A model of emotion in the prisoner’s dilemma

This paper adopts the definition ldquoan emotion is a scalar summary of complex environmental circumstancesrdquo in the context of the iterated prisonerpsilas dilemma. Prisonerpsilas dilemma players, represented both as look-up tables and artificial neural nets, are evolved with and without emotion and noise. The availability of emotion is found to have a substantial impact on the evolution of cooperation and interacts with noise in a complex manner. The impact of having a single bit of emotional information on lookup tables is also found to be different from the analogous impact on neural nets. The emotion used in this experiment is a single bit which is set if an agentpsilas opponent has defected into cooperation more often than the agent itself has done so. This simple emotion has an substantial, non-uniform impact on the behavior of evolving populations of prisonerpsilas dilemma agents. The emotion implemented in this study is only one of many possible emotions, suggesting that even this limited and mathematically tractable definition of emotion yields a rich collection of possible research topics. The most recognizable impact of adding emotion to agents in this study is to move their behavior away from the middle of the behavioral spectrum toward either sustained cooperation or defection.