AI-Based Evaluation of Teacher Ethics in Higher Education Using Classroom Text Data

With the rapid development of educational digitalization and artificial intelligence, data-driven approaches to teacher evaluation have become an important research focus. To address the limitations of traditional evaluations of teacher ethics and professional conduct, which often rely on subjective judgment and lack process-based evidence, this study proposes an AI-assisted evaluation framework based on classroom text data. Classroom teaching videos from multiple disciplines were transcribed into text and evaluated across six dimensions: political stance and values, teaching and nurturing, professional competence, role modeling, social responsibility, and professional commitment. Human ratings were used as the criterion reference, and AI-generated scores were evaluated using correlation and error indicators. The results indicate that AI-generated scores were positively correlated with human ratings, with higher consistency in dimensions more directly reflected through classroom language and observable teaching behaviors. Larger discrepancies remained in dimensions involving value judgment, implicit attitudes, and normative expectations. These findings suggest that classroom text can support AI-assisted screening and preliminary analysis of teacher ethics and professional conduct, but it should not be interpreted as a sufficient basis for fully automated or high-stakes evaluation. The study provides preliminary empirical evidence for process-based and data-driven teacher evaluation and clarifies the scope and limitations of applying large language models to teacher ethics assessment.