E66 Credibility and Appropriateness in AI-Based Simulation as Empathy Training Tools (Short paper) Proceedings of the 21st Chais Conference for the Study of Innovation and Learning Technologies: Learning in the Digital Era I. Blau, A. Caspi, Y. Eshet-Alkalai, N. Geri, Y. Kalman, D. Olenik-Shemesh, Y. Sidi, & N. Brandel (Eds.), Ra'anana, Israel: The Open University of Israel Credibility and Appropriateness in AI-Based Simulation as Empathy Training Tools (Short paper) Ronen Grinberg Technion – Israel Institute of Technology ronen.g@campus.technion.ac.il Dan Waisman Technion – Israel Institute of Technology dwaisman@technion.ac.il Rinat B. Rosenberg-Kima Technion – Israel Institute of Technology rinatros@technion.ac.il אמינות והתאמת סימולציה מבוססת בינה מלאכותית ככלי להכשרת תקשורת אמפתית )מאמר קצר( רינת ב. רוזנברג - קימה הטכניון – מכון טכנולוגי לישראל rinatros@technion.ac.il דן וייסמן הטכניון – מכון טכנולוגי לישראל dwaisman@technion.ac.il רונן גרינברג הטכניון – מכון טכנולוגי לישראל ronen.g@campus.technion.ac.il Abstract Empathic communication is a central competency for educational rehabilitation counselors, yet existing training methods often struggle to provide the socially grounded practice needed for developing such skills. Advances in generative artificial intelligence offer new opportunities for scalable simulation-based training. Guided by Vygotsky’s Sociocultural Theory and the Media Equation and Computers Are Social Actors (CASA) frameworks, this study examines a chatbot-based system designed to support empathy training through a Virtual Participant (VP) simulating a young adult with a disability. Using a qualitative case study approach, eight senior social workers from a rehabilitation organization evaluated two VP-practitioner interaction transcripts. Findings regarding the VP revealed six credibility challenges: verbal credibility, emotional and behavioral credibility, credibility in expressing resistance and initiative, inconsistencies between verbal utterances and described behavior, absence of physical presence, and absence of shared relational history. These issues limited the VP’s perceived authenticity and reduced its effectiveness as a simulated partner for practicing empathic dialogue. Overall, the findings highlight both the potential and limitations of AI-mediated empathy training. While generative agents hold great potential for practice, meaningful development of empathic communication appears to require human facilitation to contextualize, refine, and carefully prompt AIgenerated interactions. Keywords: Simulation-based learning, Simulated Participant, the Media Equation, GenAI, AIED.
RkJQdWJsaXNoZXIy Mjk0MjAwOQ==