Misrepresentation of group contributions undermines conditional cooperation in a human decision making experiment

Abstract

Cooperative behaviour can evolve through conditional strategies that direct cooperation towards interaction partners who have themselves been cooperative in the past. Such strategies are common in human cooperation, but they can be vulnerable to manipulation: individuals may try to exaggerate their past cooperation to elicit reciprocal contributions or improve their reputation for future gains. Little is known about the prevalence and the ramifications of misrepresentation in human cooperation, neither in general nor about its cultural facets (self-sacrifice for the group is valued differently across cultures). Here, we present a large-scale interactive decision making experiment (N = 870), performed in China and the USA, in which individuals had repeated cooperative interactions in groups. Our results show that (1) most individuals from both cultures overstate their contributions to the group if given the opportunity, (2) misrepresentation of cooperation is detrimental to cooperation in future interactions, and (3) the possibility to build up a personal reputation amplifies the effects of misrepresentation on cooperation in China, but not in the USA. Our results suggest that misrepresentation of cooperation is likely to be an important factor in (the evolution of) human social behaviour, with, depending on culture, diverging impacts on cooperation outcomes.

Figure 1

Overview of the experimental design. The experiment was run online and consisted of four experimental treatments: all combinations of the possibility for misrepresentation (with or without) and the possibility for reputation (with or without). Participants received an endowment of five Points and had to enter a contribution to the group project on the ‘Decision screen’. They entered either their actual contribution (treatments without misrepresentation, left side of figure in yellow) or both an actual contribution and a stated contribution (treatments with misrepresentation, right side of the figure in purple). On the ‘Results screen’, participants were shown the actual contributions (treatments without misrepresentation) or the stated contributions (treatments with misrepresentation) of the group members. In the treatments without reputation, individuals were only shown the total (stated) contributions of all group members, whereas they were shown all individual (stated) contributions in the treatments with reputation. In all treatments, the Results screen also showed the participant’s own earnings of the round, which consisted of an equal share of the total Points invested in the group project multiplied with m (which could take any value between 1.5 and 2.5, see “Methods”) and the Points the individual did not invest in the group project. After the Results screen, the participants started a new round on the Decision screen (in the same group), for a total of ten rounds.

Figure 2

Most participants across both cultures tend to overstate the contributions they made to the group project. Bars show the fractions of individuals that were ‘understaters’ (understating their contribution more often than they overstated them) ‘honest’ (individuals who understated and overstated equally often, which are mostly made up of individuals who never misrepresented at all) and ‘overstaters’ (overstating more often than understating). ‘Regular’ understaters are defined as individuals who understated in at least five more rounds than they overstated, while ‘casual’ understaters are individuals who understated in fewer than five more rounds than they overstated (the same reasoning applies to overstaters). The figure shows pooled data for the treatments with and without the possibility for reputation.

Figure 3

Misrepresentation undermines cooperation by weakening conditional cooperation and eliciting lower contributions. (A) If misrepresentation was possible, individuals were less responsive to the contributions of the other group members than in the absence of misrepresentation opportunities. Circles indicate how much individuals contributed on average depending on the average observed previous contribution of their fellow group members (i.e. stated contributions in case of the misrepresentation treatments). The radius of the circles indicate number of observations. The lines are given by a simple linear regression that included only misrepresentation, the previous average contribution of the fellow group members, and their interaction as predictors. (B) Increasing misrepresentation of fellow group members in the previous round led to decreasing contributions. The circles indicate the average contribution made depending on the average degree of misrepresentation of the other group members in the previous round. The radius of the circles indicate number of observations. The line is given by a simple linear regression that included only a natural cubic spline with two degrees of freedom in function of the previous average misrepresentation of the fellow group members as a predictor (this is consistent with how this variable was modelled in our overall statistical analysis, see “Methods”). Negative values indicate cases where individuals underrepresented their contributions, whereas positive values indicate overrepresentations.

Figure 4

Reputation effects resulted in large contribution differences between the treatments with and without misrepresentation in China, but not in the USA. Boxplots show median, interquartile range (box) and the 10% and 90% quantiles (whiskers). Red lines show the differences in mean contribution rates between the treatments with and without misrepresentation. ‘ns’ (P > 0.05) and ‘***’ (P < 0.001) indicate the significance of Tukey HSD based on a linear mixed model; see “Methods” for details.