Mechanism Design for Resource Bounded Agents Noa E. Kfir-Dahav Dov Monderer Moshe Tennenholtz Faculty of Industrial Engineering and Management Technion – Israel Institute of Technology Haifa 32000, Israel December, 1998 Abstract The theory of mechanism design deals with the design of proto- cols for non-cooperative multi-agent systems. A major task of this theory is the design of protocols that will maximize the social welfare of the agents. An ideal mechanism will optimize social welfare and will be strategy-proof, i.e. the dominant strategy of each agent will be to participate in the mechanism and to reveal his true goal and worth, as well as budget-balanced, i.e., the mechanism should not im- pose any payments from the center/organizer to the agents. Indeed, the Clarke’s mechanism, which is central to information economics and to games with incomplete information satisfies these properties. However, we show that the Clarke’s mechanism employs the use of procedures for optimizing social welfare, which are NP-hard. Hence, these procedures should be replaced by heuristics. We present a set of natural properties (axioms) of such heuristics that, when satisfied, enable to obtain the desired strategy-proofness and budget balance properties. Our result enables to extend the central protocol of the theory of mechanism design to the context of resource-bounded agents. 1