Hostname: page-component-7b7567c8c-fvjsw Total loading time: 0 Render date: 2026-11-17T07:27:50.396Z Has data issue: false hasContentIssue false

Training Psychiatry with AI: Simulator with 100 Cases and Image-Based Multi Level Communication

Published online by Cambridge University Press:  31 July 2026

D. Novak*
Affiliation:
Department of Cybernetics, Czech Technical University in Prague, Prague
F. Bodnár
Affiliation:
Department of Cybernetics, Czech Technical University in Prague, Prague
P. Kubíček
Affiliation:
Faculty of Medicine in Pilsen, Charles University, Pilsen, Czech Republic
J. Vevera
Affiliation:
Faculty of Medicine in Pilsen, Charles University, Pilsen, Czech Republic
*
*Corresponding author.

Abstract

Core share and HTML view are not available for this content. However, as you have access to this content, a full PDF is available via the 'Save PDF' action button.
Introduction

We present a conversational-AI platform for simulation-based psychiatric training. The platform includes 100 clinically grounded use cases across three communication-difficulty tiers (cooperative, resistant, and non-cooperative), enabling practice in adaptive questioning and de-escalation. During each interview, the system renders a time-ordered series of simple avatar images. When a medication is mentioned, students can open the integrated drug module to view the package leaflet.

Objectives

We aim to extend our conversational-AI virtual-patient platform into a comprehensive, clinically curated set of 100 psychiatry use cases, and to examine how graded communication difficulty shapes learners’ questioning strategies. To support presence and engagement, we added simple avatar visuals that depict only the patient’s physical appearance. When medications arise in dialogue, an integrated drug module reveals the package leaflet enabling context-aware medication education.

Methods

Users interview a virtual patient via text or voice; a large language model (LLM) pipeline maps utterances to intents and generates guarded, diagnosis-consistent replies. The library now comprises 100 clinician-curated patient profiles derived from real cases, de-identified records, and carefully synthesised data, spanning diverse diagnostic scenarios (Fig. 1). During interviews, the system renders a simple avatar sequence to visualise the patient’s physical appearance (not mental-status change). Built-in tools let learners inspect relevant physiological/clinical examinations (Fig. 2), while an embedded assessment engine evaluates adherence to ethical and methodological standards and verifies diagnostic accuracy (Fig. 3).

Results

Building on our prior pilot, where psychiatrists rated dialogue realism highly and students reported improved confidence with an average diagnostic score of ~71%, the expanded platform functioned reliably at scale with 100 curated cases. System latency remained low; trainees engaged smoothly with three communication tiers, and interaction logs showed expected shifts toward more targeted, clarifying questions in “so-so” and non-cooperative scenarios. The avatar visuals were well-received for anchoring patient identity without implying clinical state. When medications were mentioned, learners successfully used the integrated drug module.Image 1:

A screenshot of a medical simulation software interface featuring a chat conversation with a virtual patient named Lucie Vitková. See long description.Image 2:

A digital interface for a Virtual Patient simulation showing a chat history with medical test results and a patient profile for Lucie Vítková. See long description.Image 3:

A digital interface for a Virtual Patient medical simulation showing a chat history, patient profile, and diagnostic results. See long description.

Conclusions

A conversational-AI psychiatry simulator with graded communication difficulty, simple avatar visuals, and an in-context drug module is feasible and well-accepted, supporting medication-safety reasoning. Interaction analytics indicate more targeted questioning as difficulty increases, while learners efficiently retrieve drug information in context. The platform is ready for curriculum integration with automated feedback and educator dashboards; comparative outcomes will be reported.

Disclosure of Interest

None Declared

Information

Type
Abstract
Creative Commons
Creative Common License - CCCreative Common License - BY
This is an Open Access article, distributed under the terms of the Creative Commons Attribution licence (https://creativecommons.org/licenses/by/4.0/), which permits unrestricted re-use, distribution, and reproduction in any medium, provided the original work is properly cited.
Copyright
© The Author(s), 2026. Published by Cambridge University Press on behalf of European Psychiatric Association
Submit a response

Comments

No Comments have been published for this article.