Abstract

Several evolutionary mechanisms have been proposed to explain how natural selection leads to cooperation between competing individuals. Social dilemmas modelled with the aid of game theory capture the essence of this problem, and computer simulation is usually the technique used to test and formalise those explanatory mechanisms. However, scarce attention has been paid to what the notion of 'mechanisms' means and involves in the literature. Also, the key issue about when a computer simulation provides a good explanation tends to be ignored. In this article, we tackle these two drawbacks in the literature by calling attention to the implications of the notion of 'social mechanism' along different conceptual dimensions, such as ontological status, regularity, transparency, intelligibility, and reduction. We also claim that computer simulation, and specially agent-based modelling, provides a plausible explanation to social cooperation only if it satisfies some criteria of empirical adequacy instead of just being capable of generating cooperation in a virtual system. Finally, we relate these issues to five evolutionary mechanisms that explain the evolution of cooperation. We review and briefly describe the literature on these mechanisms, and we explain their most important features, how they are to be considered along the conceptual dimensions used to describe the notion of mechanism, what is the empirical and computational evidence to support them, and which are the shortcomings that each of them has as explanatory hypotheses for the evolution of cooperation.