A1994
Title: Distributional intersectional fairness in AI-supported job matching
Authors: Sabrina Muehlbauer - Institute for Employment Research (Germany) [presenting]
Enzo Weber - University of Regensburg and Institute for Employment Research (Germany)
Paula Ziethmann - Institute for Employment Research (Germany)
Abstract: Most algorithmic fairness evaluations remain limited to single-attribute group comparisons or individual-level prediction errors. A distributional approach to intersectional fairness is introduced that evaluates how machine learning models reshape the allocation of outcomes across intersecting social groups. Using large-scale administrative labour-market data from public employment services, an AI-supported job matching system is analysed. Intersectional disparities are measured via Jensen-Shannon divergence and observed occupational distributions are compared to model-implied allocations. Results show that unconstrained models do not merely reproduce existing labour-market structures but systematically amplify intersectional disparities across occupations. Fairness-aware regularization reduces aggregate disparities but introduces a pronounced trade-off with predictive performance and reallocates distortions across occupations rather than eliminating them. To identify localized fairness risks, an occupation-level audit mechanism is proposed that flags cases with unusually large distributional shifts. The findings demonstrate that fairness in large-scale decision-support systems must be evaluated at the level of outcome distributions. They highlight the limits of purely technical mitigation and the need to combine fairness-aware modelling with explainability, monitoring, and human oversight in real-world deployment.
