**ClassMeta: Designing Interactive Virtual Classmate to Promote VR Classroom Participation** **Abstract:** Peer influence is crucial for promoting classroom participation, where active student behaviors can enhance the collective learning experience. However, the availability of active students varies and is not guaranteed. Inspired by large language models (LLMs) like GPT, which can simulate diverse human behaviors, we designed ClassMeta, a GPT-4-powered agent to promote classroom participation in virtual reality (VR) classrooms. ClassMeta interacts with instructors and students using spoken language and body gestures, exhibiting behaviors such as taking notes, responding to instructors, correcting distracted students, and participating in discussions. We conducted a comparative study to evaluate the effectiveness of ClassMeta in improving the overall learning experience. **Key Contributions:** - A novel approach to promoting classroom participation through virtual agents. - An interaction design for virtual agents that includes proactive engagement with both students and instructors. - A between-group comparative user study to evaluate the effectiveness of the interaction design. - A template for tuning the agent's behavior using GPT, facilitating future educators to implement customized agents. **System Overview:** ClassMeta is developed using the Unity3D platform and runs on the Oculus Quest Pro headset. The VR classroom setup includes six student seats, with two ClassMeta agents positioned to ensure visibility. Real-time conversational context is collected using Microsoft Azure SDK and uploaded to Google Cloud Firestore. The system integrates lecture slides and uses eye-tracking to track student attention. GPT-4 is integrated to generate dynamic responses based on predefined prompts, enabling actions like note-taking, discipline reminders, and question-answering. **User Study:** We conducted a user study with 24 participants, split into two conditions: a baseline VR classroom with only human instructors and students, and a ClassMeta session with GPT-empowered agents. The study measured learning gains in conceptual knowledge and logical thinking skills through pre-post tests. Eye-tracking data and note quality were also evaluated. Results showed that students paid more attention to ClassMeta agents, and note-taking quality improved significantly. The ClassMeta agents reduced the need for instructor interventions, particularly in question-raising and discussion phases. **Conclusion:** ClassMeta effectively promotes classroom participation in VR environments by leveraging the capabilities of LLMs. The study demonstrates the potential of virtual agents to enhance student engagement and learning outcomes.**ClassMeta: Designing Interactive Virtual Classmate to Promote VR Classroom Participation** **Abstract:** Peer influence is crucial for promoting classroom participation, where active student behaviors can enhance the collective learning experience. However, the availability of active students varies and is not guaranteed. Inspired by large language models (LLMs) like GPT, which can simulate diverse human behaviors, we designed ClassMeta, a GPT-4-powered agent to promote classroom participation in virtual reality (VR) classrooms. ClassMeta interacts with instructors and students using spoken language and body gestures, exhibiting behaviors such as taking notes, responding to instructors, correcting distracted students, and participating in discussions. We conducted a comparative study to evaluate the effectiveness of ClassMeta in improving the overall learning experience. **Key Contributions:** - A novel approach to promoting classroom participation through virtual agents. - An interaction design for virtual agents that includes proactive engagement with both students and instructors. - A between-group comparative user study to evaluate the effectiveness of the interaction design. - A template for tuning the agent's behavior using GPT, facilitating future educators to implement customized agents. **System Overview:** ClassMeta is developed using the Unity3D platform and runs on the Oculus Quest Pro headset. The VR classroom setup includes six student seats, with two ClassMeta agents positioned to ensure visibility. Real-time conversational context is collected using Microsoft Azure SDK and uploaded to Google Cloud Firestore. The system integrates lecture slides and uses eye-tracking to track student attention. GPT-4 is integrated to generate dynamic responses based on predefined prompts, enabling actions like note-taking, discipline reminders, and question-answering. **User Study:** We conducted a user study with 24 participants, split into two conditions: a baseline VR classroom with only human instructors and students, and a ClassMeta session with GPT-empowered agents. The study measured learning gains in conceptual knowledge and logical thinking skills through pre-post tests. Eye-tracking data and note quality were also evaluated. Results showed that students paid more attention to ClassMeta agents, and note-taking quality improved significantly. The ClassMeta agents reduced the need for instructor interventions, particularly in question-raising and discussion phases. **Conclusion:** ClassMeta effectively promotes classroom participation in VR environments by leveraging the capabilities of LLMs. The study demonstrates the potential of virtual agents to enhance student engagement and learning outcomes.