In multi-agent systems, agents receive incomplete local information, and they must achieve global tasks. We consider this ability to determine the appropriate action to correspond exactly to “norms of behavior” in the human sense. In this study, we constructed a competitive social system consisting of selfish autonomous agents. Each agent had an independent evaluation table for their actions. The Agents’ strategies were adapted on the basis of their individual evaluations, and agents’ norms of behavior satisfy. In such systems, only the agents that adapt not only their strategy but also their norms of behavior to the environment were able to survive.