An Evolutionary Analysis of Narrative Selection

Abstract

We compare the evolutionary selection of agents who employ different heuristics to select the narratives through which they interpret new information. Agents' survival depends on the accuracy of their beliefs. Agents who choose narratives to align with the population's average belief have higher survival rates than those who use alternative heuristics. Conforming reduces an agent's belief variance but inherits the bias of the average belief, so it pays when the latter is close to the state's probability. Evolutionary selection helps maintain this proximity by eliminating inaccurate agents. The other heuristics survive as well, with their shares depending on the degree of uncertainty: agents who tend to form mild beliefs survive more when uncertainty is high, whereas those who tend to form extreme beliefs survive more when uncertainty is low. Conformism is usually considered detrimental, as it discards private information. In our model, conformist agents read their private information in light of the public information, and the consensus they anchor to consists of the agents that survived evolutionary selection, so the cost of conformism is limited.