Evaluating Provider-Fairness Mitigation in Price-Aware Recommendation
Abstract
Price-aware recommender systems incorporate economic signals into ranking, offering practical benefits but their implications for provider-side fairness remain largely unexplored. In this work, we present the first systematic evaluation of fairness mitigation in price-aware recommendation. We quantify provider-side exposure disparities across groups defined by continent, country, and price segment using three representative price-aware models on two datasets. Our analysis reveals systematic provider-side disparities, indicating that price-aware ranking can exhibit substantial provider imbalance. We then evaluate three post-processing re-ranking strategies (PFair, PFAR, and Calibration) applied to top-k recommendations to mitigate these disparities. Results show that mitigation effectiveness varies substantially across methods and settings. PFair achieves the strongest reductions for interaction-based exposure disparities, whereas PFAR is generally more effective under catalog-based reference distributions but exhibits substantial instability. Calibration remains comparatively conservative, producing only limited changes in disparity. These findings highlight the fairness risks of price-aware recommendation and show that the effectiveness of existing mitigation strategies is conditional on both the fairness reference distribution and the underlying ranking model.