Work

Bias Mitigation in Hiring-Related NLP: Interactions Between Masking, Rewriting, and Adversarial Debiasing

Year: 2026

Type: conference-paper

Abstract: AI-driven language technologies are increasingly used in hiring, but they may encode and reproduce harmful social stereotypes.Prior work often studies bias mitigation methods in isolation and outside realistic application settings.We examine the combined effects of data-level and model-level debiasing in a hiringrelated context, using Norwegian-language academic bios and a proxy STEM/non-STEM classification task.Specifically, we study masking sensitive information, GenWriter-based rewrites (Soundararajan and Delany, 2025), and adversarial debiasing (Han et al., 2021(Han et al., , 2022)).We evaluate these interventions using downstream task performance, group fairness metrics, intrinsic bias tests based on WEAT (Caliskan et al., 2017), and measures of gender leakage from hidden representations.We find that combining masking, GenWriter rewrites, and adversarial debiasing substantially reduces gender leakage while maintaining or improving downstream performance.However, effects on fairness gaps and intrinsic bias are mixed, underscoring the need for downstream, contextsensitive evaluation of bias mitigation methods in hiring-related NLP.

Source: International Conference on Natural Language Processing for Digital Humanities (NLP4DH)

DOI: 10.18653/v1/2026.nlp4dh-1.19

Authors: Alexandre Puttick, Rami El-Wazzi

Institution: Bern University of Applied Sciences

Language: English

Topic: Ethics and Social Impacts of AI

Subfield: Safety Research

Field: Social Sciences

Domain: Social Sciences

SDG: Reduced inequalities

Open Access status: hybrid

Funders: European Commission, Staatssekretariat für Bildung, Forschung und Innovation