General Equilibrium Markets as Distributed Computational Systems

One principal capability of markets is the coordination of information distributed among market participants who typically have competing incentives. Despite distributed information and conflicting incentives, competition among market participants often leads to socially desirable resource allocation. In this sense a market is a distributed multi-agent mechanism that, in effect, finds an optimal solution to a computational problem that is not known to any individual market participant. The purpose of this paper is to demonstrate that capacity of markets by defining automated agents that interact with one another in a distributed system that implements a socially desirable resource allocation. In this sense, market agents extend in several meaningful directions our concept of the function and capabilities of artificial intelligence. Market agents have been developed for single (partial equilibrium) markets going back over thirty-five years. This paper extends that research strategy in two directions. The paper reports results of simulations with automated agents in a general equilibrium market and, in addition, those agents interact in their market with human subjects in a market experiment that includes both human subjects and automated agents, so that we can better assess the bargaining capacity of the agents relative to human capabilities.