Métriques d’équité en Apprentissage Automatique et droit de l’Union Europénne en matière de non-discrimination
(2023) Conférence Nationale en Intelligence Artificielle — Location: Strasbourg, France (3.July.2023)
Files
CNIA2023_Mesures_d_equite_et_droit_EU.pdf
Open Access - Adobe PDF
- 114.84 KB
Details
- Authors
Yousefi, Yasaman 
Koutsoviti Koumeri, Lisa Legay, Axel Vanhoof, Koen
- Abstract
- (en) Machine Learning (ML) models have been shown to present biases leading to discrimination against certain social groups. Our research studies the extend to which ML techniques and fairness definitions can ensure compliance with the EU non-discrimination legal framework. Using classification models trained with different fairness constraints, we evaluate how effective the bias mitigation process is and discuss the results using both an ML approach and legal informatics methodology.
- Affiliations
Citations
Legast, M., Yousefi, Y., Koutsoviti Koumeri, L., Legay, A., Schommer, C., & Vanhoof, K. (2023). Métriques d’équité en Apprentissage Automatique et droit de l’Union Europénne en matière de non-discrimination. Conférence Nationale en Intelligence Artificielle, Strasbourg, France.