An Emergence Model of Sense of Values in a Competitive Social System

Abstract

The objective of this study is the construction of a multi-agent model for competitive multi-robotics systems from the viewpoint of sense of values (i.e., values which influence behavior). In this model, each agent has his own sense of values. Each agent is able to adapt his own sense of values and his strategies to: his environment, other agents’ sense of values, and other agents’ strategies. To represent these sense of values, evolutionary game models are used (Ishida, 1996). Iterated Prisoner’s Dilemma (IPD) is one of the most famous game models with 2-players (Axelrod, 1984). Our model has a dynamic payoff function based on variables which are named “Karma.” Karma are the independent variables that are affected by each agent’s actions. Experiments show that agents acquire common sense which made cooperative behavior possible. It was revealed that this payoff function is effective in controlling exploitation of agents in the long term, and that this function is extremely robust in spite of externally introduced noise which modified the sense of values. These results show the possibility of cooperative behavior through generating selfish agents that have no common sense. This model also implies that the emergence of sense of values in competitive multi-robotics systems is possible.