Introduction of social learning in board games

This paper discusses the use of social learning theory in training agentsto play hoard games. Specifically it aims to provide a proof of concept of howreinforcement learning methods can he improved by inducing social effects onagents as opposed to self-play methods used in teaching agents that playtwo-player games. The paper introduces social learning theory and also discussesthe initial findings on an experiment with the two player board gameTic-Tac-Toe. The socially taught agents exhibit behavior that is superior to theself play agents. The social agents are able to play the game at a, higherskill level than the self taught agent even though they are trained with thesame number of games. There is now reason to carry through more research onformalizing social frameworks for two player game agent systems. ICIC International