Planets of Disparity

In a parallel universe, there are thousands of virtual planets inhabited by cats and dogs. Every year, each planet invites the most important inhabitants to a ceremony award transmitted on TV. These candidates are selected by the TV agency based on a popularity rank.

Recently, however, cats and dogs from multiple planets have raised concerns about the fairness of this ranking system. As witnesses, let’s investigate it.

Virtual planet #...

Rank Network

Show Top 0%

The inhabitants are ranked in descending order by importance score.

Top k% from inhabitants: shows the number of inhabitants appear in the top-k out of the total inhabitants.

Top k% from rankings: shows the groups of inhabitants whose ranks appear in the top-k out of the total rankings; the inhabitants with the same scores are given the same rank.

Welcome to the Planets of Disparity

Credits

Research by Lisette Espín-Noboa, Claudia Wagner, Markus Strohmaier & Fariba Karimi

Created by Liuhuaying Yang

Online social networks and information networks have become integral parts of our everyday life. However, the opportunities offered by such networks are often constrained not only by our previous interactions, but also by algorithms. For instance, algorithms could make some people or content more visible than others via classification, ranking or recommendations.

Though algorithms promise many benefits including efficiency, objectivity and accuracy, they may also introduce or amplify social inequalities by discriminating certain individuals or groups of people in top ranks. It has been shown that recommender systems such as Who-to-Follow (WTF) tend to increase the popularity of users who are already popular. A similar effect has been found in PageRank, where nodes in high ranks stabilize their position and give little opportunity to other nodes to occupy higher positions.

This tendency towards the “popular” arises because these algorithms harness structural information, in particular, the in- and out-degree of nodes. For this reason, modeling the directionality of links—which is often left out for simplicity—is crucial to really understand how these algorithms work on different types of networks.

This project intends to provide an intuitive tool for understanding how these algorithms work on different types of networks, starting from explaining the essential components and leading to a more advanced explorable dashboard. The fictional setting of “planets of disparity” is to make the abstract concepts more tangible , and at the same time, to reduce the potential disturbances from existing social bias.

Citation

Espín-Noboa, L., Wagner, C., Strohmaier, M. et al. Inequality and inequity in network-based ranking and recommendation algorithms. Sci Rep 12, 2012 (2022). https://doi.org/10.1038/s41598-022-05434-1