Inequality and Inequity in Network-based Ranking and Recommendation Algorithms

Lisette Espín-Noboa, Claudia Wagner, Markus Strohmaier, Fariba Karimi

Introduction

Network algorithms, such as PageRank, are designed to identify "importance" based on connectivity. However, these "rich-get-richer" mechanisms can systematically disadvantage minority groups. This research explores how the interaction between homophily (the tendency to connect with similar others) and preferential attachment (the tendency for popular nodes to get more links) creates a "glass ceiling" effect in digital spaces.

Defining the Problem: Inequality vs. Inequity

The authors distinguish between two critical types of algorithmic unfairness:

  1. Inequality: Refers to the gap between the "most important" and "least important" individuals, typically characterized by a highly skewed score distribution where a few capture most of the visibility.
  2. Inequity: Refers to the gap between a group’s actual size in a population and its representation in top-tier rankings. This is a measure of group-level unfairness.
A classification framework for distinguishing between individual inequality and group inequity in rank distributions.

Methodology: The DPAH Model

To study these effects, the researchers introduced the Directed Preferential Attachment with Homophily (DPAH) model. This generative model allows for the systematic tuning of network properties:

  • Minority Fraction ($f_m$): The relative size of the minority group.
  • Homophily ($h$): The probability of nodes connecting to others within the same group.
  • Activity ($\gamma$): The skewness of the out-degree distribution (how many links individuals send).
  • Preferential Attachment: The mechanism where nodes with high in-degrees are more likely to receive new incoming links.

By applying PageRank and "Who-to-Follow" (WTF) algorithms to networks generated by DPAH, the authors could measure how social biases translate into algorithmic disparities.

Visualizing the impact of network homophily on ranking inequality and group inequity across different scenarios.

Key Insights

The Majority Holds the Key

A central finding is that a minority group’s visibility is heavily dependent on the behavior of the majority. If the majority group is homophilic (preferring its own members), the minority is consistently under-represented in top ranks. Conversely, if the majority is heterophilic (preferring the minority), the minority can achieve over-representation.

Inequality and Inequity are Distinct

While both are present, they are driven by different factors. Inequality (individual skewness) is primarily a product of preferential attachment and network density. Inequity (group representation), however, is driven by the interplay between homophily and the size of the minority group. Regression analysis showed that homophily within the majority group explained 61% of the global inequity in PageRank.

The Persistence of Bias

The research highlights that increasing the "selection pool" (e.g., looking at the top 20% instead of the top 5%) does not necessarily mitigate inequity. If an algorithm is biased at the very top, that bias tends to persist deep into the rankings.

Conclusion and Future Directions

The study demonstrates that ranking algorithms do not just identify importance; they amplify existing social biases. While the DPAH model provides a robust framework for understanding these dynamics, it currently assumes binary attributes and uniform homophily. Future work is needed to address the complexities of intersectionality and to develop algorithmic interventions—such as strategic recommendations—that can balance the scales before ranking occurs.

Summary updated

Abstract

Though algorithms promise many benefits including efficiency, objectivity and accuracy, they may also introduce or amplify biases. Here we study two well-known algorithms, namely PageRank and Who-to-Follow (WTF), and show to what extent their ranks produce inequality and inequity when applied to directed social networks. To this end, we propose a directed network model with preferential attachment and homophily (DPAH) and demonstrate the influence of network structure on the rank distributions of these algorithms. Our main findings suggest that (i) inequality is positively correlated with inequity, (ii) inequality is driven by the interplay between preferential attachment, homophily, node activity and edge density, and (iii) inequity is driven by the interplay between homophily and minority size. In particular, these two algorithms reduce, replicate and amplify the representation of minorities in top ranks when majorities are homophilic, neutral and heterophilic, respectively. Moreover, when this representation is reduced, minorities may improve their visibility in the rank by connecting strategically in the network. For instance, by increasing their out-degree or homophily when majorities are also homophilic. These findings shed light on the social and algorithmic mechanisms that hinder equality and equity in network-based ranking and recommendation algorithms.
BibTeX
@inproceedings{espinnoboa2022inequality,
  author = {Lisette Espín-Noboa and Claudia Wagner and Markus Strohmaier and Fariba Karimi},
  title = {{Inequality and Inequity in Network-based Ranking and Recommendation Algorithms}},
  booktitle = {Scientific},
  year = {2022},
  doi = {10.1038/s41598-022-05434-1},
}
Figure