Development and Application of a ChatGPT-Based Simulation Training Platform for Korean Medicine

• Published in: 동의신경정신과학회지 Vol. 35; no. 4; pp. 413 - 427
• Main Authors: Jin-woo Ryu, Chan-young Kwon, Jea-seong Park, Seong-ryong Lim, Hye-li Jeon, Hui-jun Kim, Sung-hee Kim
• Format: Journal Article
• Language: Korean
• Published: 대한한방신경정신과학회 31.12.2024
• Subjects: ChatGPT; Clinical performance examination; Education; Korean medicine; Korean medicine; Education; ChatGPT; Clinical performance examination
• ISSN: 1226-6396; 2234-4942

Objectives: This study aimed to develop a ChatGPT-based stimulation training platform for Korean medicine (KM) education to provide students with virtual patient interactions and enhance their clinical competencies and evaluate its performance. Methods: A web-based platform integrating ChatGPT API was developed using Node.js and MongoDB with a react-based frontend interface. The platform included functions for instructors to create patient scenarios and for students to conduct simulated clinical consultations. The system was piloted with 47 fourth-year KM students in a KM Neuropsychiatry course, focusing on suicide risk assessment. Student perceptions were evaluated through an anonymous survey measuring satisfaction with simulated consultation experience and AI chatbot usability. Results: Of a total of 47 students, 42 completed the survey with a response rate of 89.4%. Simulation experience received high satisfaction scores across multiple dimensions, including interest (4.44± 0.62), appropriateness (4.71±0.38), knowledge acquisition (4.24±0.79), and skill development (4.14± 0.62). The chatbot system demonstrated strong usability scores in usefulness (4.40±0.54), accessibility (4.42±0.56), and overall discoverability (4.43±0.63). However, students expressed some skepticism about the authenticity of virtual patient responses (3.69±0.95). Conclusions: The platform shows promise as an accessible and standardized tool for clinical education in KM. While it effectively addresses resource constraints in traditional standardized patient programs, its low authenticity ratings suggest that it should be used to complement rather than replace conventional clinical training methods. Future developments should focus on enhancing response authenticity and expanding the range of clinical scenarios while maintaining high levels of accessibility and user satisfaction.