Memory retrieval in the demand game with a few possible splits: Unfair conventions emerge in fair settings

Abstract

Our study examines the long-run evolutionary outcome emerging in scenarios where two populations engage in a demand game with three potential splits. These populations differ in the sample sizes used when best responding to retrieved information from the past. Our findings reveal the existence of a threshold in the setting's fairness (i.e., the fairness of unfair splits) such that, below the threshold (i.e., in an unfair setting), the emerging convention is the fair one, while above the threshold (i.e., in a fair setting), the emerging convention is unfair, favoring the agents with the longer sample size. The threshold gets lower as the difference in the sample sizes increases.