It's complicated. The relationship of algorithmic fairness and non-discrimination regulations for high-risk systems in the EU AI Act

Kristof Meding

Keywords: AI Act, algorithmic fairness, non-discrimination, legal regulations

TL;DR: The paper highlights the complexity of aligning legal non-discrimination standards with algorithmic fairness in the context of the EU AI Act, concluding that resolving these challenges requires nuanced, collaborative interdisciplinary approaches.

Abstract: What constitutes a fair decision? This question is not only difficult to answer for humans but becomes more challenging when Artificial Intelligence (AI) models are used. In light of problematic algorithmic outcomes, the EU has recently passed the AI Act, which mandates specific rules for (among others) high-risk systems. For high-risk systems, the AI Act might point towards traditional legal non-discrimination regulations and machine learning based algorithmic fairness concepts. This paper aims to bridge these two concepts as reflected in the AI Act by providing: (1) a high-level introduction of the AI Act targeting computer science-oriented scholars, and (2) an analysis of the relationship between the AI Act’s legal non-discrimination and algorithmic fairness provisions. Finally, we consider future steps for the application of non-discrimination law and the AI Act’s provisions. This paper serves as a foundation for future interdisciplinary collaboration between legal scholars and machine learning researchers with a computer science background by reducing the gap between these two fields.

Submission Number: 31