Bias-Aware Hierarchical Clustering for detecting the discriminated groups of users in recommendation systems
作者：
Highlights：
• Unsupervised clustering methods may detect user disparities in recommendations.
• Bias-Aware Hierarchical Clustering detects more severe discrimination than other methods for different recommenders on two datasets.
• The method does not require information about sensitive attributes or user characteristics.
• A post-hoc model trained with descriptive user features provides explanations of these groups.
• The method may be used to design and evaluate fair recommender systems.
摘要
•Unsupervised clustering methods may detect user disparities in recommendations.•Bias-Aware Hierarchical Clustering detects more severe discrimination than other methods for different recommenders on two datasets.•The method does not require information about sensitive attributes or user characteristics.•A post-hoc model trained with descriptive user features provides explanations of these groups.•The method may be used to design and evaluate fair recommender systems.
论文关键词：Recommender system,System fairness,Bias detection,Model interpretability,Collaborative filtering
论文评审过程：Received 27 August 2020, Revised 30 December 2020, Accepted 14 January 2021, Available online 2 February 2021, Version of Record 2 February 2021.
论文官网地址：https://doi.org/10.1016/j.ipm.2021.102519
