Abstract: While artificial intelligence ( AI) textbooks offer the advantage of enabling largescale personalized teaching, they face the new challenge of difficulty in visualizing massive learning data. This study systematically analyzed the learning behaviors of 436 students using the AI textbookMedical Biochemistry, addressing three critical challenges in data visualization: ① visualizing interclass variations across parallel large classes; ② analyzing the fine-grained subgroup learning profiles within a parallel large class; ③ profiling the in-depth thinking reflected in AI interactions within a parallel large class. In this research, seven key parameters were used to demonstrate the differences in learning characteristics between parallel large classes. Within one parallel large class, students were subdivided into 9 subgroups based on behavioral heterogeneity, with personalized differences analyzed across subgroups. Using a self-developed scaffold questioning framework for AI interactions, cluster analysis was conducted on all interaction questions between students and AI in one parallel large class, enabling visual evaluation of students’in-depth thinking effects. Future research should develop an automatic display for AI textbook data visualization based on these strategies to support data-driven decision-making in large-scale personalized teaching.

Key words: