Classroom Integration of AI-Simulated Patients and Automated Feedback in Doctoral Psychotherapy Training: Longitudinal Mixed Methods Study
Interactive Journal of Medical Research · 11.11.2026 · 15 · e93199-e93199
DOI: 10.2196/93199 ↗Kurz gefasst
Mixed-Methods-Studie in der Zeitschrift Interactive Journal of Medical Research, online erschienen am 11.11.2026. Verfasst von Ellen Converse und 7 weiteren Autor:innen. Beteiligte Einrichtung: PAU-Stanford PsyD Consortium, University Palo Alto, 1791 Arastradero Road, Palo Alto, CA, 94304, United States, 1 650-498-9111. Die Arbeit nennt 56 Literaturangaben. Der Volltext ist frei in PubMed Central verfügbar.
Abstract (Originalsprache)
Background
Psychotherapy education increasingly emphasizes technology-assisted methods that extend deliberate practice beyond human supervision. AI has enabled automated skill assessment and patient simulations, yet few studies have examined how students experience such tools in real-world settings alongside established practice modalities.
Objective
We aimed to evaluate students’ experiences with AI-augmented deliberate practice embedded in an existing psychotherapy course sequence. Secondary aims included identifying implementation barriers, ethical and pedagogical considerations, and factors influencing students’ acceptance of AI-assisted learning.
Methods
Participants were 29 first-year clinical psychology doctoral students in a 2-quarter psychotherapy sequence at an American Psychological Association–accredited program. In the fall course, students completed weekly peer role-plays and could optionally submit transcripts to CARE for AI feedback to supplement peer feedback. In the winter course, students completed required weekly deliberate practice through Skillsetter video vignettes and could optionally practice with CARE’s voice-based AI-simulated patients, receiving AI and instructor feedback afterward. Surveys with Likert-type and free-response items were voluntary in the fall and required in the winter; 9 and 25 students completed them, respectively. Analyses included descriptive statistics, matched-pairs testing, and thematic analysis.
Results
In the fall course (n=9), where privacy concerns and survey nonresponse limited data on AI feedback, students rated peer feedback higher than AI feedback on being constructive and helpful (median 3, IQR 3-3 vs median 2, IQR 2-2; P =.02) and recognizing areas to improve (mean 5.22, SD 0.83 vs mean 4.33, SD 0.87; P =.046), citing transcription errors and a narrow focus on empathy while valuing AI’s alternative phrasing. In the winter course (n=25), Skillsetter was rated higher than CARE only for learning from mistakes (mean 4.52, SD 2.00 vs mean 3.20, SD 2.04; P =.03); immersion, spontaneous responding, communication beyond words, and alliance building did not differ significantly. Students valued CARE for interactive practice not possible with vignettes but reported undetected vocal pauses (20/25, 80%), insufficient realism (14/25, 56%), and limited emotional expression (13/25, 52%), and raised concerns about AI replacing peer support (22/25, 88%) and peer interaction (20/25, 80%). Asked about AI broadly, students expressed mixed views, citing concerns about supplanted relational dynamics, nonconsensual data use, and generative AI’s environmental impacts.
Conclusions
Embedding AI-augmented deliberate practice into an existing course sequence was feasible, but students’ experiences favored established methods: they preferred human over AI feedback and described video vignettes as more realistic than voice-based AI interactions, while valuing AI-simulated patients for interactive practice other modalities could not provide. Implementation barriers included feedback quality, transcription accuracy, and instructor time required to supplement AI feedback. Perceptions of AI’s instructional utility appeared contingent on students’ trust in the technology, suggesting future deployments should weigh a sample’s attitudes toward AI and its value relative to existing training resources.
Schlagwörter
Stichwörter der Autor:innen
Autor:innen und Institutionen
- PAU-Stanford PsyD Consortium, University Palo Alto, 1791 Arastradero Road, Palo Alto, CA, 94304, United States, 1 650-498-9111.
- Department of Computer Science, Stanford University, 353 Jane Stanford Way, Stanford, CA, 94305, United States.
- Department of Psychiatry and Behavioral Sciences, Stanford University School of Medicine, 401 Quarry Road, Stanford, CA, 94304, United States.
Zitieren
APA
Converse, E., Louie, R., Safer, D., Pacheco, J. P., Fang, W., Kent, J., Arnow, B., & Yang, D. (2026). Classroom Integration of AI-Simulated Patients and Automated Feedback in Doctoral Psychotherapy Training: Longitudinal Mixed Methods Study. Interactive Journal of Medical Research, 15, e93199-e93199. https://doi.org/10.2196/93199
Vancouver
Converse E, Louie R, Safer D, Pacheco JP, Fang W, Kent J, et al. Classroom Integration of AI-Simulated Patients and Automated Feedback in Doctoral Psychotherapy Training: Longitudinal Mixed Methods Study. Interactive Journal of Medical Research. 2026;15:e93199-e93199. doi:10.2196/93199
BibTeX
@article{Converse2026,
author = {Converse, Ellen and Louie, Ryan and Safer, Debra and Pacheco, Juan Pablo and Fang, William and Kent, Jamie and Arnow, Bruce and Yang, Diyi},
title = {Classroom Integration of AI-Simulated Patients and Automated Feedback in Doctoral Psychotherapy Training: Longitudinal Mixed Methods Study},
journal = {Interactive Journal of Medical Research},
year = {2026},
volume = {15},
pages = {e93199--e93199},
doi = {10.2196/93199}
}
EUVIMED