Emotions as strategic signals

In this article, we ask how much, if anything, of Robert Frank's (1988, 2004) theory of emotions as evolved strategic commitment devices can survive rejection of its underlying game-theoretic model. Frank's thesis is that emotions serve to prevent people from reneging on threats and promises with enough reliability to support cooperative equilibria in prisoner's dilemmas and similar games with inefficient dominant equilibria. We begin by showing that Frank, especially in light of recent revisions to the theory, must be interpreted as endorsing a version of so-called 'constrained maximization' as proposed by Gauthier (1986). This concept has been subjected to devastating criticism by Binmore (1994), which we endorse: no consistent mathematical sense can be made of games in which constrained maximization is allowed. However, this leaves open the question of whether Frank has identified a genuine empirical phenomena by means of his confused theoretical model. We argue that he in fact has; but that seeing this depends on our rejecting a muddled folk-psychological model of emotions, which Frank himself follows, according to which emotions are inner states of people. Instead, following Dennett (1987, 1991) and other so-called 'externalist' philosophers of cognitive science, we argue that emotions, properly speaking, are social signals coded in culturally evolved intentional conventions that find their identity conditions outside of individuals, in the social environment. As such, their evolutionary proper functions lie in their capacity to enable individuals to solve what we call 'game determination' problems - that is, coordination on multiple-equilibrium meta-games over which base-games to play. This allows emotions to indeed serve as commitment devices in assurance games (though not in prisoner's dilemmas). Thus the empirical core of Frank's thesis is recovered, though only by way of drastic revisions to both the game theory and the psychology incorporated in his model.