Abstract

Broadband networking allow agents to share distributed resources. In a peer-to-peer networks, a data file can be shared nonrivalrously, because any number of agents can read it at the same time. In grid computing, computing power is shared rivalrously, but it can be multiplexed through time, since if one agent is not using his computer at some instance then another agent may. The raison d’etre for such systems is that they are self-financing and that each agent is better off contributing resources to a shared pool than he would be otherwise. The problem for a system designer is to arrange things, using simple enforceable rules, fees, etc., so that heterogenous agents are incentivized to make socially efficient decisions about the sizes of their contributions to the common resource pool. Free-riders should be prevented, since this can make the system collapse. (Think of a large grid computing network in which the sharing rule is “equal shares”. A given agent may as well decide to contribute 0, since he stills obtain the same share as any other agent. But if all agents contribute 0 then the system becomes useless.) I will present a optimization model that captures salient features of the above problem. A reoccuring theme of “solutions” is that we must prevent agents who value the system highly from masquerading as those who do not, and so it may be optimal to prevent the latter type of agent from participating at all.