God games: An experimental study of uncertainty, superstition, and cooperation

Abstract

This paper uses a novel lab experiment to test claims about the origins and functions of religion. We modify the standard public goods game, adding a computer-based agent that adjusts earnings in ways that might depend on players' contributions. Our treatments employ three different descriptions of the adjustment process that loosely correspond to monotheistic, atheistic, and agnostic interpretations of the computer's role. The adjustments neither mask players' contributions nor magnify their impact. Yet players in all three adjustment treatments contribute much more than those who play the standard public goods game. Players' contributions and survey responses show that adjustments induce superstitions in all treatments, with the strongest superstitions appearing in the quasi-monotheistic treatment and the weakest in the quasi-atheistic treatment. Text-based communication raises contributions and strengthens coordination. But when paired with the quasi-monotheistic description, communication also promotes counterproductive quests for winning numbers.