EUVIMEDEUVIMED entdecken
Open Access · CC BY-NC 4.0

Artificial Clients, Genuine Interactions? Assessing ChatGPT4's Process‐Level Performance in Clinical Simulations Using Qualitative Content Analysis

Xihe Tian, Yanhong Liu, Sarah Litt, Kahyen Shin, Tanqiu Jiang

Counselling and Psychotherapy Research · 2025-12 · 25 (4)

DOI: 10.1002/capr.70063 ↗

Kurz gefasst

Fachartikel in der Zeitschrift Counselling and Psychotherapy Research, online erschienen am 14.12.2025. Verfasst von Xihe Tian und 4 weiteren Autor:innen. Die Arbeit wurde bisher 2-mal zitiert (Crossref) und nennt 31 Literaturangaben. Der Volltext ist frei zugänglich (Open Access, CC BY-NC 4.0).

AbstractBibliografische AngabenZitierenVerwandte Studien

Abstract (Originalsprache)

Objective

With rapid advances in technology, growing attention has been directed towards evaluating Artificial Intelligence (AI) performance in human tasks; however, limited research has examined AI's performance as clinical clients in counselling simulations, particularly at the process level. This study explores how ChatGPT‐4.0 (GPT‐4), an advanced AI Large Language Model (LLM) with voice functioning, performs as a simulated client and how its responses reflect counselling processes within counselling interactions.

Method

Twelve fictional client scenarios were developed to assess GPT‐4's performance across diverse counselling situations. Using qualitative content analysis (QCA), we examined GPT‐4‐simulated client responses on the content level.

Results

GPT‐4 demonstrated multiple elements of counselling process, including the ability to form therapeutic alliance, attune to counsellors' theoretical orientations and adapt to different stages of the counselling process. Analysis also showed the GPT‐4‐simulated clients expressed strong motivation for positive changes, advanced mental health literacy and low but selective defence mechanisms. Additionally, a detectable yet limited amount of non‐verbal information emerged through tone and pacing within the model's voice output.

Conclusion

Findings suggest that GPT‐4 can produce process‐level behaviours that align with key components of counselling interactions and has the potential to become a strong supplementary training tool for counselling role‐play practices.

Zitieren

APA

Tian, X., Liu, Y., Litt, S., Shin, K., & Jiang, T. (2025). Artificial Clients, Genuine Interactions? Assessing ChatGPT4's Process‐Level Performance in Clinical Simulations Using Qualitative Content Analysis. Counselling and Psychotherapy Research, 25(4). https://doi.org/10.1002/capr.70063

Vancouver

Tian X, Liu Y, Litt S, Shin K, Jiang T. Artificial Clients, Genuine Interactions? Assessing ChatGPT4's Process‐Level Performance in Clinical Simulations Using Qualitative Content Analysis. Counselling and Psychotherapy Research. 2025;25(4). doi:10.1002/capr.70063

BibTeX
@article{Tian2025,
  author = {Tian, Xihe and Liu, Yanhong and Litt, Sarah and Shin, Kahyen and Jiang, Tanqiu},
  title = {Artificial Clients, Genuine Interactions? Assessing ChatGPT4's Process‐Level Performance in Clinical Simulations Using Qualitative Content Analysis},
  journal = {Counselling and Psychotherapy Research},
  year = {2025},
  volume = {25},
  number = {4},
  doi = {10.1002/capr.70063}
}
Kostenlos starten →