Ontology-guided multi-agent framework targets explainable differential diagnosis

[arXiv] Roxana Ramona Szomiu, George Ursu and Adrian Groza propose an ontology-guided, multi-agent differential diagnosis framework that replicates physicians' iterative reasoning. According to the authors, large language models perform well on medical benchmarks but often rely on pattern recognition from training data rather than grounded clinical evidence, can produce plausible yet unsupported diagnostic conclusions, and remain largely opaque. The framework grounds diagnostic inference in a structured, expert-curated medical knowledge base, decomposes diagnosis into explicit, auditable reasoning steps, and is said by the authors to constrain inference to evidence-backed associations, reducing hallucination and improving transparency, traceability and consistency. The system also identifies the most discriminative missing symptoms and generates targeted follow-up questions to reduce uncertainty. The authors say these components improve the reliability, explainability and diagnostic accuracy of AI-assisted differential diagnosis. The work is a preprint and has not been peer-reviewed; source code is publicly available.