Learning Optimal Seller Strategies with Intelligent Agents: Application of Evolutionary and Reinforcement Learning

The role of automated agents in the electronic marketplace has been growing steadily and has been attracting a lot of research from the artificial intelligence community as well as from economists. We consider the problem of homogeneous sellers of a single raw material or component vying for business from a single large buyer, and present artificial agents that learn near-optimal seller strategies. Standard game-theoretic analysis of the problem assumes completely rational and omniscient agents to derive Nash equilibrium seller policy. We show that in our problem such an equilibrium is unstable, and present simple reinforcement and evolutionary learning agents that learn strategies with better than Nash payoffs.