Effects of population polarization and information availability on group cooperation: agent-based modelling of the weakest-link game

Authors

  • Evan Koblentz Center for Social Complexity, Department of Computational and Data Sciences, George Mason University, Fairfax, VA
  • William Kennedy Center for Social Complexity, Department of Computational and Data Sciences, George Mason University, Fairfax, VA

DOI:

https://doi.org/10.13021/jssr2026.5660

Abstract

People often fail to cooperate despite their best interest. For instance, voters may knowingly oppose a candidate whose positions best align with their interests after members of their social network suggest the candidate is untrustworthy. This project applies game theory to examine the roles of polarized trust structures and information availability on group cooperation. Using NetLogo to simulate the weakest-link game, agents form groups and then decide whether to cooperate with, or defect from, their group. If all agents in a group cooperate, each receives the maximum payoff. If one or more defect, cooperating agents receive no payoff and defectors a small payoff. Agents’ uncertainty centers on not knowing whether all others in their group can be trusted to cooperate to deliver the highest payoff. Nine agent-based models were created that varied in two aspects: how trust in other agents to cooperate is structured and how information is made available regarding agents’ prior cooperation decisions. Each model was run through 100 simulations (50 agents and 25 rounds per run). Mean number of rounds until all agents cooperate, percent of groups that fully cooperate over all rounds, and average payoff accumulated by agents were compared across models using Kruzkal-Wallis tests. Results suggest that trust polarization tied to a subpopulation’s identity is more harmful to both group cooperation and individual payoff than polarization based around a single agent. Findings also suggest cooperation is weakest when information is obtained through word of mouth rather than personal experience or common knowledge.

Published

2026-09-24

Issue

Section

College of Science: Department of Computational and Data Sciences