Multi-agent system improves feedback from online class videos
MAVE: A Multi-Agent Framework for Grounded Pedagogical Feedback from Online Class Videos
Computers and Society
Summary
Providing useful and timely feedback to teachers based on student behavior in online class videos is challenging because videos are long and have many distractions. The authors designed a system called MAVE that uses four specialized roles to watch classroom videos, understand what students are doing, identify gaps in learning, and suggest teaching tips. They created a test set called MAVEBench with many video clips from real STEM and Humanities classes to check how well the system works. Their experiments show MAVE gives better feedback that matches what’s happening in class and offers practical advice compared to other methods.
What this means in practice
- •For online education platforms: Automate tailored teaching advice by analyzing student behavior from online class videos to improve instructor guidance.$Commercial implications: This enables educational platforms to offer personalized, scalable instructor feedback services that enhance course effectiveness and user satisfaction.
- •For education technology developers: Build enhanced tools that interpret classroom video data to generate actionable feedback supporting teacher development and student engagement.
Authors
Qingyue Chen, Huaiyuan Yao, Wanpeng Xu, Varun Shourie, Qianwen Wang, Hua Wei
Abstract
Timely pedagogical feedback is difficult to scale in online classes, where long and noisy videos make student responses hard to locate and automated systems often produce generic advice weakly grounded in actual classroom behavior. The key challenge is to connect the instructional goal of a specific lesson segment with temporally aligned multimodal evidence of student responses, and then translate the observed gap into an actionable teaching move. We introduce MAVE (Multi-Agents for Video-based Educational Feedback), a multi-agent framework for grounded pedagogical feedback from online-class videos. MAVE decomposes feedback generation into four roles: the Curator structures the lesson and infers expected student activity; the Observer extracts aligned speech, visual, and interaction evidence; the Assessor estimates the activity supported by student behavior; and the Advisor generates gap-conditioned teaching recommendations. To evaluate this capability, we introduce MAVEBench, a benchmark designed to measure whether feedback systems can ground diagnoses in multimodal student evidence, and generate actionable pedagogical feedback. MAVEBench contains 173 knowledge-point segments from approximately 20 hours of authentic online-class videos across STEM and Humanities, together with comprehensive simulation. Experiments show that MAVE outperforms direct prompting, Reflexion, and TextGrad in diagnostic correctness, instructional alignment, and actionability. Our code is available at https://github.com/DaRL-GenAI/MAVE