ClinicBot: A Guideline-Grounded Clinical Chatbot with Prioritized Evidence RAG and Verifiable Citations
Researchers have developed ClinicBot, an AI system designed to provide accurate and verifiable clinical support by grounding responses in official medical guidelines. The system addresses the hallucination problem common in large language models by employing a novel retrieval-augmented generation (RAG) approach. ClinicBot prioritizes evidence based on clinical significance and guideline structure, rather than simple textual similarity, and presents answers with verifiable citations. It has been demonstrated using diabetes-related questions and the American Diabetes Association's Standards of Care. AI
IMPACT This system could improve the reliability of AI in high-stakes medical contexts by ensuring answers are grounded in established guidelines.