The evolution of biased perceptions

We show that in a large class of pairwise strategic interactions, individuals who have perception biases and update their beliefs in a non Bayesian fashion will not only survive in the long run, but also prosper and take over the entire population. This result holds even when the interacting individuals do not always observe their rivals’ types and even when individuals sometimes play against nature. This result suggests that in general there is no reason to believe that evolution will lead to a population of rational agents who revise their beliefs using Bayesian updating. To prove our results, we develop a simple methodology that allows us to study the long run evolution of types in the population starting from (almost) any initial distribution of types in a fully dynamic evolutionary context.