New research from East China Normal University sheds light on classroom dynamics using large-scale AI analysis

Classroom teaching plays a crucial role in improving students’ learning, building interpersonal relationships, peer engagement, and provides foundation for future academic success. Analyzing classroom teaching patterns can help create an environment conducive to learning, promote diverse learning styles, and enhance student learning outcomes. This involves the collection of substantial quantity of classroom videos and evaluating them manually. However, manual annotation consumes more time and human resources.

Against this backdrop, Yihe Gao and Xiaozhe Yang from East China Normal University have now utilized AI to revolutionize classroom analysis. Their study, published online on March 24, 2025 in ECNU Review of Education, introduces the High-Quality Classroom Intelligent Analysis Standard (CEED) system. “This AI-driven system applies machine learning and multimodal data processing techniques to classify discourse, analyze behavior, and evaluate teacher-student interactions. By automating large-scale classroom video analysis, the CEED system marks a significant advancement over traditional manual observation methods, which are labor-intensive and time-consuming,” explains Gao.

The CEED system is based on the data handling ability of AI throughout various aspects such as speech, behavior, and psychology to evaluate classroom conditions, with emphasis on students’ learning scenarios. This multi-dimensional approach can provide a comprehensive understanding of the high-quality standards of classroom efficiency, equity, and democracy.

The research team collected classroom videos of grade 1 to grade 9 with Chinese language as medium of instruction. By analyzing 1,008 primary and secondary school classes, the researchers reveal that teacher-centered instruction prevails. On average, teacher presentation occupied 51.9% of classroom time and teacher-student interaction accounted for 30.5%. Additionally, individual tasks and group activities comprised 12.3% and 5.3%, respectively.

These findings confirm that teacher-led instruction remains predominant, particularly in higher grade levels, where open-ended questioning decreases significantly. The research challenges common assumptions that older students engage in more critical discussions, revealing instead that teachers tend to favor structured, closed-ended questions.

The researchers further examined the pattern of classroom time span spent across different grades and found that first grade had the lowest proportion of time during group activities, lower than the third grade, fourth grade, sixth grade, and seventh grade. Also, the amount of time spent on teaching increases as the grade level increases. Conversely, the proportions of teacher-student interactions, group activities, and individual tasks decrease as the grade level increases. Furthermore, teachers in higher grades tend to engage in lecturing as an instructional approach.

The study underscores the potential of AI in transforming classroom evaluation and teacher training, emphasizing the implications on education policy and practice. Educators can use the CEED system to analyze their teaching methods, assess classroom engagement, and refine instructional strategies based on data-driven insights. Assessing the indicators such as teacher-student interactions, teacher presentation, group activities, and individual tasks can assist in improving the quality of teaching. These indicators can help in designing training programs to streamline teaching strategies.

"Our study not only provides a snapshot of current teaching practices, but also provides educators with a tool to facilitate instructional adjustments and enhance classroom interactions," Yang noted. "AI-driven analytics open up new avenues to increase the efficiency of instructional improvement and foster more student-centered learning environments."

While the AI system has proven highly effective and accurate in analyzing language-based subjects, challenges remain in applying it to fields like physical education. Further research is needed to refine AI’s capabilities in assessing diverse classroom activities across subjects. Additionally, concerns around algorithmic bias and data interpretation highlight the need for careful implementation in educational settings.

This study represents a significant step toward integrating AI in education research and practice, offering a scalable, data-driven approach to understanding and improving classroom dynamics.
Reference
Titles of original paper: Are China’s Classes Predominantly Centered Around Teacher-Presentation Instruction?—A Large-Scale Data Analysis Based on Classroom Intelligent Analysis Systems
Journal: ECNU Review of Education
DOI: 10.1177/20965311251322181