Provider fairness across continents in collaborative recommender systems

作者：

Highlights：

• We assess unfairness for groups of providers belonging to different geo-graphic continents, considering state-of-the-art recommendation models.

• We propose a multiclass re-ranking algorithm to introduce provider fairness following a notion of equity that distributes the recommendations according to the representation of the groups in the input data.

• We evaluate our algorithm in two recommendation domains and study its effectiveness at producing fair but effective recommendations.

摘要

•We assess unfairness for groups of providers belonging to different geo-graphic continents, considering state-of-the-art recommendation models.•We propose a multiclass re-ranking algorithm to introduce provider fairness following a notion of equity that distributes the recommendations according to the representation of the groups in the input data.•We evaluate our algorithm in two recommendation domains and study its effectiveness at producing fair but effective recommendations.

论文关键词：Recommender systems,Bias,Provider fairness,Geographic groups,Data imbalance,Disparate impact

论文评审过程：Received 14 April 2021, Revised 20 July 2021, Accepted 9 August 2021, Available online 12 October 2021, Version of Record 12 October 2021.

论文官网地址：https://doi.org/10.1016/j.ipm.2021.102719