A Focused Survey on Patient Simulation with Large Language Models

Abstract

The rapid advancement of generative artificial intelligence technologies, particularly large language models such as ChatGPT, has opened new avenues in healthcare education. One of the most promising applications is the simulation of virtual patients to enhance clinical training. This study aims to critically analyze and compare ten recent peer-reviewed research articles that explore the integration of ChatGPT-based virtual or standardized patients into medical education. The selected studies vary in methodology, scope, and application, ranging from history taking and empathy training to gamified learning environments, emotion recognition, and structured simulation platforms. A qualitative synthesis was conducted, focusing on implementation strategies, learner engagement, user satisfaction, and educational outcomes. The review reveals that large language model-powered virtual patients show high potential in improving communication skills, diagnostic reasoning, and learner autonomy. Several studies report positive user feedback and notable cost and scalability advantages compared to traditional simulation methods. In addition, novel training techniques targeting the emotional awareness of artificial intelligence agents are emerging as an important direction for future development. However, limitations such as response accuracy, emotional realism, and the absence of non-verbal communication remain key challenges. Most authors agree that large language models should complement rather than replace human interaction in simulation-based learning. This paper contributes to a deeper understanding of how generative artificial intelligence can be effectively used in medical education, providing valuable insights for educators, curriculum designers, and technology developers seeking to integrate artificial intelligence tools into training programs.

Ähnliche Arbeiten