On fair selection in the presence of implicit and differential variance

DOI10.1016/J.ARTINT.2021.103609MaRDI QIDQ2060744zbMATH OpenOpenAlexFDO

Authors Vitalii Emelianov, Nicolas Gast, Krishna P. Gummadi, Patrick Loiseau

Publication date 13 December 2021

Published in Artificial Intelligence (Search for Journal in Brave)

Full work available at URL https://arxiv.org/abs/2112.05630

zbMATH Keywords

differential variancefairnessselection problem

Mathematics Subject Classification ID

Matching models (91B68)

Abstract: Discrimination in selection problems such as hiring or college admission is often explained by implicit bias from the decision maker against disadvantaged demographic groups. In this paper, we consider a model where the decision maker receives a noisy estimate of each candidate's quality, whose variance depends on the candidate's group -- we argue that such differential variance is a key feature of many selection problems. We analyze two notable settings: in the first, the noise variances are unknown to the decision maker who simply picks the candidates with the highest estimated quality independently of their group; in the second, the variances are known and the decision maker picks candidates having the highest expected quality given the noisy estimate. We show that both baseline decision makers yield discrimination, although in opposite directions: the first leads to underrepresentation of the low-variance group while the second leads to underrepresentation of the high-variance group. We study the effect on the selection utility of imposing a fairness mechanism that we term the $gamma$-rule (it is an extension of the classical four-fifths rule and it also includes demographic parity). In the first setting (with unknown variances), we prove that under mild conditions, imposing the $gamma$-rule increases the selection utility -- here there is no trade-off between fairness and utility. In the second setting (with known variances), imposing the $gamma$-rule decreases the utility but we prove a bound on the utility loss due to the fairness mechanism.

Recommendations

• Impartial selection with additive approximation guarantees
• Impartial selection with additive approximation guarantees
• Selection problems in the presence of implicit bias
• A near-optimal mechanism for impartial selection
• On the fairness of random procedures
• Optimization hierarchy for fair statistical decision problems
• Fairness in selecting multiple objects under diversity requirements
• Achieving fairness in the stochastic multi-armed bandit problem
• On asymptotic fairness in voting with greedy sampling
• Fairness and learning: An experimental examination

Cites work

• Power-law distributions in empirical data
• Selection problems in the presence of implicit bias

Cited in

(3)

• scientific article; zbMATH DE number 1189335 (Why is no real title available?)
• Selection problems in the presence of implicit bias
• FAC-fed: federated adaptation for fairness and concept drift aware stream classification

This page was built for publication: On fair selection in the presence of implicit and differential variance

Report a bug (only for logged in users!)Click here to report a bug for this page (MaRDI item Q2060744)