20 November 2007
Sindhu Joseph

How to model an institutional agent that can autonomously reason about violating a norm? BDI theory of agency helps to model an intentional agent driven towards action. Yet it is inadequate to model an autonomous agent simply because it has no mechanism to choose a coherent subset from its set of cognitions when conflicts arise. Further it has no tools to maintain or adapt its cognitions given its social or institutional interactions. This is in general the case for most existing agent models. In this presentation we introduce an agent model based on coherence theory. Theory of coherence has been well studied in the field of cognitive science and is a general theory to describe the world. The intuition behind this theory is that elements (propositions, concepts, theories) can be associated and a positive association between two elements support each other while a negative association tend to exclude one. A coherent set is one in which the elements are mutually supported maximally. For example creationism and evolution does not form a coherent set as they have many negative associations between the two theories. We give a formalization of Thagard's theory on coherence as constraint satisfaction and use it to explain the reasoning process of an institutional agent that permits the agent to drop beliefs or to violate norms in order to keep a maximal state of coherence.