Evolutionary Games and Social Conventions

Some thirty years ago Lewis published his Convention: A Philosophical Study (Lewis, 2002). This laid the foundation for a game-theoretic approach to social conventions, but became more famously known for its seminal analysis of common knowledge; the concept receiving its canonical analysis in Aumann (1976) and which, together with the assumptions of perfect rationality, came to be defining of classical game theory.

However, classical game theory is currently undergoing severe crisis as a tool for exploring social phenomena; a crisis emerging from the problem of equilibrium selection around which any theory of convention must revolve. In response, the so-called evolutionary turn has developed. While retaining the broad framework, in which games are described in terms of strategies and payoffs, this marks a transition from the classical assumptions of perfect rationality and common knowledge to assumptions characterising agents as conditioned for playing certain strategies upon the population of which evolutionary processes operate. By providing accounts of equilibrium selection and stability properties of behaviours, the resulting frameworks have been brought to work as well-defined metaphors of individual learning and social imitation processes, from which a revised theory of convention may be erected (see Sugden 2004, Binmore 1993 and Young 1998). This paper makes a general argument in support of the evolutionary turn in the theory of convention by a progressive exposition of its successful application to a variety of simple, but paradigmatic games. In doing this, it further examines and qualifies on what may be said within this framework about the relations between social conventions on the one hand, and phenomena such as Pareto-efficiency, risk, discrimination, self-interest and cooperation on the other. For most of the arguments, the formalisation will be kept at a minimum as well as restricted to two-player interactions.