Imitation and Learning

As a learning rule in the context of bounded rationality, imitation is both common and plausible. Its strength is that it relies on minimal information about the environment. In its general form, imitation is nevertheless quite complex since the agent has to specify whom, when and with which propensity to imitate. Some forms of imitation perform well while others do not. This chapter demonstrates, for a variety of different environments, which forms of imitation can be individually or socially desirable. One basic intuition is that when agents within a population face similar choices under uncertainty, imitation can lead to a form of information aggregation. Play within the population serves as a memory of how the different actions performed on average. Special circumstances arise in strategic environments where one may be learning from competing agents. Long-run predictions of imitative behavior are presented in pure decision problems and strategic settings and related to benchmarks such as evolutionary stability.