An argumentation based semantics for agent reasoning

Abstract. A key challenge for agent architectures and programming paradigms is to account for defeasible reasoning over mental attitudes and to provide as-sociated conflict resolution mechanisms. A growing body of work is looking to address these challenges by proposing argumentation based approaches to agent defeasible and practical reasoning. This work conforms to Dung’s seminal ar-gumentation semantics. In this paper we review our previous work in which we extend Dung’s semantics to allow for inclusion of arguments that express pref-erences between other arguments. In this way we account for the fact that pref-erence information required to resolve conflicts is itself defeasible and may be conflicting. We then propose the extended semantics as a semantics for agent defeasible and practical reasoning, and substantiate this claim by showing how our semantics can characterise, and indeed provide a framework for extending, existing approaches to agent reasoning over beliefs, goals, and actions.