Generative AI Empowering the Reform and Practice

Abstract

Technical characteristics of generative AI (large-scale pre-trained models), including natural language understanding and generation, code semantic modeling, contextual reasoning and multi-turn interaction, match well with the functions of teaching evaluation tasks like generating tasks, collecting evidence, interpreting performances, giving feedbacks and adjusting. To solve some problems existed in evaluating the Python Programming course, e.g., limitation of summative measurement, fragmentation of process evidence, delay of feedback adjustment, lack of interpretability of judgment, we build a logical framework for generative AI-empowered diversified teaching evaluation by mapping the technical characteristic-evaluation function, course competency objective-evaluation dimension and empowerment mechanism-structural elements. Construct an evaluation system from three aspects: multi-dimensional evaluation indicators oriented towards progressive programming competency; Process evaluation data generation and semantic analysis; Human-AI collaboration intelligent feedback and dynamic adjustment. Explore on how to reconfigure roles and allocate tasks when embedding generative AI within the evaluation process, how to intelligently interpret evaluation results and transform it into teaching decisions, quality assurance and boundary regulation. Our research provides an interpretable, traceable and adjustable support path for the evaluation of the Python Programming course.