From experience to evidence: How Shanghai schools are using AI to transform classroom analysis

Traditional classroom observation and analysis have long depended heavily on manual recording and experiential judgment, making the process time-consuming, subjective, and difficult to scale. A new case report published in the ECNU Review of Education reveals how primary and secondary schools in Shanghai are successfully overcoming these long-standing limitations by integrating the High-Quality Classroom Intelligent Analysis System into their routine instructional practices.

Authored by a research team including Chenlu Liu and Xin Zheng from East China Normal University, along with Bei Ding from Shanghai Jiangwan Middle School, the study details the practical application of this AI-based classroom analysis system. Unlike traditional methods, the platform automatically processes uploaded videos and analyzes multimodal data, including teacher–student dialogue, interaction structures, and time allocation. Within approximately 12 minutes, it translates raw classroom data into structured, visualized, and theory-informed analytic reports, making previously elusive interaction processes clearly visible.

The report presents evidence illustrating how schools use AI-generated reports to support instructional improvement across two distinct levels and four specific types. At the individual teacher level, educators use the system for "same teacher, optimized designs" to immediately compare the effectiveness of different instructional strategies for the same topic. In one instance of optimizing lesson designs, a mathematics teacher at Qibao High School taught the same topic twice in a single day. AI analysis revealed a distinct shift in classroom dynamics: The proportion of teacher-led instruction decreased from 54% in the first lesson to 34% in the second, while student self-directed activity increased from 22% to 41%. The findings demonstrate how specific design adjustments directly impact student agency.

Teachers also use the system to track professional growth over time through "same teacher, longitudinal improvement". For example, AI data collected over six years for a novice mathematics teacher showed teacher–student interaction increasing from 13% to 24%, and the proportion of open-ended questions rose from 1.92% to 11.11%. This continuous tracking provided longitudinal evidence of the teacher's effort to return more learning time to students, turning the abstract concept of student-centered teaching into an empirically grounded reality.

Beyond individual reflection, the AIC system is also transforming how teachers collaborate within China's Teaching and Research Groups (TRGs). The technology enables a more objective evaluation when teachers engage in "same lesson, different designs" to analyze how varying pedagogical approaches shape classroom structures. In a Physics TRG at Jiangwan Junior High School, two teachers taught the same "Linear Motion" lesson, and the AI report highlighted distinct but equally valuable instructional strengths. One teacher's design fostered logical reasoning with 19.3% open explanatory responses, while the other excelled in metacognitive feedback, accounting for over 21% of the evaluation phase. By making these differences visible, the system enables teachers to learn from and build on one another's instructional strengths.

This increased visibility allows groups to build consensus and integrate different strengths, ultimately driving the "collective advancement of a TRG". For instance, a Grade 5 Chinese-language TRG at Kongjiang No. 2 Primary School utilized AI reports to identify heavy teacher-led instruction. The group collaboratively introduced new strategies like role-play and mind mapping, reducing teacher talk from 69% to 55%. The group then extended these successful strategies across the entire unit and disseminated them to other grade levels, showcasing sustained, unit-level instructional optimization.

While AI provides timely and precise data that significantly deepen the understanding of student learning processes, the researchers issue a critical caution regarding its implementation in schools. The authors emphasize that the AI system should serve primarily as a collaborative partner or a supportive resource, rather than a rigid set of prescriptive indicators.

They warn that an overreliance on AI-generated metrics could inadvertently encourage "teaching to indicators," which risks weakening a teacher's independent professional judgment. As the report concludes, although artificial intelligence has the power to reshape the modes, processes, and overall efficiency of classroom analysis, the professional reflection, informed judgment, and collective wisdom of human teachers remain absolutely indispensable.