The structure of multiplex networks predicts play in economic games and real-world cooperation

TL;DR

This study shows that the structure of multiplex social networks, specifically the overlap of ties across domains, predicts cooperative behavior in both experimental and real-world settings across different communities.

Contribution

It introduces the use of multiplex network overlap as a predictor of cooperation, linking social structure to cooperative behavior at individual and community levels.

Findings

Higher multiplex overlap predicts increased cooperation.

Individual and community overlap influence cooperative tendencies.

Multiplex network structure explains inconsistencies in cooperation studies.

Abstract

Explaining why humans cooperate in anonymous contexts is a major goal of human behavioral ecology, cultural evolution, and related fields. What predicts cooperation in anonymous contexts is inconsistent across populations, levels of analysis, and games. For instance, market integration is a key predictor across ethnolinguistic groups but has inconsistent predictive power at the individual level. We adapt an idea from 19th-century sociology: people in societies with greater overlap in ties across domains among community members (Durkheim's "mechanical" solidarity) will cooperate more with their network partners and less in anonymous contexts than people in societies with less overlap ("organic" solidarity). This hypothesis, which can be tested at the individual and community level, assumes that these two types of societies differ in the importance of keeping existing relationships as opposed to recruiting new partners. Using multiplex networks, we test this idea by comparing cooperative tendencies in both anonymous experimental games and real-life communal labor tasks across 9 Makushi villages in Guyana that vary in the degree of within-village overlap. Average overlap in a village predicts both real-world cooperative and anonymous interactions in the predicted direction; individual overlap also has effects in the expected direction. These results reveal a consistent patterning of cooperative tendencies at both individual and local levels and contribute to the debate over the emergence of norms for cooperation among humans. Multiplex overlap can help us understand inconsistencies in previous studies of cooperation in anonymous contexts and is an unexplored dimension with explanatory power at multiple levels of analysis.