User-oriented Fairness in Recommendation

Abstract

As a highly data-driven application, recommender systems could be affected by data bias, resulting in unfair results for different data groups, which could be a reason that affects the system performance. Therefore, it is important to identify and solve the unfairness issues in recommendation scenarios.In this paper, we address the unfairness problem in recommender systems from the user perspective. We group users into advantaged and disadvantaged groups according to their level of activity, and conduct experiments to show that current recommender systems will behave unfairly between two groups of users. Specifically, the advantaged users (active) who only account for a small proportion in data enjoy much higher recommendation quality than those disadvantaged users (inactive). Such bias can also affect the overall performance since the disadvantaged users are the majority. To solve this problem, we provide a re-ranking approach to mitigate this unfairness problem by adding constraints over evaluation metrics. The experiments we conducted on several real-world datasets with various recommendation algorithms show that our approach can not only improve group fairness of users in recommender systems, but also achieve better overall recommendation performance. CCS CONCEPTS• Information systems → Recommender systems; • Computing methodologies → Artificial intelligence.

The Unfairness of Active Users and Popularity Bias in Point-of-Interest Recommendation

“…Fairness is becoming one of the most influential topics in recommender systems in recent years [7,27,22]. Burke et al [6] classified fairness in the recommendation system based on general beneficiaries, consumers (C), providers (P), and both (CP).…”

“…Deldjoo et al [9] proposed a flexible framework to evaluate consumer and provider fairness using generalized cross-entropy. Li et al [22] focused on the fairness in the recommendation systems from the user's perspective in the e-commerce domain, i.e., C-fairness. They created user groups based on their activity level into two groups: advantaged and disadvantaged.…”

“…Furthermore, they analyzed whether such algorithmic popularity bias affects users of different genders. Contrary to the works mentioned above that focus on consumers [22,3,2], or provider fairness [1,20,38], our work studies the fairness of recommended items from both perspectives (i.e., CP-Fairness). Mehrotra et al [26] showed that blindly optimizing for consumer relevance might hurt supplier fairness.…”

“…Traditional recommender systems (RS) focused on maximizing customer satisfaction by tailoring the content according to individual customers' preferences, thereby disregarding the interest of the producers. Several recent studies [22,2] have revealed how such customer-centric designs may impair the well-being of the producers. Thus, in recent years, two-sided marketplaces have steadily emerged as a pivotal problem of fairness topics in RS, requiring optimizing supplier preferences and visibility.…”

Consumer Fairness in Recommender Systems: Contextualizing Definitions and Mitigations

“…As fairness is an abstract concept, an abundance of consumer fairness notions have been proposed, along with algorithmic procedures for mitigating unfairness in recommendations according to the proposed notions. Examples of mitigation procedures have been applied in pre-processing [13], by transforming the input data, in-processing [18,6,15,28], by constraining the training process of stateof-the-art models, and post-processing [19,23,3], by ranking again the originally recommended items. Moreover, the evaluation protocol adopted to assess their impact has been often heterogeneous (e.g., different data sets, train-test splits) and limited to showing that the proposed mitigation is better than doing nothing, making the landscape convoluted.…”