A Tripartite Evolutionary Game of Green Agricultural Technology Diffusion: Conservative Oscillation, Critical Thresholds, and State-Dependent Policy

Understanding when coordinated cooperation emerges in multi-agent systems, and when it fails, is a central question in the study of complex socio-economic dynamics. This paper develops a tripartite evolutionary game among three boundedly rational populations—local governments, agribusinesses, and smallholder farmers—and applies it to the diffusion of green agricultural technology, using conservation tillage in Northeast China’s black soil region as a concrete setting. The model incorporates a performance-based subsidy that is disbursed only upon realized adoption, thereby foreclosing subsidy capture. We derive the replicator dynamics, analyze the asymptotic stability of all eight pure-strategy equilibria, and, for the regime in which the system fails to converge, construct a first integral analytically and verify its conservation numerically to machine precision. Four findings emerge. First, the intuitively ideal fully coordinated state is dynamically unstable; the stable desirable outcome is one in which the government withdraws while the market sustains adoption. Second, under low initial willingness the system does not converge to any evolutionarily stable strategy but enters a sustained, conservative (center-type) oscillation—a policy-reversal trap that transient interventions cannot extinguish, placing the dynamics within the class of Hamiltonian-like evolutionary systems. Third, a sharp critical threshold in initial willingness partitions the state space into distinct basins of attraction, rendering policy effectiveness state-dependent. Fourth, within the modeled trap, credible penalties and enhanced agribusiness service capacity can engineer escape, whereas untargeted farmer subsidies are markedly less effective and may even deepen the oscillation. These results recommend treating public intervention as a transitional, regime-aware catalyst with a built-in exit. Beyond the agricultural setting, the analysis illustrates how conservative oscillations and multiple basins shape the controllability of multi-agent cooperation systems.