Q-learning driven adaptive decision rules and environmental transformation in three strategy evolutionary games

Cyclic dominance is often viewed as a robust mechanism for sustaining diversity and cooperation in evolutionary systems. In this work, it is shown that the persistence of such dynamics is far more fragile than commonly assumed when adaptive behavior and environmental feedback are explicitly accounted for. An eco-evolutionary game is examined in which, alongside cooperators and defectors, a third class of agents alters the shared environment rather than engaging in direct social interactions. Through this environmental transformation, the conditions under which all strategies compete are continuously reshaped. The results demonstrate that environmental modification generates strong feedback effects that can either promote or undermine cooperation, depending on the nature of the feedback. It is further revealed that evolutionary outcomes depend on how agents adapt. In imitation-based adaptation, cyclic dominance is maintained, and persistent spatial organizations support cooperation. In contrast, when adaptation occurs through reinforcement learning, cyclic dominance is destabilized, long-lived transients emerge, and the system converges toward learning-dependent dominance or mixed evolutionary states that are not inferable from payoff structure alone. Most notably, when environmental modifiers are allowed to adapt to their ecological impact, environmental feedback itself becomes an evolving property of the system. Under these conditions, cooperation, dominance, or coexistence arise from the coupled evolution of behavior, learning, and environmental change rather than from static strategic interactions. These findings demonstrate that cyclic dominance cannot be regarded as a universal stabilizing principle. Instead, long-term cooperation emerges from the dynamic interplay among behavioral plasticity, experience-based decision-making, spatial structure, and evolving environmental feedback, offering a deeper understanding of cooperation in complex adaptive systems.