This paper models self-enforcing agreements and the evolution of preferences for reciprocity when all agents are rational, Bayesian optimisers. In this 'indirect' evolutionary model, player types are defined not by their "strategies", but by their "preferences". The paper studies the evolution...