Abstract
This paper proposes a co-evolution model toward the understanding of the emergence of social morals. The model has two layers, agents' layer and scores' layer. An agent means an individual in Iterated Prisoner's Dilemma (IPD) and a score means a payoff matrix in dilemma games. In this model, behaviors of IPD agents and each value in payoff matrix are co-evolved. Simulation results showed that scores of IPD games are gradually evolved toward the cooperative score which gives more payoff for cooperative strategies and less payoff for defective strategies under all parameters we have tested. However, under a certain limitation of the scores, we observed that strategies of IPD agents could be evolved toward defective, even if the scores are cooperative. Furthermore, we could observe a periodical changes of the scores and the strategies as follows under a certain limitation of the scores. First, all values of all scores move to the cooperative scores and most agents take cooperative strategies because defective strategies of agents earn less payoff. Under the cooperative circumstances, values of scores which have no influence on cooperative moves are mutated freely. Then, this mutation allows invasion by defective strategies created by mutations. Then the cooperative scores which allow the invasion die out, and the world is dominated by the cooperative scores again.
