Criteria of Choice in Non-Constant-Sum Games

Experiments were performed on single-shot 2×2×2×2×2 games (five players with two strategies each). Variants of the following types of games were used: (1) no conflict, in which the “cooperative” strategy C is always more advantageous than the “noncooperative” strategy D; (2) Steering Others' Fate, in which any single player is indifferent between C and D but, by choosing D, increases the others' payoffs; (3) five-person Prisoner's Dilemma; (4) Justified Trust, in which it is most advantageous to choose C only if everyone else does so; and (5) five-person Chicken. As expected, the fraction of players choosing C was largest in no-conflict games and smallest in Prisoner's Dilemma. The behavioral results were compared with players' and control nonplayers' predictions about the fractions of C choices. Fractions predicted were consistently larger than those observed. In all games, Ss who chose predominantly C expected higher levels of C than those who chose predominantly D.