From prediction to intervention: Paradigm shifts in causal AI for precision medicine and large‐scale cohorts

Abstract In precision medicine, the rapid expansion of large‐scale databases and routinely collected clinical data has enabled the development of increasingly accurate predictive models utilizing artificial intelligence. However, predictive accuracy does not guarantee the provision of actionable guidance for clinical interventions. In this review, we examine causal AI, a sophisticated tool that integrates causal inference principles with contemporary machine learning methodologies. This integration signifies a paradigm shift from association‐centric modeling toward decision‐relevant, intervention‐oriented evidence. We commence with an exposition of the methodology for constructing causal models from cohort data, accompanied by a thorough elucidation of the technical concepts underpinning this process. Furthermore, how to leverage causal models to mine the causal relationships is provided, as well as real examples of translational applications of causal AI results in clinical practice. Finally, we outline existing challenges and possible future research directions and applications of causal AI.