A Methodological Framework for Designing Clinically Coherent Generative AI Simulated Patients: A Physiotherapy...

Area of Science:

• Health Informatics
• Artificial Intelligence in Education
• Medical Simulation

Background:

• Generative artificial intelligence (AI) presents opportunities for scalable health professions education.
• Current AI-driven simulated patient initiatives often lack standardized methodologies.
• Ad hoc approaches limit the effectiveness and consistency of AI simulations.

Purpose of the Study:

• To develop and evaluate a methodological framework for designing generative AI mock patients.
• To provide a structured approach for creating AI-driven simulated patients in health professions education.
• To address the need for evidence-based design in AI educational simulations.

Main Methods:

• Development of a methodological framework integrating clinical reasoning, communication, layered prompt architecture, scenario constraints, and guardrails.
• Application of the framework using a physiotherapy education chatbot as an exemplar.
• Evaluation of the framework's effectiveness in supporting consistent patient-persona behavior and scenario progression.

Main Results:

• The framework facilitated consistent patient-persona behavior and guided scenario progression in AI simulations.
• The study identified specific interaction challenges within the AI simulations.
• Mitigation strategies for identified challenges were developed and described.

Conclusions:

• The proposed framework offers a practical methodology for the evidence-based design of generative AI simulations.
• This work contributes to the health informatics education community by providing a structured approach.
• The framework supports the development of more consistent and effective AI-driven simulated patients for training health professionals.