This paper takes up the EU’s new rule on discriminatory credit underwriting. I build on earlier work, exploring how anti-discrimination law fares when fitting algorithmic credit scoring and creditworthiness evaluation into the regime of EU direct/indirect discrimination and US disparate treatment/disparate impact doctrine. I suggest that anti-discrimination law, when faced with redundant encoding, runs into doctrinal and practical problems. These manifest in proving but-for causation, in scenarios resembling algorithmic redlining, and in a more fundamental misfit between anti-discrimination law and big data analytics. The paper summarizes these challenges, addresses recent changes in US law, and submits that the new EU Directive has missed the chance to provide for a modern rule, fit to cope with algorithmic credit underwriting.
in: Spiecker, I./Burchard, C. (eds.): Algorithmic Transformation and Diffusion of Power: Trust, Conflict, Uncertainty and Control, pp. 61-90, Studien zum Datenschutz, Nomos,
2025
