On the Role of Expectations in Multi-agent Reasoning and Decision Making

In game theory, the notion of expectation is made implicit. That is, most game-theoretic solution concepts, such as equilibrium, rationalizability, assume that agents expect one another to play their respective strategies according to the solution(s) induced by these solution concepts. This assumption limits the applicability of these solution concepts to many multi-agent systems situated in open and dynamic environments due to obvious reasons: agents being resource-bounded, agents being rational-bounded, and agents not knowing some relevant aspects of the game they are playing. In this paper, by making agents' expectations explicit we provide a generic and flexible framework to allow concepts and results of game theory to be more readily mapped to the reasoning and decision making mechanisms in multi-agent systems.