Figure 1

The experimental setup consisted of a series of “global” and “local” interactions.

(a,b) In each round, participants first took part in a global interaction stage and then in a pairwise local interaction stage. (a) In the global stage, groups of on average 39 participants (min = 17, max = 60, sd = 10.28, N = 646) played 20 rounds of the Public Goods Game (PGG). In each round, participants were endowed with 20 MUs: they chose how many of these MUs to contribute to a common pool and how many to keep for themselves. The contributed units were doubled and split equally among all group members. (b) In the pairwise interaction stage, participants were connected to two other group members on a ring-structured network (in experiment 1; for differences to experiment 2, see SI Section 1). In each round, participants played a Prisoner’s Dilemma (PD) with each neighbour: they could choose to cooperate by paying 6 units to give 18 units to their neighbours; or defect by doing nothing. Thus mutual cooperation yielded a benefit of 12 for both, unilateral cooperation cost cooperators 6 units while providing defectors with 18 units, and mutual defection did not alter the payoff of either participant. (c,d) The control and treatment conditions differed in what participants could observe about their neighbours. (c) In the control condition, participants were not told how many MUs their neighbours contributed in the PGG stage. (d) In the treatment condition, conversely, participants were informed of their neighbours’ contributions in the PGG while making their pairwise decisions in the PD.