Decision-theoretic GOLOG with qualitative preferences
Christian Fritz, Sheila A. McIlraith · 2006
Personalization is becoming increasingly important in agent programming, particularly as it relates to the Web. We propose to develop underspecified, task-specific agent pro-grams, and to automatically personalize them to the pref-erences of individual users. To this end, we propose a framework for agent programming that integrates rich, non-Markovian, qualitative user preferences expressed in a lin-ear temporal logic with quantitative Markovian reward func-tions. We begin with DTGOLOG, a first-order, decision-theoretic agent programming language in the situation calcu-lus. We present an algorithm that compiles qualitative pref-erences into GOLOG programs and prove it sound and com-plete with respect to the space of solutions. To integrate these preferences into DTGOLOG we introduce the notion of multi-program synchronization and restate the semantics of the lan-guage as a transition semantics. We demonstrate the utility of this framework with an application to personalized travel planning over the Web. To the best of our knowledge this is the first work to combine qualitative and quantitative prefer-ences for agent programming. Further, while the focus of this paper is on the integration of qualitative and quantitative pref-erences, a side effect of this work is realization of the simpler task of integrating qualitative preferences alone into agent programming as well as the generation of GOLOG programs from LTL formulae. 1