Individually Fair Ranking

Abstract

We develop an algorithm to train individually fair learning-to-rank (LTR) models. The proposed approach ensures items from minority groups appear alongside similar items from majority groups. This notion of fair ranking is based on the definition of individual fairness from supervised learning and is more nuanced than prior fair LTR approaches that simply ensure the ranking model provides underrepresented items with a basic level of exposure. The crux of our method is an optimal transport-based regularizer that enforces individual fairness and an efficient algorithm for optimizing the regularizer. We show that our approach leads to certifiably individually fair LTR models and demonstrate the efficacy of our method on ranking tasks subject to demographic biases.

Fairness in Recommendation: A Survey

“…In such setting, the fairness metrics are usually relevant to the exposure of the candidates belonging to different groups, so that the average exposure of groups can be controlled to be proportional to their average relevance to the search query. Different from probability-based fairness notions, the exposure/attention-based fairness can quantify not only group fairness [135,167], but also individual fairness [18,26] according to specific formalization. Group fairness can be measured through the difference of the average exposure between different groups 𝐺 1 and 𝐺 2 [209]:…”