Title:
Can a Large Language Model Behave Like a Doctor?
Thinking, Acting, and Learning to Reason — a Clinician’s Perspective
Short Description:
Large language models pass medical examinations, but a clinician’s day is not made of examination questions. Real care means choosing what to ask next, holding hypotheses open under uncertainty, knowing when to stop, and letting the patient’s values decide among defensible options.
Short Bio:
Xi Chen, MD, is a postdoctoral researcher and orthopaedic clinician at West China Hospital, Sichuan University, China, working in the field of musculoskeletal and sports medicine. His research sits at the intersection of artificial intelligence and clinical medicine, with a focus on developing and validating AI systems that work inside real clinical workflows. He has published more than ten international papers as first or co-first author, including several highly cited works, contributed as a co-author to an international reporting guideline for generative AI research in healthcare, and holds more than ten patents and other intellectual property as a principal contributor. Working as a practising clinician, he is particularly interested in how AI systems behave once they leave benchmark settings and enter routine care.
His work follows two main directions. The first is AI-assisted diagnosis and clinical decision making, where he uses large language models to support dynamic diagnostic reasoning that follows the way clinicians actually gather information and reach a diagnosis, to simulate multidisciplinary team discussion through multiple collaborating agents, and to provide end-to-end support for rare diseases across diagnosis, treatment, and prognosis. The second is chronic disease management, where he combines large language models with wearable sensing and computer vision to automate patient assessment, generate individualised treatment and exercise plans, and deliver long-term intervention and follow-up, with knee osteoarthritis as the primary clinical setting.