Addressing Marketing Bias in Product Recommendations

Abstract

Modern collaborative filtering algorithms seek to provide personalized product recommendations by uncovering patterns in consumerproduct interactions. However, these interactions can be biased by how the product is marketed, for example due to the selection of a particular human model in a product image. These correlations may result in the underrepresentation of particular niche markets in the interaction data; for example, a female user who would potentially like motorcycle products may be less likely to interact with them if they are promoted using stereotypically 'male' images.In this paper, we first investigate this correlation between users' interaction feedback and products' marketing images on two realworld e-commerce datasets. We further examine the response of several standard collaborative filtering algorithms to the distribution of consumer-product market segments in the input interaction data, revealing that marketing strategy can be a source of bias for modern recommender systems. In order to protect recommendation performance on underrepresented market segments, we develop a framework to address this potential marketing bias. Quantitative results demonstrate that the proposed approach significantly improves the recommendation fairness across different market segments, with a negligible loss (or better) recommendation accuracy.

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

“…Moreover, we observe that a primary challenge many models face is the unfairness of popularity bias. As for future work, we plan to extend our experiments on more datasets from different domains (e.g., marking in e-commerce [34], media-streaming, e-fashion [13]) and models such as sessionbased [28], neural [8] and content-based systems [15]. Investigating the different type of bias, such as gender bias, can be a potential direction for future research.…”

Facets of Fairness in Search and Recommendation

“…But it would be problematic if Spotify only recommends songs from a few popular artists to the consumers because: 1) it gives low exposure to less popular artists, and 2) the consumer may not find the recommendations interesting after sometime. There has been a few recent works discussing and addressing these concerns [18,5,24,11,4,8]. There are several classes of multi-sided recommendation: 1) Multi-receiver recommendation -when a target audience is a group of people rather than an individual, e.g., students on an education platform; 2) Multi-provider recommendation -when several suppliers provide the recommendation content, and the platform needs to choose between them, e.g., Airbnb and Spotify, and 3) Side stakeholder recommendation -when there are parties other than suppliers and consumers involved in the marketplace, the recommendations need to consider their preferences as well, e.g., drivers in the Uber Eats platform.…”

Fairness-Aware Explainable Recommendation over Knowledge Graphs

“…Farnadi et al [17] claim that no model can be fair in every aspect of metrics. Previous work has explored the fairness problem in recommendation from the perspective of selection aspects [21,33,35], marketing bias [36], popularity bias [42], multiple stakeholders [5] in terms of consumers and providers, among others. Existing research on fairness has shown that protected groups 1 , defined as the population of vulnerable individuals in terms of sensitive features such as gender, age, race, religion, etc., are easily treated in a discriminatory way.…”