Understanding the evolution of learning by explicitly modeling learning mechanisms

Abstract

Abstract Models of the evolution of learning often assume that learning leads to the best solution to any task, and disregard the details of the learning and decision-making process along with its potential pitfalls. These models therefore do not explain instances in the animal behavior literature in which learning leads to maladaptive behaviors. In recent years a growing number of theoretical studies use explicit models of learning mechanisms, offering a fresh perspective on the issue by revealing the dynamics of information acquisition and biases arising from it. These models have pointed out possible learning rules and their adaptive value, and shown that the value of learning may crucially depend on such factors as the layout of the physical environment to be learned, the structure of the payoffs offered by different alternatives, the risk of failure, characteristics of the learner and social interactions. This review considers the merits of explicit modeling in studying the evolution of learning, describes the kinds of results that can only be obtained from this modeling approach, and outlines directions for future research.