Artificial Intelligence AI-Driven VR Platform for Education : A User Experience Study on Conflict Resolution Training Through a Proof-of- Concept Simulator

다국어 초록 (Multilingual Abstract)

Conflicts are inevitable in human social interactions, impacting relationships, mental health, and collaboration across settings such as workplaces, schools, and cross-cultural exchanges. While traditional conflict resolution training programs rely on theoretical models like the Thomas-Kilmann Conflict Mode Instrument (TKI), their reliance on questionnaire- based exercises or human-dependent role-playing limits ecological validity and scalability. However, previous research attempting to apply AI technologies in this domain has exhibited shortcomings in areas such as immersion and cross-cultural adaptability. Recent advancements in AI and VR technologies offer transformative potential for immersive, data- driven training. This study proposes the AI-based VR Scenario Experiment Platform (AVSEP), a novel framework integrating Large Language Models (LLMs) and VR to simulate realistic conflict scenarios, measure user responses, and deliver personalized feedback. The AVSEP framework was validated by developing the Student Conflict Simulator VR, a proof-of-concept application targeting Korean university students. The platform combines realistic virtual characters and VR environment design, Meta Quest Pro and sensors for multimodal interaction and data collection (e.g., facial expressions, gaze, and other physiological signals), and LLM-driven AI assistants for dynamic role-playing. Virtual characters were designed with distinct personalities and culturally tailored backstories, guided by the Six Steps of the No-Lose Method to structure conflict resolution stages. Speech-to-Text (STT) and Text-to-Speech (TTS) enabled natural voice interactions, while Retrieval-Augmented Generation (RAG) enhanced contextual awareness. We conducted a multifaceted user experience study on the Student Conflict Simulator VR to assess the platform ’s feasibility and user experience. Data collection integrated validated user experience scales (Multimodal Presence Scale and Virtual Reality System Usability Questionnaire), and in-VR questionnaires for capturing subjective perceptions on the scenario realism and training utility. The findings revealed most participants’ effective engagement with the immersive environment and AI-mediated role-playing. The absence of participant avatars influenced the presence results regarding self-presence, underscoring design trade-offs between usability and embodied immersion in educational applications. In addition, we collected chat log data and multiple sensor data to initially test AVSEP's data collection and real-time application features. The study demonstrates AVSEP’s potential to bridge theoretical training and practical skill application, offering scalable, immersive conflict resolution education. Future work will expand scenarios (e.g., workplace, healthcare), refine emotional depth via advanced LLMs, and enhance cross-cultural adaptability. By modularizing content creation and leveraging multimodal interactions, AVSEP establishes a foundation for next-generation VR-AI training tools, showing the potential for both academic and commercial applications in conflict resolution training and other interpersonal skill development.

목차 (Table of Contents)

• Chapter 1. Introduction 1
• 1.1 Related Work 3
• 1.1.1 Communication and Conflict Resolution Training and Assessment Research 3
• 1.1.2 Incorporating AI into Communication and Conflict Resolution Training 5
• 1.1.3 Research on Applications Integrating AI and VR Technologies 7
• 1.2 Problems 10
• 1.3 Objectives of the Study 11
• Chapter 2. Research Questions and Methodology 13
• 2.1 Research Questions 13
• 2.2 Development Method 13
• 2.3 User Experience Study Method 15
• 2.3.1 Environment and Device Configuration 18
• 2.3.2 Participants 20
• 2.3.3 Data Collection Instruments 20
• 2.3.4 Experimental Procedure 21
• 2.3.5 Data Analysis Methods 22
• 2.3.6 Ethical Considerations 23
• Chapter 3. Result 24
• 3.1 AI-based VR Scenario Experiment Platform Design 24
• 3.1.1 Selection and Usage of Conflict Resolution Models 29
• 3.1.2 Design of Virtual Character Personalities and Backstories 31
• 3.1.3 Design of Virtual Characters and Virtual Environments 35
• 3.1.4 AI Assistants 38
• 3.1.5 Voice Chat System 40
• 3.1.6 Scenario Contents Manager 41
• 3.1.7 Tutorial Scene 41
• 3.1.8 Conflict Resolution Scenario Configuration Workflow 42
• 3.2 User Experience Study Result 44
• 3.2.1 Demographics and Previous VR Experience 44
• 3.2.2 In-VR Questionnaire Results 44
• 3.2.3 Presence 50
• 3.2.4 Usability and Satisfaction 52
• Chapter 4. Discussion 54
• 4.1 Motivation and Contributions 54
• 4.2 Interpretation of the Results 55
• 4.2.1 User Experience Study Results 55
• 4.2.2 Comparison with Existing Studies 57
• 4.3 Implications 60
• 4.4 Limitations 63
• Chapter 5. Conclusion 65
• 5.1 Summary of the Results 65
• 5.2 Answering the Research Questions 66
• 5.3 Future works 69
• References 70
• Appendices 79
• Appendix A. Multimodal Presence Scale (English and Korean). 79
• Appendix B. Virtual Reality System Usability Questionnaire (English and Korean) 81
• Appendix C. In-VR questionnaire questions (English and Korean) 82
• Appendix D. Six Steps of the No-Lose Method (English and Korean) 84
• Appendix E. Virtual Character Personality Design 86
• Appendix F. Backstory Design (English and Korean) 87