What is it about?
Social media platforms decide which content users see and in what order. We studied how two common features of these systems, personalization and giving extra weight to active engagement such as likes and shares, can affect the news people choose to read. We first measured how people interact with ranked news and used these observations to build a simple model of the feedback between users and rankings. We then tested the model in an experiment where the rankings changed in response to the actions of real participants. We found that combining personalization with rewards for active engagement shifted news consumption toward more politically extreme and like-minded content.
Featured Image
Photo by Piotr Cichosz on Unsplash
Why is it important?
It is difficult to determine how ranking algorithms affect users because real social media platforms are complex and their algorithms are rarely open to researchers. Our study provides a controlled way to isolate and test one important feedback mechanism. In our human experiment, changing from a neutral ranking to one that combined personalization with a strong reward for active engagement caused users to consume more extreme and like-minded news. The model also helps explain how these effects emerge and predicts how changing the strength of personalization or engagement rewards can alter the outcome. This provides evidence that seemingly common choices in the design of ranking algorithms can systematically shape the political information people consume.
Perspectives
What I find most interesting about this work is that it brings together a simple mathematical model and experiments with real people to study a feedback mechanism that is usually very difficult to observe directly. Social media algorithms are often discussed in broad terms, but their effects emerge from repeated interactions between users and the rankings they see. I particularly like that in this study we could first measure some of those behavioral tendencies, use them to make predictions, and then test those predictions in a setting where the ranking actually evolved in response to participants' actions. I hope the paper helps make these feedback effects easier to understand and study experimentally.
Vicenç Gómez
Universitat Pompeu Fabra
Read the Original
This page is a summary of: Rewarding Engagement and Personalization in Popularity-Based Rankings Amplifies Extremism and Polarization, August 2026, ACM (Association for Computing Machinery),
DOI: 10.1145/3770855.3818037.
You can read the full text:
Contributors
The following have contributed to this page
