The University of Manchester
The University of Manchester is a leading research-intensive university with internationally recognised expertise in artificial intelligence, data science and biomedical and health research. Within the Department of Computer Science, the National Centre for Text Mining (NaCTeM) conducts research in natural language processing (NLP), text mining, large language models (LLMs), knowledge discovery and biomedical and clinical AI. NaCTeM has extensive experience in developing methods and tools for extracting, integrating and reasoning over complex biomedical knowledge, with strengths in biomedical information extraction, evidence synthesis, causal and mechanistic knowledge discovery, trustworthy AI and the evaluation of generative AI systems.
Role within AIRIS
Within AIRIS, the University of Manchester contributes expertise in NLP, LLMs, multi-agent AI, biomedical knowledge extraction, scientific reasoning, evidence synthesis, evaluation and trustworthy AI.
UOM has a major role in the development of AIRIS's AI-assisted hypothesis generation and scientific discovery capabilities. This includes developing multi-agent approaches that integrate scientific literature, multimodal data and mechanistic knowledge to generate, critique, refine and prioritise biomedical hypotheses. Emphasis will be placed on evidence-grounded reasoning, including the assessment of the plausibility and novelty of generated hypotheses and the identification of supporting and conflicting scientific evidence.
UOM will also contribute to the evaluation and trustworthiness of AIRIS GenAI systems, including methods for assessing factuality, hallucination, robustness, bias and fairness, and the quality of AI-generated reasoning. This will include examining potential biases in training data, scientific literature, biomedical datasets, model outputs and agent interactions, as well as assessing whether system performance and generated hypotheses vary across relevant demographic, clinical, geographic or other population groups. UOM will support the development of bias-aware evaluation protocols, mitigation strategies and transparent reporting practices. This work will support AIRIS's broader objective of establishing rigorous evaluation and benchmarking methodologies for trustworthy, fair and responsible GenAI in biomedical research.
Through NaCTeM's longstanding expertise in biomedical text mining and evidence synthesis, UOM will additionally contribute methods for connecting information extracted from the scientific literature with AIRIS's mechanistic and causal modelling components, helping ensure that generated hypotheses and explanations are traceable to scientific evidence and that potential biases or gaps in the underlying evidence are identified and communicated.