A Preliminary Study of Motivational Interviewing Training Using a Large Language Model (ChatGPT): Effects on Empathy and Motivational Interviewing Confidence among Social Work Students

Background: Motivational interviewing (MI) requires repeated practice, individualized feedback, and ongoing coaching that shortterm group education cannot sufficiently provide. Large language models (LLMs) such as ChatGPT may offer realistic conversational simulation and immediate feedback. This pilot study examined whether an LLM (ChatGPT)–integrated MI training program improves empathy and MI Confidence among social work students.Methods: A one-group repeated measures quasi-experimental design was applied to 15 undergraduate social welfare students. The four-session program used ChatGPT (GPT-5.5) as a virtual client for open-ended questions, affirmations, reflective listening, and summaries (OARS) practice with automated Motivational Interviewing Treatment Integrity-based feedback. Empathy, MI Skill Confidence, and Global Interviewing Confidence were measured at pretest, posttest, and 4-week follow-up, and analyzed using Friedman and Wilcoxon signed-rank tests. Qualitative content analysis of participants’ experiences was also conducted.Results: Empathy improved significantly across time points (χ2=6.037, P=0.049), increasing from pretest to posttest (P=0.014) and remaining higher at follow-up (P=0.037). MI Skill Confidence showed an upward trend without statistical significance (χ2=2.561, P=0.278). Global Interviewing Confidence changed significantly (χ2=6.682, P=0.035), being higher at follow-up than pretest (P=0.021). Qualitative analysis yielded four themes: realistic simulation, immediate feedback, skill mastery through repeated practice, and requests for enhanced realism.Conclusions: An LLM (ChatGPT)–integrated MI training program may improve empathy and overall interviewing confidence and function as an effective auxiliary tool complementing traditional MI education. LLMs can support learners’ repeated practice and immediate feedback.