Reinventing Clinical Dialogue: Agentic Paradigms for LLM Enabled Healthcare Communication
Xiaoquan Zhi, Hongke Zhao, Likang Wu, Chuang Zhao, Hengshu Zhu
2025-12-11
Summary
This paper explores how artificial intelligence, specifically large language models, are changing in the field of medical diagnosis and treatment, moving from simply generating text to acting more like intelligent assistants.
What's the problem?
Current large language models are really good at *sounding* correct, but they don't always prioritize being *actually* correct, which is a huge problem in healthcare where mistakes can be dangerous. They also tend to react to information without remembering past interactions or having a plan, unlike a doctor who builds a case over time.
What's the solution?
The paper doesn't offer a single solution, but instead analyzes different approaches researchers are taking to build more reliable medical AI. It categorizes these approaches into four main types based on how they use medical knowledge and how much independent decision-making they have. It breaks down each type, looking at how they plan, remember information, take action, and learn over time, all while trying to balance being helpful and staying safe.
Why it matters?
This work is important because it provides a clear framework for understanding the rapidly evolving field of medical AI. By categorizing different approaches and highlighting their strengths and weaknesses, it helps researchers and developers build better, safer, and more effective AI tools for healthcare.
Abstract
Clinical dialogue represents a complex duality requiring both the empathetic fluency of natural conversation and the rigorous precision of evidence-based medicine. While Large Language Models possess unprecedented linguistic capabilities, their architectural reliance on reactive and stateless processing often favors probabilistic plausibility over factual veracity. This structural limitation has catalyzed a paradigm shift in medical AI from generative text prediction to agentic autonomy, where the model functions as a central reasoning engine capable of deliberate planning and persistent memory. Moving beyond existing reviews that primarily catalog downstream applications, this survey provides a first-principles analysis of the cognitive architecture underpinning this shift. We introduce a novel taxonomy structured along the orthogonal axes of knowledge source and agency objective to delineate the provenance of clinical knowledge against the system's operational scope. This framework facilitates a systematic analysis of the intrinsic trade-offs between creativity and reliability by categorizing methods into four archetypes: Latent Space Clinicians, Emergent Planners, Grounded Synthesizers, and Verifiable Workflow Automators. For each paradigm, we deconstruct the technical realization across the entire cognitive pipeline, encompassing strategic planning, memory management, action execution, collaboration, and evolution to reveal how distinct architectural choices balance the tension between autonomy and safety.