Background This study presents a novel approach to educational role-playing through an AI-based bot, leveraging GPT-4 to simulate anamnesis interviews in various learning scenarios. Developed collaboratively by an interdisciplinary team of university lecturers, the bot provides a platform for students of different health study programs to engage in complex patient-health professional conversations, offering an alternative to traditional role plays with actors or real patients.
Methods This study utilized a GPT-4 based digital teaching assistant to train anamnesis interviews in virtual settings with students from different online health care study programs. Student’s satisfaction, virtual patient’s accuracy, its realism, and quality was evaluated with a quantitative survey.
Results The evaluation of the bot focused on student feedback, highlighting a preference for the AI-driven method due to its immersive and interactive nature. Preliminary results show that students consistently rate the language ability of the AI model positively. More than 80% of students rated the professional and content-related precision of the virtual patient as good to excellent. Even as a text-based chatbot, the vast majority of students see a fairly close to very close relationship to a real anamnesis interview. The results further indicate that students even prefer this training approach to traditional in-person role-plays.
Conclusions The study underscores the bot's potential as a versatile tool for enriching learning experiences across multiple health disciplines, signaling a meaningful shift in educational practices towards the integration of AI technologies.