Logical Formalizations Built on Game-Theoreti Argument about Commitments

摘要

The formalization of commitment is a topic of continuing interest in Artificial Intelligence (AI)'s understanding of human cooperative activity and organization. Such formalizations are crucial for clarifying rational behavior. AI research on commitments, however, has been focusing on describing systems of agents, neglecting the individual incentives to perform certain actions. We argue in this paper that an understanding of a system of agents needs to incorporate not only a logical system of possible actions, but also an incentive structure related to the actions and the interdependence of agents involved in interactions between more agents. As an example we will discuss the use of commitments in interactions between two agents. By adding game-theoretic reasoning, we will not only be able to describe different commitment systems in various (legal) settings, but we can also determine whether or not such commitment system is expected to be socially efficient, desirable, and able to influence human behavior.