Maintenance and collapse of cooperation driven by bounded rationality in structured populations

The maintenance of long-term stability in human cooperation presents fundamental challenges from self-interested behaviors in social dilemmas. We construct a micro-dynamic model unifying reinforcement learning and Prospect Theory to investigate the evolution of human cooperation driven by bounded rationality. Through Monte Carlo simulations on a square lattice, we explore the emergence, maintenance, and collapse of cooperation in public goods games, with emphasis on psychological reference points and initial conditions. The system converges to a high-cooperation static equilibrium when cooperators exhibit lower greediness than defectors. Conversely, reversed greediness asymmetry triggers cascading collapse via localized clusters comprising low-probability cooperators. Supercritical initial cooperation enables defector-resistant static equilibrium, whereas subcritical values drive the system toward low-cooperation dynamic equilibrium. The framework establishes predictive criteria for cooperation sustainability versus collapse, offering actionable insights for public governance and collective risk mitigation. • Prospect Theory and reinforcement learning unified in a bounded rationality model. • Low cooperation greediness and high defection greediness promote cooperation stability. • Low-probability cooperation clusters trigger cascading collapse of global cooperation.