Sufficient Cause Learning


Disease processes are complex, and there is a need to move beyond single-factor epidemiology towards an epidemiological framework that incorporates this complexity. Artificial intelligence methods provide us with an opportunity to rethink traditional epidemiological methods for etiological research. We are developing ‘Sufficient Cause Learning’ to bridge the epidemiological theory of sufficient causes and causal inference with a novel machine learning approach for interpreting artificial neural networks. With this approach, we may come up with new hypotheses of disease etiology. Such hypotheses would need to be rigorously tested using causal models in new cohorts, and hopefully, some of these generated hypotheses of complex interacting causes may lead us to new or more effective and targeted health interventions.

Contact person: Andreas Rieckmann (