AI赋能通识选修课全过程评价改革研究——以某工科院校的物联网概论课程为例

AI-Empowered Whole-Process Evaluation Reform in General Education Elective Courses—A Case Study of an Introduction to IoT Course at a University of Science and Technology

摘要: 在高等教育数字化转型的浪潮中，如何借助人工智能技术突破通识选修课长期面临的评价方式单一、标准固化、反馈迟滞、实践评价虚化等结构性困境，已成为教学改革的重要议题。文章以面向全校文、理、工多专业学生开设的通识选修课《物联网概论》为案例，系统剖析了传统评价体系的制度性缺陷与认识论根源，构建了AI驱动的全过程评价改革方案。该方案以“评价–诊断–反馈–改进”四维闭环模型(DEFI模型)为核心，整合了全过程多元评价、分层分类差异化评价、即时反馈与干预、虚拟仿真实践评价四大模块。通过在超星学习通平台的实践验证，改革后学生学习投入度提升32%，教学反馈周期从天级缩短至分钟级，课程满意度达91.6%，教师批改负担降低约70%。研究在肯定AI赋能评价改革成效的同时，亦对算法偏见、数据隐私、评价异化等潜在风险进行了批判性反思，并审慎界定了研究成果的适用边界——该模型目前已在工科院校线上通识课中验证有效，推广至其他场景时需依据课程属性、学生结构与技术条件进行适应性调整。研究为同类课程的数字化评价改革提供了可参考的理论框架与实践路径。

Abstract: In the context of the digital transformation of higher education, leveraging artificial intelligence to address the chronic challenges of general education elective courses—such as uniform evaluation methods, rigid criteria, delayed feedback, and the difficulty of assessing hands-on practice in online settings—has become an urgent pedagogical reform issue. Taking Introduction to the Internet of Things, a general education elective course open to students majoring in humanities, science, and engineering, as a case study, this paper systematically analyzes the structural deficiencies and epistemological roots of traditional evaluation systems and proposes an AI-driven whole-process assessment reform framework. The framework establishes a closed-loop four-dimensional evaluation model—“Diagnosis, Evaluation, Feedback, Improvement” (DEFI model)—which integrates four core modules: multidimensional whole-process evaluation, stratified and differentiated assessment, real-time feedback and intervention, and virtual simulation-based practical evaluation. Through implementation on the Chaoxing Xuexitong platform, the reform has led to a 32% increase in student learning engagement, reduced the feedback cycle from days to minutes, achieved a course satisfaction rate of 91.6%, and reduced teachers’ grading workload by approximately 70%. While affirming the effectiveness of AI-empowered evaluation reform, this paper also offers critical reflections on potential risks including algorithmic bias, data privacy, and evaluative alienation. It further delimits the applicable boundaries of the findings: the model has been validated in online general education courses at engineering-oriented institutions, and its extension to other contexts requires adaptive adjustments based on course attributes, student demographics, and technological conditions. This study provides a referable theoretical framework and practical pathway for the digital transformation of evaluation in similar courses.