The Unfairness of Multifactorial Bias in Recommendation

TL;DR

Multifactorial bias, the interplay of popularity bias and positivity bias, degrades item exposure fairness in recommender systems. The authors simulate its impact and adopt a percentile-based rating transformation on item profiles to mitigate bias, reporting improvements in exposure fairness with negligible or favorable effects on accuracy across six algorithms and four datasets. They also show that this pre-processing can enhance the efficiency of existing post-processing fairness pipelines, enabling strong fairness with smaller initial lists. The work demonstrates that simple, data-driven pre-processing can meaningfully reduce multifactorial bias and improve practical fairness in recommender systems.

Abstract

Popularity bias and positivity bias are two prominent sources of bias in recommender systems. Both arise from input data, propagate through recommendation models, and lead to unfair or suboptimal outcomes. Popularity bias occurs when a small subset of items receives most interactions, while positivity bias stems from the over-representation of high rating values. Although each bias has been studied independently, their combined effect, to which we refer to as multifactorial bias, remains underexplored. In this work, we examine how multifactorial bias influences item-side fairness, focusing on exposure bias, which reflects the unequal visibility of items in recommendation outputs. Through simulation studies, we find that positivity bias is disproportionately concentrated on popular items, further amplifying their over-exposure. Motivated by this insight, we adapt a percentile-based rating transformation as a pre-processing strategy to mitigate multifactorial bias. Experiments using six recommendation algorithms across four public datasets show that this approach improves exposure fairness with negligible accuracy loss. We also demonstrate that integrating this pre-processing step into post-processing fairness pipelines enhances their effectiveness and efficiency, enabling comparable or better fairness with reduced computational cost. These findings highlight the importance of addressing multifactorial bias and demonstrate the practical value of simple, data-driven pre-processing methods for improving fairness in recommender systems.

Figures (13)

• Figure 1: Popularity distribution of the rating data in multiple datasets used in this paper. The horizontal axis shows the percentage of the items and the vertical axis shows the percentage of the accumulated ratings.
• Figure 2: Distribution of ratings of items in the MovieLens dataset. The horizontal axis shows the rating values in the dataset and the vertical axis shows the percentage of times each rating value is seen in the dataset.
• Figure 3: The relationship between average ratings and popularity of items in four public datasets.
• Figure 4: The relationship between the average ratings and popularity of items in four datasets after artificially mitigating for popularity-positivity bias.
• Figure 5: The performance of recommendation model on modified rating data in our simulation study on three datasets. Experiments are performed using BiasedMF algorithm.