Using ChatGPT as a virtual client in counseling training: counselors‑in‑training’s perceptions across counseling self‑efficacy levels
摘要
This study explores counselors-in-training (CIT)’s perceptions of using ChatGPT as a virtual client for psychotherapy training. Specifically, it examines the perceived usefulness and challenges of ChatGPT in role-play sessions and investigates how these perceptions differ based on counseling self-efficacy. A case study approach was employed with 43 CITs in South Korea. Participants engaged in five AI-based psychotherapy training sessions using ChatGPT as a virtual client. Qualitative data were collected through written interviews and analyzed using thematic analysis. Additionally, epistemic network analysis (ENA) was applied to compare perceptions based on counseling self-efficacy levels. Participants found ChatGPT useful for practicing psychotherapy, reducing anxiety, and receiving feedback. However, challenges included unrealistic client responses, difficulty in building rapport due to a lack of emotional expression, and technical issues. ENA results showed that those with high counseling self-efficacy focused on ChatGPT’s ability to simulate realistic clients, while those with low self-efficacy emphasized anxiety reduction and convenience. ChatGPT can be a valuable supplementary tool for psychotherapy training but has limitations in emotional expression and system stability. Enhancing AI’s human-like interaction and integrating it into immersive platforms may improve its effectiveness for training future psychotherapists.