Benevolent and Malevolent Ellsberg Games

Traditionally, real experiments testing subjective expected utility theory take for granted that subjects view the Ellsberg task as a one-person decision problem. We challenge this view: Instead of seeing the Ellsberg task as a one-person decision problem, it can be perceived as a two-player game. One player chooses among the bets. The second player determines the distribution of balls in the Ellsberg urn. The Nash equilibrium predictions of this game depend on the payoff of the second player, with the game ranging from a zero-sum one to a coordination game. Meanwhile, the predictions by ambiguity aversion models remain unchanged. Both situations are implemented experimentally and yield different results, in line with the game-theoretic prediction. Additionally, the standard scenario (without explicit mention of how the distribution is determined) leads to results similar to the zero-sum game, suggesting that subjects view the standard Ellsberg experiment as a game against the experimenter.