New Publication in Nature Computational Science

In a recent Comment published in Nature Computational Science, Christoph Kern, Unai Fischer Abaigar and Frauke Kreuter of the SODA lab, jointly with colleagues of the Munich School of Management at LMU, Carnegie Mellon University and the University of Cambridge, argue that reliable algorithmic decision-making (ADM) systems need to be grounded in causal reasoning.

The reason is simple: ADM systems don’t just predict outcomes, they change them.
If we want our models to be meaningful in the real world, they must understand and model cause-and-effect relationships.The Comment shows that identifiability is key, which means that we can express the causal effect of an intervention using observed data. Without identifiability, we don't know whether we measure the right thing. Identifiability and causal assumptions are critical for building reliable and robust algorithms that are aligned with decision-making objectives in real-world tasks.
The paper further highlights current research avenues to extend ADM, including promising directions around uncertainty quantification, robustness, performativity, evaluation and benchmarking. This will help ADM systems to adapt safely across environments and make sound decisions under real-world constraints.

Link: https://www.nature.com/articles/s43588-025-00814-9

Free access: https://rdcu.be/enVCs