VTutor for High-Impact Tutoring at Scale: Managing Engagement and Real-Time Multi-Screen Monitoring with P2P Connections

TL;DR

VTutor addresses the scalability gap in high-impact tutoring by uniting real-time multi-student screen sharing with animated pedagogical agents and a teacher-facing dashboard. The architecture comprises a VTutor Student Client, a Tutor Frontend Dashboard, and a Node.js backend, using WebRTC for peer-to-peer screen sharing and WebSocket signaling with a PostgreSQL datastore for logs. Animated avatars provide just-in-time, context-aware feedback, reducing tutor cognitive load while preserving interpersonal engagement. The approach offers a pathway to scalable, data-informed tutoring in hybrid classrooms, with future work focusing on bandwidth optimization, avatar expressiveness, advanced learner-state detection, and longitudinal outcomes.

Abstract

Hybrid tutoring, where a human tutor supports multiple students in learning with educational technology, is an increasingly common application to deliver high-impact tutoring at scale. However, past hybrid tutoring applications are limited in guiding tutor attention to students that require support. Specifically, existing conferencing tools, commonly used in hybrid tutoring, do not allow tutors to monitor multiple students' screens while directly communicating and attending to multiple students simultaneously. To address this issue, this paper introduces VTutor, a web-based platform leveraging peer-to-peer screen sharing and virtual avatars to deliver real-time, context-aware tutoring feedback at scale. By integrating a multi-student monitoring dashboard with AI-powered avatar prompts, VTutor empowers a single educator or tutor to rapidly detect off-task or struggling students and intervene proactively, thus enhancing the benefits of one-on-one interactions in classroom contexts with several students. Drawing on insight from the learning sciences and past research on animated pedagogical agents, we demonstrate how stylized avatars can potentially sustain student engagement while accommodating varying infrastructure constraints. Finally, we address open questions on refining large-scale, AI-driven tutoring solutions for improved learner outcomes, and how VTutor could help interpret real-time learner interactions to support remote tutors at scale. The VTutor platform can be accessed at https://ls2025.vtutor.ai. The system demo video is at https://ls2025.vtutor.ai/video.

Figures (3)

• Figure 1: VTutor System Architecture. The VTutor Student Client (left) shows how learners interact with educational technologies or learning materials (e.g., IXL) alongside a virtual agent. A chat interface enables conversation with remote tutors, whose utterances can be spoken aloud by the avatar. Tutors use a Tutor Frontend Dashboard (top-right) to view multiple student screens simultaneously and send real-time prompts through the avatar. Communication channels include WebSocket connections with backend (for chat, logs, events) and WebRTC peer-to-peer connections (for screen sharing). The VTutor Backend Node.js Server (bottom-right) handles user authentication, session management, and logging with a PostgreSQL database.
• Figure 2: Tutor Dashboard and Messaging Interface for individual student. The tutor's view shows an "Anonymous" student currently screen-sharing IXL, where the student attempts a math problem. On the right, the tutor and student exchange messages in real-time; any tutor messages sent here are spoken aloud by the VTutor avatar on the student's screen. The lower panels provide status information (e.g., last tutor interaction), letting the tutor quickly detect off-task behavior and intervene with targeted guidance.
• Figure 3: Student Interface During Tutoring Session. The student is solving an algebraic equation on IXL ("Solve for $k$"), while the VTutor avatar, an animated panda agent, waves and offers animated guidance in the lower-right corner. Students can also chat directly with tutors; messages from tutors is spoken aloud by the VTutor avatar.