[Paper Review] Fairness and Diversity in Recommender Systems: A Survey

Yuying Zhao, Yu Wang+4 more|arXiv (Cornell University)|Jul 10, 2023

Mind wandering and attention7 citations

TL;DR

A comprehensive survey linking fairness and diversity in recommender systems, expanding diversity to include user-level considerations and reinterpreting fairness through a diversity lens.

ABSTRACT

Recommender systems are effective tools for mitigating information overload and have seen extensive applications across various domains. However, the single focus on utility goals proves to be inadequate in addressing real-world concerns, leading to increasing attention to fairness-aware and diversity-aware recommender systems. While most existing studies explore fairness and diversity independently, we identify strong connections between these two domains. In this survey, we first discuss each of them individually and then dive into their connections. Additionally, motivated by the concepts of user-level and item-level fairness, we broaden the understanding of diversity to encompass not only the item level but also the user level. With this expanded perspective on user and item-level diversity, we re-interpret fairness studies from the viewpoint of diversity. This fresh perspective enhances our understanding of fairness-related work and paves the way for potential future research directions. Papers discussed in this survey along with public code links are available at https://github.com/YuyingZhao/Awesome-Fairness-and-Diversity-Papers-in-Recommender-Systems .

Motivation & Objective

• Introduce fairness and diversity concepts in recommender systems and motivate their joint study.
• Provide an expanded notion of diversity that includes user-level diversity.
• Explore the connections between fairness and diversity at both item and user levels.
• Review measurements and debiasing/diversification methods for fairness and diversity.
• Identify challenges and opportunities guiding future research.

Proposed method

• Survey the literature on fairness and diversity in recommender systems across item and user dimensions.
• Categorize fairness measurements (exposure, quality discrepancy, discriminator performance) and debiasing methods (pre-, in-, post-processing).
• Categorize diversity sources, measurements, and methods, and extend diversity concepts to user-level diversity.
• Analyze the interplay between fairness and diversity, showing how fairness can be viewed through a diversity lens.

Experimental results

Research questions

• RQ1 How are fairness and diversity defined and measured in recommender systems at both item and user levels?
• RQ2 What is the relationship between fairness and diversity, and how can diversity inform fairness strategies?
• RQ3 What methods exist to promote fairness and diversity across preprocessing, in-processing, and post-processing stages?
• RQ4 What are the key challenges and opportunities for integrating fairness and diversity in practice?

Key findings

• Fairness and diversity are strongly connected, and expanding diversity concepts to the user side clarifies this relationship.
• Item-level fairness often centers on popularity bias and exposure disparities, while user-level fairness focuses on quality discrepancy and embedding-based discrimination.
• Diverse strategies (pre-, in-, post-processing) can mitigate biases, and re-ranking methods offer flexible post-hoc fairness adjustments.
• Diversity enhancement can serve as a mechanism to improve fairness, and fairness strategies can be interpreted through a diversity perspective.
• The survey provides a unified framework and updated references to recent work, highlighting gaps and future directions.