CAREER: Mechanism Design for Resource-Bounded Agents: Indirect Revelation and Strategic Approximations

Harvard University, Cambridge MA

Investigators

Abstract

Systems with distributed computation across open networks share many characteristics with economies, with individual computational devices that are largely autonomous and represent the self-interest of multiple users, including both individuals and businesses. A fundamental challenge is to design useful mechanisms to solve distributed problems in these open systems despite the self-interest of individual agents. This work proposes a foundational study into the issues that arise in computational mechanism design with computationally-bounded agents. One component proposes the study of indirect revelation mechanisms, in which agents are able to participate without computing, or revealing, complete information about their preferences or local constraints. Another component proposes the study of mechanisms that provide approximations to game-theoretic desiderata, such as strategy proofness. In both cases, the research agenda calls both for new theories and for new computational methods, that leverage the theory constructively within systems. The broader impact of the work promises the continued integration of game-theoretic and economic methods into computational systems, with applications ranging from automated mediation in e-marketplaces, to automated negotiation between devices in open systems. An important component involves the development of innovative graduate and undergraduate curricula, to help to educate students about problems at the interface between computer science and economics.