In Proceedings of the Second International Conference on Multiagent Systems ( ICMAS - 96 ) , Kyoto , Japan , December 1996 Self � ful lling Bias in Multiagent Learning

Learning in a multiagent environment is com plicated by the fact that as other agents learn the environment e ectively changes Moreover other agents actions are often not directly ob servable and the actions taken by the learning agent can strongly bias which range of behav iors are encountered We de ne the concept of a conjectural equilibrium where all agents expec tations are realized and each agent responds op timally to its expectations We present a generic multiagent exchange situation in which compet itive behavior constitutes a conjectural equilib rium We then introduce an agent that executes a more sophisticated strategic learning strategy building a model of the response of other agents We nd that the system reliably converges to a conjectural equilibrium but that the nal result achieved is highly sensitive to initial belief In essence the strategic learner s actions tend to ful ll its expectations Depending on the start ing point the agent may be better or worse o than had it not attempted to learn a model of the other agents at all