Our paper on "An Empirical Investigation of Bias and Unfairness in Case-Based Adaptation" addresses an increasingly pressing question in artificial intelligence development: When systems encounter bias against protected groups (whether from real-world conditions or training data) do they merely reproduce these patterns, mitigate them, or potentially amplify them?

Our research specifically targets the "reuse" step within Case-Based Reasoning (CBR) systems, focusing on case-based adaptation in k-nearest neighbor classification tasks, i.e. in a methodology commonly employed in automated decision-making across various sectors.

The investigation highlights relations to earlier work addressing similar concerns during the retrieve and retain phases of CBR systems, now extending analysis to the stage where practical adaptations occur. Our key findings suggest that architectural choices made by developers significantly impact whether bias propagates through digital systems unchanged, gets reduced, or intensifies downstream applications. This raises questions about accountability for decisions made by technologies whose internal workings may obscure exactly how such outcomes materialize.

The 34th International Conference on Case-Based Reasoning (ICCBR-26) will be held in Bremen, Germany, between 13-16 August 2026. ICCBR-26 will be co-located with ECAI/IJCAI 2026 and hosted and organized at the University of Bremen. You can find the full paper here.