Authors: Andrii Kliachkin, Eleni Psaroudaki, Jakub Marecek, Dimitris Fotakis

Published on: March 28, 2024

Impact Score: 7.6

Arxiv code: Arxiv:2403.19419

Summary

• What is new: A randomized method for post-processing rankings without needing the protected attribute.
• Why this is important: Lack of fairness in machine learning for ranking-related problems, with the complication of not always having the protected attribute and having multiple measures of fairness.
• What the research proposes: A new post-processing method that randomizes rankings to improve fairness without requiring protected attribute information.
• Results: The method showed robustness with respect to P-Fairness and improved effectiveness in terms of Normalized Discounted Cumulative Gain (NDCG) compared to existing methods.

Technical Details

Technological frameworks used: Randomized post-processing

Models used: P-Fairness and NDCG

Data used: Extensive numerical study

Potential Impact

Online advertising, recommender systems, HR automation industries could benefit or be disrupted.

Want to implement this idea in a business?

We have generated a startup concept here: EquiRank.