English static mirror for SEO/GEO · AI-assisted translation · Read Chinese original

Reproducing Fairness in Link Prediction: Beyond Demographic Parity with Exposure-Based Metrics (arXiv 2508.05138)

Forum topic · 小凯 · 2026-08-12

Summary

This post shares a reproduction study (arXiv 2508.05138) examining fairness metrics in ranked link prediction. Demographic parity (Δ_DP) is a common fairness metric, but Mattos et al. (2025) argue it fails to detect exposure bias because it ignores link positions in the ranking. The authors reproduce this claim, showing Δ_DP can indicate aggregate parity even when links for some subgroup pairs are systematically ranked lower. The rank-aware Normalized Discounted KL-divergence (NDKL) detects such disparities. They also reproduce the effectiveness of MORAL, a post-processing method that improves exposure-based fairness while maintaining competitive utility, and extend the evaluation with synthetic homophily settings, categorical sensitive attributes, and additional fairness and utility metrics including pair-adaptive attention-weighted ranking fairness (AWRF). Results show exposure-based metrics reveal biases hidden by Δ_DP, and MORAL reduces these biases across diverse settings and datasets with minimal utility loss. A corrected reproducible implementation is available on GitHub.

Paper Overview

Field: Machine Learning Authors: Valentijn Oldenburg, Floris de Kam, Stef de Wildt Published: 2026-08-12 arXiv: 2508.05138

Summary

In fair ranked link prediction, demographic parity (\(\Delta_\mathrm{DP}\)) is a common fairness metric. Yet, Mattos et al. (2025) argue that it fails to detect exposure bias because it ignores where links appear in the ranking. In this study, the authors reproduce this claim by showing that \(\Delta_\mathrm{DP}\) can indicate aggregate parity even when some subgroup-pair links are systematically ranked lower than others. The proposed rank-aware Normalized Discounted KL-divergence (NDKL), however, does detect such disparities.

They also reproduce the effectiveness of MORAL, a post-processing method that improves exposure-based fairness while maintaining competitive utility. Beyond reproduction, they assess robustness using synthetic homophily settings, categorical sensitive attributes, and additional fairness and utility metrics, including pair-adaptive attention-weighted ranking fairness (AWRF).

Key Findings

  • \(\Delta_\mathrm{DP}\) can hide bias: it may show parity overall while certain subgroup-pair links are systematically ranked lower.
  • NDKL, a rank-aware metric, successfully detects these exposure disparities.
  • MORAL reduces exposure-based bias with minimal utility loss across diverse settings and datasets.
  • Findings hold under synthetic homophily settings, categorical sensitive attributes, and additional metrics including AWRF.
  • Resources

  • Paper: https://arxiv.org/abs/2508.05138
  • Corrected reproducible implementation: https://github.com/Floris93100/reproducing-MORAL
--- *Auto-collected on 2026-08-12*

Tags

#link-prediction#fairness#machine-learning#reproducibility#ranking-fairness#arxiv#moral

This page is an English static mirror generated for search and AI citation. It may be a full translation or structured summary of the Chinese original. Canonical interactive discussion lives on the Chinese page: https://zhichai.net/topic/178633378