AI & Causality

When analyzing large amounts of diverse medical data for clinical decision making, it is not only important to understand statistical relationships between factors, but also the causes and correlations that reveal something about biological processes. The use of models based on both data and cause-and-effect relationships can help to make these models better, more representative and more comprehensible.

Such models enable not only statistical but also causal relationships between different factors to be shown, learned and used in clinical decision-making. This applies to all kinds of clinical data as well as comprehensive biological data. In contrast to purely data-driven methods, these models offer clear insights into the dependencies of the observed factors. For example, in a predictive model, they can explain why certain events occur and thus make them easier to understand.

By working together with experts, these correlations can be checked and improved. This is particularly important in the medical field in order to integrate data from different sources and populations in a meaningful way. This involves different types of observations, experiments, therapies and measurements. Research is also focusing on how to make such models scalable and how to correct any biases in the data. Another exciting development is the combination of these causal models with modern approaches to machine learning and knowledge graphs, which is currently being driven forward in the SciKnow Cluster of Excellence.