Self-interested cooperation in humans : an agent-based model of self-interest and an experimental investigation of cooperation willingness

Cooperation is a fundamental aspect of human behaviour, enabling us to achieve goals that would be impossible to accomplish individually. However, its success relies on the willingness of the group members to contribute to the collective effort. This thesis investigates self-interested cooperation in group decision-making tasks under diminishing reward functions. In these tasks, a few cooperators can significantly improve the group decision quality, generating rewards that outweigh the costs of cooperation. However, as more individuals cooperate, the marginal increase in group reward diminishes, which eventually causes the cooperation costs to exceed the benefits. This dynamic results in an equilibrium of cooperators and free-riders. In three agent-based simulation studies, we investigated factors affecting the equilibrium of cooperators and free-riders. Therefore, we conducted evolutionary simulations based on a group decision-making metaphor, where the group was given a set of alternatives, each representing a specific reward and had to choose one based on noisy cues that were assessed with decreasing validity. The first study focussed on environmental effects. We found that the cooperation rate was increased under lower costs, more alternatives to choose from, and when cues followed a non-compensatory structure. Our second study investigated the impact of group structure. Building on the simulation from Study 1, we introduced diversity in group members. Results showed that diversity in cognitive styles and information sources led to increased cooperation rates, whereas diversity in ability had no such effect. In the third study, we investigated how different costs in the acquisition of cue information would affect the cooperation rates and depth of information acquisition of cooperators. We designed the simulation with noisy cues with steeply increasing validities, where the first cue was the least valid and the cheapest. The setup incentivised the group to achieve the highest rewards when cooperators fully informed themselves. We found that high information costs hindered the sustainability of fully informed decision-making, especially when acquiring richer information was significantly more expensive. In two experimental studies, we investigated how humans follow the equilibrium predicted by self-interested cooperation. In Study 4, individual subjects played a game where they chose between cooperating and free-riding within a group of preprogrammed agents. Cooperation consistently increased the other group members’ rewards, while the choice of cooperating and free-riding offered varying rewards for the participants themselves under different conditions. The participants were explicitly informed of the reward for both options. Participants generally cooperated when it aligned with their self-interest. In the fifth study, groups of actual participants played a game where everyone had to choose between cooperation and free-riding. The reward had a diminishing increase with more cooperators. We found that the equilibrium between cooperators and free-riders affected group cooperation rates, showing that selfinterested cooperation plays a key role in group cooperation. This thesis provides a comprehensive theoretical and empirical investigation of self-interested cooperation in humans. Through simulations, we have shown that environmental and group-structural factors affect the equilibrium of cooperators and free-riders. Furthermore, we prove this equilibrium shapes human decision-making in group cooperation contexts.