Social norms and game theory: Harmony or discord?

摘要

Recent years have witnessed an increased number of game-theoretic approaches to social norms, which apparently share some common vocabulary and methods. We describe three major approaches of this kind (due to Binmore, Bicchieri, and Gintis), before comparing them systematically on five crucial themes: generality of the solution, preference transformation, punishment, epistemic conditions, and type of explanation. This allows us to show that these theories are, by and large, less compatible than they seem. We then argue that those three theories struggle to account for three phenomena pertaining to social norms (namely context dependence, conflicting norms, and self-evidence), with which any complete game-theoretic account should in principle be able to deal. 1 Introduction2 Accounts 2.1 Binmore: Social norms as equilibrium selection solvers 2.2 Bicchieri: Social norms as conditional behavioural rules 2.3 Gintis: Social norms as choreographers3 Comparing the Accounts 3.1 Games, cues, and generality of solutions 3.2 Norms and other-regarding preferences 3.2.1 Other-regarding preferences 3.2.2 Artificial utility functions 3.3 The role of sanctions 3.4 Getting the epistemic conditions right 3.4.1 Gintis 3.4.2 Bicchieri 3.5 Proximate and ultimate accounts 3.6 Taking stock4 Challenges 4.1 The challenge of context-dependent behaviour 4.1.1 Context-dependence: Varying the game 4.2 Conflicting norms 4.3 Self-evidence5 Conclusion