AI & ML Paradigm Shift

A massive field study (9,000+ users) proves that algorithmic shifts can reduce affective polarization without sacrificing user engagement.

March 23, 2026

Original Paper

The Prosocial Ranking Challenge: Reducing Polarization on Social Media without Sacrificing Engagement

Jonathan Stray, Ian Baker, George Beknazar-Yuzbashev, Ceren Budak, Julia Kamin, Kylan Rutherford, Mateusz Stalinski, Tin Acosta, Chris Bail, Michael Bernstein, Mark Brandt, Amy Bruckman, Anshuman Chhabra, Soham De, Kayla Duskin, Sara Fish, Beth Goldberg, Andy Guess, Dylan Hadfield-Menell, Muhammed Haroon, Safwan Hossain, Michael Inzlicht, Gauri Jain, Yanchen Jiang, Alexander P. Landry, Yph Lelkes, Hongfan Lu, Peter Mason, Jennifer McCoy, Smitha Milli, Paul Resnick, Emily Saltz, Martin Saveski, Lisa Schirch, Max Spohn, Siddarth Srinivasan, Alexis Tatore, Luke Thorburn, Joshua A. Tucker, Robb Willer, Magdalena Wojcieszak, Manuel Wüthrich, Sylvan Zheng

arXiv · 2603.19626

The Takeaway

This is the first large-scale empirical evidence challenging the industry assumption that 'prosocial' or 'bridging' content necessarily conflicts with engagement-driven business models. It shows that specific ranking changes can improve societal outcomes while maintaining (or even increasing) time spent on platforms like X/Twitter.

From the abstract

We report the first direct comparisons of multiple alternative social media algorithms on multiple platforms on outcomes of societal interest. We used a browser extension to modify which posts were shown to desktop social media users, randomly assigning 9,386 users to a control group or one of five alternative ranking algorithms which simultaneously altered content across three platforms for six months during the US 2024 presidential election. This reduced our preregistered index of affective po

Read the original paper →

← Back to today's papers