of Jon Doyle and Richmond Thomason about qualitative decision theory (AI Magazine, 1999), AI provides qualitative methods for treating preferences that can improve or complement numerical methods for treating preferences from classical decision theory. Preferences are essential to treat conflicting information in nonmonotonic reasoning, reasoning about action and time, planning, diagnosis, configuration, and other areas in knowledge representation and reasoning. In constraint programming, preferences are used to treat soft constraints, and to reduce search effort. Preferences are complementary to constraints and represent an AI counterpart to objective or utility functions. AI permits complex preference representations and thus allows to reason with and about preferences. Hence, AI provides a new perspective for formalizing information that is essential for many decision making problems, e.g. web-based configuration, scheduling, robot planners. The purpose of this workshop is to provide a forum for exchanging experiences with different (symbolic) approaches for treating and applying preferences, for comparing and bridging gaps between these approaches, and for identifying challenging questions for future research. It addresses theoretical approaches, algorithms, and implemented systems.