Partner choice and cooperation in social dilemmas can increase resource inequality

Cooperation is crucial for societal well-being, but social dilemmas, where individual self-interest conflicts with collective good, frequently impede it. The ability to choose interaction partners is known to promote cooperation, allowing cooperative individuals to avoid exploitation by defectors. Existing research largely assumes equal capabilities for cooperation among individuals. However, this study investigates the effect of partner choice on cooperation within unequal societies where individuals possess varying resources and productivity. This inequality may skew partner selection, leading high-resource individuals to choose similar partners, leaving low-resource individuals to interact among themselves. This could lead to a vicious cycle where the rich get richer and the poor get poorer. The hypothesis is that in societies characterized by pre-existing resource inequality, partner choice, rather than fostering cooperation, will amplify those inequalities. This study employs an experimental design to test this hypothesis, using a repeated public goods game to model this social dilemma. The context is important as it explores the interplay between individual agency (partner choice) and structural constraints (resource inequality) in shaping cooperation and ultimately resource distribution.

The literature extensively documents the importance of cooperation in social dilemmas for group and societal functioning. A lack of cooperation is linked to negative outcomes like group dissolution, polarization, and conflict. Past research demonstrates that allowing individuals to choose their interaction partners enhances cooperation. Partner choice enables cooperative individuals to self-select into groups, thus avoiding exploitation by defectors and creating a positive feedback loop that incentivizes defectors to cooperate. However, most studies focused on homogeneous groups. A growing body of work now addresses the impact of inequality on cooperation in social networks, suggesting that inequality can undermine cooperation. However, the impact of partner choice on cooperation remains an open question in contexts of resource inequality. This research directly addresses this gap, by explicitly incorporating variations in individual resource endowment and productivity levels, factors that influence partner attractiveness and cooperation outcomes.

This study uses a 2 (condition: partner choice vs. assigned partner) × 4 (type: HH, HL, LH, LL) between-subjects experimental design. 336 participants were recruited online via Leiden University and Prolific, forming 42 groups of eight participants each. Each group included two individuals of each of the four types, which were determined by a combination of randomly assigned high or low endowments and productivity factors (HH = High Endowment/High Productivity, HL = High Endowment/Low Productivity, LH = Low Endowment/High Productivity, LL = Low Endowment/Low Productivity). Participants played 24 rounds of a two-person public goods game. In each round, participants simultaneously decided how much of their endowment to contribute to the public good. Contributions were multiplied by the individual's productivity factor and distributed equally within the pair. In the assigned partner condition, partners were pseudo-randomly assigned to ensure equal interaction with each type. In the partner choice condition, participants ranked partner types from most to least preferred each round, and pairing was based on mutual preference. The study collected data on partner preferences, cooperation rates, and resource accumulation. The primary analyses involved multilevel (logistic) models to address the nested structure of the data, accounting for participants nested within groups. Statistical analyses were performed using the lme4 package in R, testing for differences in cooperation rates, partner preferences, and resource accumulation across conditions and participant types, controlling for various factors like round number and participant type.

The study found significant segregation under partner choice. Participants were paired with their first choice in 45% of the rounds, leading to a higher frequency of same-type pairs compared to the assigned partner condition (MLLM, z = −7.56, p < 0.001). Same-type pairs were also more stable, interacting for more consecutive rounds (MLM, t(292) = 12.14, p < 0.001). The most preferred partner type was the HH type (65.1% of first choices), while the LL type was the most frequently rejected (75% unable to pair with their first choice). This led to HH types being paired together in 73% of rounds in the partner choice condition, and LL types in 66.9%. This clustering of the extremes resulted in segregation and unequal success rates in interacting with one's preferred partner, disadvantaging LL types. Partner preferences evolved over time, with non-HH types increasingly preferring same-type pairings in later rounds (MLLM, z = −9.24, p < 0.001, for preference for HH types decreasing over time). Although partner choice led to higher overall cooperation, it was not equal. Cooperation levels depended on partner type; participants cooperated more with HH partners and less with LL partners. This disparity in cooperation contributed to increased resource inequality. HH types accumulated significantly more resources in the partner choice condition than in the assigned partner condition (MLM, z = 49.09, p < 0.001), while LL types accumulated significantly less (MLM, z = −41.05, p < 0.001). The Gini index revealed a steeper skew in resource distribution under partner choice (0.44) compared to the assigned partner condition (0.24), demonstrating partner choice amplified pre-existing resource disparities.

This study's findings challenge the notion that partner choice invariably promotes cooperation. While partner choice did lead to higher overall cooperation, the benefits were not equally distributed. In an unequal society, the freedom to choose partners exacerbated existing resource inequalities. The observed segregation, driven by a preference for high-resource partners and the consequent exclusion of low-resource individuals, created a self-reinforcing cycle. This highlights how individual-level choices can have significant macro-level consequences, impacting the distribution of resources and the stability of cooperation. These findings resonate with other studies showing that inequality can alter social network structures and cooperation patterns. Although this study controlled for anonymity of cooperation choices, only observing general characteristics, future research could incorporate observed cooperation behavior to better understand how cooperation behavior might compensate for structural disadvantages. The results suggest that mechanisms promoting equal opportunity and resource redistribution might be necessary to prevent partner choice from worsening inequality.

This study demonstrates that partner choice in social dilemmas can amplify existing resource inequality, particularly in contexts where individuals differ significantly in endowment and productivity. While partner choice enhances overall cooperation, the benefits are unequally distributed, leading to a self-reinforcing cycle where the advantaged accumulate more resources. Future research should investigate how incorporating information on past behavior, incentive structures, and different network configurations might modify these findings. Further explorations into interventions that mitigate the negative effects of partner choice on inequality are warranted.

The study used an artificial environment to model a complex societal phenomenon. The specific parameters of the public goods game, the sample composition, and the online recruitment method might limit the generalizability of the findings. The sample sizes of the groups recruited through different platforms could also have impacted the findings. Further research is needed to determine the extent to which these findings generalize to other contexts, and whether the effects are modified by variations in incentive structures or information available to participants.