Ongoing debates in higher education regarding whether artificial intelligence should be further integrated or deliberately constrained call for empirical research that offers a more explanatory analytical framework. However, existing studies on the human-AI collaboration (HAC) paradox are largely grounded in a binary logic of facilitation versus inhibition, leaving the dynamic mechanisms underlying complex cognitive processes insufficiently explored. To address this issue, this study adopts a di