Can negotiators outperform game theory?

During the past decade, two parallel literatures emerged and flourished: the behavioral literature on negotiation, and the game-theoretic literature on bargaining. Abundant evidence from both literatures suggests that settlement terms predicted by game-theoretic models are not accurate descriptions of actual negotiated outcomes. Furthermore, the behavioral literature provides much evidence that two common assumptions in game-theoretic models, self-interest and economic rationality, also are not accurate. Both of these discrepancies between theory and evidence - in outcome and in assumptions - primarily imply that actual negotiated outcomes are less efficient than those predicted by game theory. Evidence is beginning to emerge, however, that under certain conditions, negotiators reach more efficient outcomes than predictued by game-theoretic models. The argument that individuals can outperform game-theoretic predictions is not new. In The Strategy of Conflict, Schelling suggests that social norms may lead actual negotiations toward more efficient outcomes than those predicted by forma. game. tehory. Experimental results have supported this thesis. Axelrod (1984), in his work on the prisoners's dilemma game, focuses on the role of expected future interaction as an explanation for when interdependent parties can outperform the game-theoretic solution that dictates defection in a finite-period prisoner's dilemma game. Axelrod's empirical finding is consistent with the theoretical work of Kreps, Milgrom, Roberts, and Wilson (1892), who develop a model that deviates from standard game theory by assuming a small probability that the other party either is not rational or is not self-interested, and so will not follow the rational pure defection strategy. Our goal is to show that the outperformance of game-theoretic predictions is possible in even a one-trial context, and to offer social and cognitive explanations for such outcomes.