Medical students and clinical teachers experiences and perceptions of using AI-based simulated patients in psychiatry clinical skills training, a qualitative analysis
- 1University of Kent
Abstract
Objectives: To explore medical students' and clinical teachers' experiences and perceptions of an AI-based simulated patient using a large language model (LLM-ASP) for psychiatry clinical skills training, to inform its development and integration. Design: Qualitative focus group study with reflexive thematic analysis. Setting: Kent and Medway Medical School, UK. Student groups met in person and online; teacher groups met online. Participants: Sixteen undergraduate medical students (years 3 and 4; two focus groups) and nine psychiatry clinical teachers (six residents, three consultants; three focus groups), recruited through university mailing lists and the Royal College of Psychiatrists' South Eastern Division newsletter. Interventions: Before each discussion, participants used an LLM-ASP tool built by the research team (ChatGPT Advanced Voice Mode, Streamlit, LangChain) for about 10 minutes. Sessions were audio-recorded, transcribed verbatim and analysed using Braun and Clarke's reflexive thematic analysis with Saldana's multi cycle coding. Results: Students and teachers each generated three themes. Students described the interface (trust, realism, reliability), educational value, and ethical, cultural and governance considerations. Teachers described pedagogical value, design and usability, and ethical, cultural and governance considerations. Both groups saw LLM-ASP as a flexible, low-pressure, self-paced supplement for early communication skills practice and for rehearsing scenarios hard to access in placements, not a replacement for actor-based simulation or patient contact. Barriers included limited emotional realism, robotic delivery and response delays. Shared ethical concerns were safeguarding, data storage and consent, accountability, demographic bias in generated patients, and over-reliance weakening critical thinking. Students also raised the impact on simulated-patient actors. Students focused on trust and usability; teachers focused on pedagogical integration and oversight. Conclusions: Participants saw LLM-ASP as a potentially useful complement to existing psychiatry teaching. Adoption depends on technical reliability, realism, representation and governance, and on keeping human interaction central to the curriculum.
Related
- Perceptions and Experiences of the Foster Family Program Among Alumni and Undergraduate Medical Students at CMC Ludhiana: An Exploratory Mixed-Methods Study
- Conditional willingness to care for patients with Ebola virus disease: a qualitative study of risk-benefit perceptions and support needs of clinical students at a Ugandan medical school
- Nursing students’ experiences with using generative artificial intelligence: a descriptive qualitative study
- A qualitative exploration of community health workers' roles and motivations in delivering care during and after the COVID-19 pandemic: experiences and community perceptions in rural and urban Delhi, India.
- Recovery experiences and perceptions of quality of life after percutaneous coronary intervention for ST-elevation myocardial infarction: An interpretive descriptive qualitative study
BibTeX
@misc{gamaleldin2026medical,
title = {{Medical students and clinical teachers experiences and perceptions of using AI-based simulated patients in psychiatry clinical skills training, a qualitative analysis}},
author = {D. A. Gamaleldin and j. Rodda and S. Shergil and J. Ang and S. Hotham and R. Rice and H. Mansy and S. Benett and O. OLURE and V. Williams},
howpublished = {medRxiv},
year = {2026},
month = oct,
doi = {10.64898/2026.10.28.26366085},
url = {https://www.medrxiv.org/content/10.64898/2026.10.28.26366085v1},
note = {CC BY-NC},
}
RIS
TY - UNPB
TI - Medical students and clinical teachers experiences and perceptions of using AI-based simulated patients in psychiatry clinical skills training, a qualitative analysis
AU - D. A. Gamaleldin
AU - j. Rodda
AU - S. Shergil
AU - J. Ang
AU - S. Hotham
AU - R. Rice
AU - H. Mansy
AU - S. Benett
AU - O. OLURE
AU - V. Williams
PY - 2026
DA - 2026/10/28
PB - medRxiv
DO - 10.64898/2026.10.28.26366085
UR - https://www.medrxiv.org/content/10.64898/2026.10.28.26366085v1
AB - Objectives: To explore medical students' and clinical teachers' experiences and perceptions of an AI-based simulated patient using a large language model (LLM-ASP) for psychiatry clinical skills training, to inform its development and integration. Design: Qualitative focus group study with reflexive thematic analysis. Setting: Kent and Medway Medical School, UK. Student groups met in person and online; teacher groups met online. Participants: Sixteen undergraduate medical students (years 3 and 4; two focus groups) and nine psychiatry clinical teachers (six residents, three consultants; three focus groups), recruited through university mailing lists and the Royal College of Psychiatrists' South Eastern Division newsletter. Interventions: Before each discussion, participants used an LLM-ASP tool built by the research team (ChatGPT Advanced Voice Mode, Streamlit, LangChain) for about 10 minutes. Sessions were audio-recorded, transcribed verbatim and analysed using Braun and Clarke's reflexive thematic analysis with Saldana's multi cycle coding. Results: Students and teachers each generated three themes. Students described the interface (trust, realism, reliability), educational value, and ethical, cultural and governance considerations. Teachers described pedagogical value, design and usability, and ethical, cultural and governance considerations. Both groups saw LLM-ASP as a flexible, low-pressure, self-paced supplement for early communication skills practice and for rehearsing scenarios hard to access in placements, not a replacement for actor-based simulation or patient contact. Barriers included limited emotional realism, robotic delivery and response delays. Shared ethical concerns were safeguarding, data storage and consent, accountability, demographic bias in generated patients, and over-reliance weakening critical thinking. Students also raised the impact on simulated-patient actors. Students focused on trust and usability; teachers focused on pedagogical integration and oversight. Conclusions: Participants saw LLM-ASP as a potentially useful complement to existing psychiatry teaching. Adoption depends on technical reliability, realism, representation and governance, and on keeping human interaction central to the curriculum.
ER -