Information Processing for a Finite Allocation Mechanism.

We deal with an organization having n individual agents (the agents) and one central agent or agency (the coordinator), and the problem of allocating a fixed supply of n commodities among the agents. The coordinator wishes to distribute the commodities in such a way as to maximize an objective function whose arguments are functions known (implicitly) to the individual agents, but not to the coordinator. The more information the coordinator obtains about the individual agents' functions, the better the allocation decision he or she can make; however, gaining such information is costly, so that the coordinator must balance the potential gain against the cost of acquiring additional information. We find a satisficing solution to this problem which is computable and which terminates in a finite number of steps.