Practical Social Reasoning Systems

The Belief-Desire-Intention (BDI) model of rational agency provides an abstract architecture for practical reasoning systems and is today regarded as one of the key contributions of multiagent systems research to AI. So far, however, a similar model for practical social reasoning systems that is generic enough to be used as a unifying framework for the diverse methods used to manage the interactions between autonomous agents is still amiss. And yet, as modern day AI applications are moving increasingly towards open environments characterized by interaction among large numbers of potentially self-interested, heterogenous and mutually opaque agents, the need for such a model is felt more strongly than ever. In this paper, we propose the Expectation-Strategy-Behavior (ESB) model for practical social reasoning systems that confronts the challenge of identifying such a unifying framework. We claim that this architecture is generic enough to cover most existing approaches to reasoning about interaction described in the literature, and that it can be easily integrated with the general BDI architecture.