Leveraging social influence for the wisdom of the crowd

Author(s): Schimmelpfennig, R; Ehret, S; Voigt, S; Efferson, C

Theoretical and empirical research suggests that human social learning is both frequency- and payoff-dependent. Less evidence exists about the joint effect of both types of social information. We ran a two-staged incentivised online experiment in Kenya (N=1400). First, participants were ranked based on their performance in a Raven Matrices Task. Second, participants were grouped to make decisions in a sequential task in which participants receive a private signal and can observe the decisions of upstream participants. In a 2x2 factorial design, we manipulated whether high-performing participants would make their decisions early or late in the sequence and whether participants are informed about the ordering or not. We find strong evidence for frequency-dependent social learning across all conditions. However, while frequency- and payoff-dependent cues interact to increase the accuracy of decisions, we don’t find any significant interaction effect on social learning. To understand how learning shapes decisions in complex real-world scenarios, research should further investigate the joint effects of social learning biases for decision-making.