Impact of interaction diversity and interlayer coupling on the evolution of human-AI cooperation

In the context of a digitized society characterized by profound human-artificial intelligence (AI) interaction, the evolutionary dynamics of cooperative behavior confront novel challenges. This study develops a two-layer network evolutionary game model that couples a human decision-making layer with an AI system layer, aiming to investigate how bidirectional interlayer coupling mechanisms influence the emergence and sustainability of cooperation. We find that interlayer coupling facilitates the emergence and stabilization of cooperative behavior across both the AI and human layers. However, the efficacy of this facilitative effect is highly contingent upon the relative dependency configurations between two networks. When interaction diversity (link-strategy) is introduced within the AI layer, the system’s resilience to the temptation of defection is markedly enhanced, resulting in elevated levels of cooperation. Conversely, excessive interaction diversity within the human layer may, under certain conditions, undermine cooperative coordination, particularly in scenarios characterized by high interlayer dependency. Moreover, the AI layer exhibits substantial adaptive capacity, maintaining relatively stable behavioral patterns across varying human decision-making regimes, thereby enabling the emergence of a robust mixed equilibrium of cooperation and defection. This study proposes a computational model to explore the evolutionary dynamics of cooperation in human-AI collaborative environments, offering new insights into AI governance, multi-agent system design, and related domains.