Beyond Correlation: Towards Causal Large Language Model Agents in Biomedicine
Adib Bazgir, Amir Habibdoust Lafmajani, Yuwen Zhang
TL;DR
This paper argues that current LLMs in biomedicine mostly exploit correlations and lack genuine causal understanding, risking unreliable inferences. It proposes causal LLM agents that integrate multimodal biomedical data and grounding with knowledge graphs and formal causal inference to enable intervention-based reasoning, while addressing safety, evaluation, and reproducibility. The authors outline a research agenda spanning agent design, benchmarking, and multimodal data integration, plus concrete applications in drug discovery, personalized medicine, and public health, including examples like Mendelian Randomization workflows and Real-World Evidence generation. They emphasize the need for interdisciplinary collaboration to create reliable, controllable AI partners that can advance biomedical progress through grounded, explainable causal reasoning.
Abstract
Large Language Models (LLMs) show promise in biomedicine but lack true causal understanding, relying instead on correlations. This paper envisions causal LLM agents that integrate multimodal data (text, images, genomics, etc.) and perform intervention-based reasoning to infer cause-and-effect. Addressing this requires overcoming key challenges: designing safe, controllable agentic frameworks; developing rigorous benchmarks for causal evaluation; integrating heterogeneous data sources; and synergistically combining LLMs with structured knowledge (KGs) and formal causal inference tools. Such agents could unlock transformative opportunities, including accelerating drug discovery through automated hypothesis generation and simulation, enabling personalized medicine through patient-specific causal models. This research agenda aims to foster interdisciplinary efforts, bridging causal concepts and foundation models to develop reliable AI partners for biomedical progress.