The Evolution of Communication in a Sender/Receiver Game of Common Interest with Cheap Talk

Jasmina Arifovic and Curtis Eaton
Simon Fraser University
Arifovic@sfu.ca

Abstract

We are interested in the evolution of communication in a complex market environment in which there are many agents of different types. The framework for our analysis is a sender/receiver game. Senders send costless messages to signal their types, and receivers take actions regarding senders' types. There is a multiplicity of Nash equilibria: no-signaling equilibria with no effective communication; and perfect signaling equilibria in which effective communication occurs. Perfect signaling equilibria are evolutionarily stable. We describe players as chromosomes, which are used to determine a message to be sent for each player type and an action to be taken for each message received. Chromosomes are updated using a genetic algorithm. During the evolution, the proportion of time spent near an equilibrium is very high. The population of players can be seen as climbing an equilibrium payoff ladder. It first coordinates on an equilibrium with the lowest payoff and no effective communication, and then it climbs towards the Pareto optimal equilibrium in which all types are effectively communicated.