Local competition promotes hurtful behavior towards unknown others

Humans frequently cooperate to achieve benefits unattainable through selfish behavior. Punishment of free-riders is key for sustaining cooperation, but also costly, raising questions about the origins of such behavior. Some evolutionary models suggest that individuals may be willing to hurt unknown others in competitive environments. However, empirical evidence about the effect of the scale of competition on strategies involving hurtful behavior in human groups, such as spite, retaliation for hurting, and punishment of free-riders, is missing. Using a laboratory experiment, we manipulate the scale of competition to investigate its influence on hurtful and helpful behaviors between unknown humans in an indirect reciprocity game. We observe distinct behavioral patterns between local and global competition. When competition is local and thus confined to take place within isolated groups, we find frequent hurting as an expression of spite and retaliation. In contrast, when competition extends globally across several groups, hurting is used for the punishment of free-riders primarily, whereas helping behavior is rewarded, which together promotes cooperation. Thus, while isolated competition fosters inefficient, antisocial behavior, global competition encourages prosociality and cooperation.

Methods
156 subjects participated in the experimental sessions that were conducted at the CREED laboratory of the University of Amsterdam. Participation was voluntary, and subjects were recruited through email announcements from a standard university subject pool. Each experimental session lasted approximately 60 minutes. Subjects' total earnings in euros were determined by the relative ranking of the points (denoted by ‘francs’ in the experimental instructions) earned over all the rounds of the indirect reciprocity game. The only difference between the treatments is how the ranking is determined for these euro earnings, either within a local group or globally across several groups. By design, average earnings were 20 euros in both treatments.

Each subject participated in only one session, and none had previously participated in a similar experiment. Subjects were seated in separate cubicles, which ensured anonymity during the experiment. We consider each group of 6 subjects as one independent observation, because subjects only interacted with the other 5 subjects in their group.

The experiment was computerized and all interactions took place anonymously through computers. The supplementary online materials contain the instructions script provided to subjects. At the end of each experimental session subjects were paid their euro earnings in private and in cash. In all treatments, subjects were informed of the matching procedure, the number of rounds, and how to calculate the payoff. The experiment was neutrally framed. The donor’s choices were called ‘blue’ (help), ‘green’ (snub) and ‘purple’ (hurt).

Each session consisted of 100 rounds of the experimental game. In each round each six-subject cohort was randomly partitioned in three donor–recipient pairs. Therefore, pairing was independent of previous choice, distinguishing it from previous theoretical models. A recipient received no information about her donor and made no decision. A donor, however, learned the most recent three decisions made by her recipient.

By choosing between the three colors the donor decided whether she would help, hurt or snub the recipient. The recipient learned her donor’s chosen color and her own earnings in each round only after all donors had made and confirmed all their decisions. The round number was always visible and both the donor and the recipient could always access the information they observed and their own decisions in all past rounds. In each round a pair of matched subjects only saw the decision of the donor in their pair.