"Responsibility Trap" in the Integration of AI into Higher Education Governance: Triggers and Mitigation

The integration of generative AI into higher education governance not only enhances governance efficacy via data intelligence, but also engenders a "Responsibility Trap" characterized by four dimensions: the algorithmic dissolution of responsibility subject, the technical deviation of responsibility standards, the black-box obstruction to responsibility traceability, and the distributed myth of responsibility allocation. This governance phenomenon stems from multiple factors: cognitive biases and capacity deficits among governance subjects, the quasi-autonomy and quantification-oriented bias of AI technologies, gaps and ambiguities in governance regimes, and departmental fragmentation alongside power-responsibility imbalances in governance structures. To solve these problems, interventions should be implemented across four domains: designing a capacity-building system to elevate the competence of governance subject; developing a transparent and controllable technical framework for educational AI; refining a robust normative system for traceability and accountability; and constructing a new governance structure featuring coordination, integration, and binding. These measures will enable the full realization of AI's technological potential while facilitating the modern transformation of higher education governance.