368 Stijn Heymans, Davy Van Nieuwenborgh, and Dirk Vermeir Preferential Reasoning on a Web of Trust Stijn Heymans, Davy Van Nieuwenborgh ⋆ , and Dirk Vermeir ⋆⋆ Dept. of Computer Science Vrije Universiteit Brussel, VUB Pleinlaan 2, B1050 Brussels, Belgium {sheymans,dvnieuwe,dvermeir}@vub.ac.be Abstract. We introduce a framework, based on logic programming, for prefer- ential reasoning with agents on the Semantic Web. Initially, we encode the knowl- edge of an agent as a logic program equipped with call literals. Such call literals enable the agent to pose yes/no queries to arbitrary knowledge sources on the Semantic Web, without conditions on, e.g., the representation language of those sources. As conflicts may arise from reasoning with different knowledge sources, we use the extended answer set semantics, which can provide different strate- gies for solving those conflicts. Allowing, in addition, for an agent to express its preference for the satisfaction of certain rules over others, we can then induce a preference order on those strategies. However, since it is natural for an agent to believe its own knowledge (encoded in the program) but consider some sources more reliable than others, it can alternatively express preferences on call literals. Finally, we show how an agent can learn preferences on call literals if it is part of a web of trusted agents. 1 Introduction The current WWW is a gigantic pool of data, where one can easily imagine two web sites saying the opposite. Human users are capable of deciding which sources they find trustworthy or not (irrespective of the fact whether they actually are or not). Semantic Web software agents [18] on the other hand would have an equally vast amount of data at their disposition, but a far more difficult time differentiating between good and bad information. In this paper, we will gradually build a (abstract) software agent, i.e. an entity on a web of trust that can reason with a diverse pool of (possibly mutually inconsistent) knowledge sources. The basic underlying reasoning framework we use for such an agent is answer set programming (ASP) [13, 3], a logic programming paradigm with a stable model semantics for negation as failure. A logic program corresponds to knowledge one wishes to represent, or, more specifically, to an encoding of a particular problem, e.g. a planning problem [24, 9]; the answer sets of the program then provide its intentional knowledge, or the solutions of the encoded problem, e.g. a plan for a planning problem. ⋆ Supported by the FWO. ⋆⋆ This work was partially funded by the Information Society Technologies programme of the European Commission, Future and Emerging Technologies under the IST-2001-37004 WASP project. Y. Gil et al. (Eds.): ISWC 2005, LNCS 3729, pp. 368-382, 2005. c  Springer-Verlag Berlin Heidelberg 2005