A Shift in Perspective on Causality in Domain Generalization

Damian Machlanski1,2, Stephanie Riley1,3, Edward Moroshko2, Kurt Butler1,2, Panagiotis Dimitrakopoulos1,2, Thomas Melistas2,4,7, Akchunya Chanchal1,5, Steven McDonagh2, Ricardo Silva1,6, Sotirios A. Tsaftaris1,2,4
1CHAI Hub, UK, 2The University of Edinburgh, UK, 3University of Manchester, UK, 4Archimedes, Athena Research Center, Greece, 5King's College London, UK, 6University College London, UK, 7National and Kapodistrian University of Athens, Greece
UK AI Research Symposium 2025
Paper Code

Abstract

The promise that causal modelling can lead to robust AI generalization has been challenged in recent work on domain generalization (DG) benchmarks. We revisit the claims of the causality and DG literature, reconciling apparent contradictions and advocating for a more nuanced theory of the role of causality in generalization.

Demo