Research on the controllability of generative AI educational resource generation

Abstract

Controllable risks such as content distortion, ethical bias and algorithm black box of generative AI (GAI) educational resources threaten the reliability of digital education. In this study, a three-dimensional dynamic risk assessment model (technology/content/application dimensions, including 18 detailed indicators) was constructed, and the following strategies were proposed: (1).Tech-nical optimization: improve output stability through blockchain traceability and knowledge graph constraint generation; (2) Multiagent governance: Integrate a collaborative framework for policy compliance (e.g., China’s Interim Measures for the Management of Generative AI Services), institutional review, and AI ethics training. The model has been validated in K12 and higher education scenarios and can effectively identify high-risk events. The research results provide a quantifiable path for the application of GAI in education security, balancing innovation and regulation. In the future, it will be extended to transnational copyright governance.