Energy-Based Fairness Recommendation: A Novel Framework for Fairness Treatment in Recommender Systems

Fairness recommendation systems have become an increasingly important area of focus, as biased recommendations can negatively impact different user groups and product providers. We introduce the Energy Fairness Recommendation System (EFR), a novel approach using energy distance to measure and mitigate unfairness in recommendations. Unlike traditional statistical metrics, EFR provides a theoretically grounded framework for capturing complex distributional disparities across both users and items. Experiments on three real-world datasets show EFR reduces user group unfairness by 61 % and product distributional disparity by 44 %, while maintaining competitive accuracy (<tex xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink">$&lt;5 {\%}$</tex> NDCG reduction). Our approach effectively detects subtle biases traditional metrics miss, and scales efficiently to large applications through an optimized algorithm.