Trust and Equity Theory in Prisoner's Dilemma

Abstract

In human society, people choose to cooperate or not depending on how much trust they put in others. In addition, people often pursuit fair treatment considering the ratio of their input. We introduce these personalities, called trust and equity, to intelligent adaptive agents in prisoner's dilemma. Intelligence of players is modeled using recurrent neural network. In a game between two agents, they take different actions according to the degree of trust or equity. We also simulate the evolution of 3 societies with various degrees of trust only, equity only, and with multiple personalities. Agents who trust others at the moderate level show cooperative behaviors and finally survive. Meanwhile, agents who consider equity more than self-interest become dominant as a result of evolution. The evolution of a merged society demonstrates the different behavior of each personality.