Clinical reasoning (CR) is a vital skill in medical education, yet traditional digital simulations often fall short when it comes to realism and emotional engagement. Virtual Patients (VPs) are widely used to simulate clinical scenarios, but most of them lack the interactivity and human-like responsiveness needed to truly prepare students for real-life scenarios. As AI advances, there is a growing interest in how technologies like social robots and large language models (LLMs) could transform this learning experience.
The research brought together specialists in medical education, human-robot interaction, and conversational AI from Karolinska Institutet (KI), one of Europe’s leading medical universities, the Royal Institute of Technology (KTH) and Sorbonne University.
KI is known for its cutting-edge use of virtual patients in medical training to improve clinical education in disciplines such as nursing or dentistry. The university is also home to the Stockholm Medical AI and Learning Environments initiative, a facility aimed at accelerating the development and safe integration of AI and robotics in healthcare
The project aimed to test a new approach to CR training in rheumatology: combining a social robotic interface with an LLM-powered Virtual Patient. The study involved 15 medical students who interacted with both a traditional computer-based VP and a VP delivered through a social robot. Their experiences were measured through surveys and in-depth interviews to assess the educational impact of each platform.
The Furhat robot serves as the physical interface for the AI-powered Virtual Patient. Its expressive face, natural eye contact, and conversational abilities allowed students to engage with the simulation in a more lifelike way. Integrated with an LLM, the robot could respond intelligently to clinical questions, simulate patient emotions, and adapt to different communication styles.
Students found the robotic experience more authentic and effective for practicing clinical reasoning, emotional communication, and patient interaction. They reported stronger engagement, improved learning, and greater realism compared to the computer-only interface.
By blending robotics, AI, and medical education, this study points to a future where the digital components of clinical training are increasingly human-centered.
Read more here: https://www.kth.se/is/tmh/projects/conversational-robots-as-virtual-patients-1.1366621
