Research on the Construction of AI-Empowered College Physics Experiment Courses

Abstract

Driven jointly by the national strategic initiative on educational digitalization and the "Artificial Intelligence+" action plan, physics experiment teaching is undergoing a critical transition from traditional hands‐on operations to intelligent inquiry‐based learning. However, current research on AI‐empowered experimental teaching largely relies on expensive specialized equipment or cloud‐based computing resources, which entails high deployment costs, significant data privacy risks, and limited feasibility for widespread adoption in ordinary colleges and universities. Grounded in the principles of lightweight design, low cost, and high availability, this study systematically develops a theoretical framework and practical pathways for constructing AI‐enhanced university physics experiment courses. The proposed approach centers on open‐source large language models and retrieval‐augmented generation (RAG) techniques to build a domain‐specific knowledge system for physics experiments, and achieves data security and cost control through local deployment. We establish a tripartite construction pathway encompassing knowledge base development, vertical AI model customization for the course, and intelligent teaching applications. Furthermore, we design a human‐computer collaborative instruction model that integrates AI‐assisted preview, instructor‐led focused instruction, and a two‐stage assessment comprising AI initial evaluation followed by instructor verification, thereby forming a closed‐loop intelligent teaching system covering the entire spectrum of instruction, learning, administration, evaluation, and assessment. The findings indicate that this model effectively mitigates the hallucination issue inherent in general‐purpose large models when applied to specialized domains, improves the quality of students' experimental reports, and enhances their autonomous inquiry capabilities. It also offers a replicable, cost‐effective reform blueprint for resource‐constrained local institutions.