人工智能驱动计算机类专业过程性考核的实践与反思

Practice and Reflection on AI-Driven Process-Oriented Assessment for Computer Science Majors

摘要: 计算机类专业课程的过程性考核仍存在评价维度单一、反馈滞后和个性化支持不足等问题。基于教学改革实践，本文将大语言模型辅助评阅、课程知识图谱和认知诊断引入过程性考核，构建智能考核系统。系统对主观题和代码作答进行分维度辅助评阅，并结合Q矩阵与G-DINA模型形成知识点掌握诊断，据此提供反馈与学习支持。研究以一门专业核心课程开展准实验，比较实验组与对照组在评阅效率、学习表现和反馈利用等方面的变化，并结合访谈与课堂观察进行分析。结果显示，该机制有助于缩短反馈周期，促进学生依据反馈改进任务成果，也为教师识别共性问题和调整教学提供依据。大语言模型仅承担辅助初评和诊断支持，关键评价由教师复核确认。

Abstract: Formative assessment in computer science courses still faces several challenges, including limited evaluation dimensions, delayed feedback, and insufficient personalized support. Drawing on teaching reform practices, this study integrates large language model-assisted assessment, a course knowledge graph, and cognitive diagnosis into formative assessment, and develops an intelligent assessment system. The system conducts multidimensional auxiliary assessment of open-ended responses and programming submissions, and combines a Q-matrix with the G-DINA model to generate diagnostic information on students’ mastery of knowledge points, which is then used to support feedback and learning. A quasi-experimental study was conducted in a core professional course to compare changes in assessment efficiency, learning performance, and feedback use between an experimental group and a comparison group, supplemented by interviews and classroom observations. The results indicate that the proposed mechanism helps shorten the feedback cycle, encourages students to improve their work based on feedback, and provides evidence for instructors to identify common learning difficulties and adjust teaching. The large language model is used only for auxiliary preliminary assessment and diagnostic support, while key evaluative judgments are reviewed and confirmed by instructors.