Can One Embedding Fit All? A Multi-Interest Learning Paradigm Towards Improving User Interest Diversity Fairness

Yuying Zhao, Minghua Xu, Huiyuan Chen, Yuzhong Chen, Yiwei Cai, Rashidul Islam, Yu Wang, Tyler Derr

Keywords: Fairness, Diversity, Multi-Interest Recommendations

Abstract:

Recommender systems have gained widespread applications across various domains owing to their superior ability to understand and capture users' interests. However, the complexity and nuanced nature of users' interests, which can span a wide range of diversity, pose a significant challenge in delivering fair recommendations. In real-world scenarios, user preferences vary significantly; some users show a clear preference toward certain item categories, while others have a broad interest in diverse ones. Even though it is expected that all users should receive high-quality recommendations, the effectiveness of recommender systems in catering to this disparate interest diversity remains under-explored.
In this work, we investigate whether users in different groups with varied levels of interest diversity are treated fairly. Our empirical experiments reveal an inherent disparity: users who have a wider range of interests often receive lower-quality recommendations. To achieve fairer recommendations, we propose a multi-interest framework that uses multiple (virtual) interest embeddings, rather than the utilization of single embedding to represent individual users. Specifically, the framework consists of stacked multi-interest representation layers. Each layer includes an interest embedding generator that derives virtual interests from globally shared interest parameters, and a center embedding aggregator that facilitates multi-hop aggregation. The experiments have demonstrated the effectiveness of the proposed method in achieving better trade-off between fairness and utility across various datasets and backbones. Our code and datasets are available at: https://anonymous.4open.science/r/User-Interest-Diversity-Fairness-4BBE/.

Track: Responsible Web

Submission Guidelines Scope: Yes

Submission Guidelines Blind: Yes

Submission Guidelines Format: Yes

Submission Guidelines Limit: Yes

Submission Guidelines Authorship: Yes

Student Author: Yes

Submission Number: 2089