Authors: My H. Dinh, James Kotary, Ferdinando Fioretto

Published on: February 07, 2024

Impact Score: 8.12

Arxiv code: Arxiv:2402.05252

Summary

• What is new: This paper introduces a novel approach by integrating efficiently-solvable fair ranking models based on Ordered Weighted Average (OWA) functions into the training loop of LTR models, enabling backpropagation through constrained optimizations of OWA objectives.
• Why this is important: Conventional Learning to Rank (LTR) models often produce biased results, and existing fair LTR models either lack accuracy or efficiency.
• What the research proposes: The study proposes using optimization of OWA functions within the LTR model’s training loop to balance fairness, user utility, and runtime efficiency effectively.
• Results: The models demonstrated the ability to achieve a better balance between fairness, user utility, and efficiency compared to existing solutions.

Technical Details

Technological frameworks used: nan

Models used: Efficiently-solvable fair ranking models based on OWA functions

Data used: nan

Potential Impact

Job search platforms, healthcare information retrieval services, and social media content feed algorithms could all see significant improvements and disruptions from the insights of this paper.

Want to implement this idea in a business?

We have generated a startup concept here: RankWise.