Does mutual knowledge of preferences lead to more equilibrium play? Experimental evidence

In many experiments, the Nash equilibrium concept seems not to predict well. One reason may be that players have non-selfish preferences over outcomes. As a consequence, even when they are told what the material payoffs of the game are, mutual knowledge of preferences may not be satisfied. We experimentally examine several 2x2 games and test whether revealing players' preferences leads to more equilibrium play. For that purpose, we elicit subjects' preferences over outcomes before the games are played. It turns out that subjects are significantly more likely to play an equilibrium strategy when other players' preferences are revealed. We discuss a noisy version of the Bayesian Nash equilibrium and a model of strategic ambiguity to account for observed subject behavior