The Prosocial Ranking Challenge: Trade-Offs in Designing Social Media Algorithms

About this event

Recommendation algorithms play a central role in shaping the societal consequences of social media use. Yet, simply deactivating them yields limited prosocial benefits and often reduces user engagement. To address this, we conducted a browser-extension field experiment with 6,000 U.S. social media users to test the effects of alternative algorithms. Five treatment algorithms, designed through an international peer-reviewed competition, seamlessly re-ranked, hid, or inserted content and were compared against a control condition. Across an 18-week intervention, the treatments collectively reduced affective polarization without decreasing engagement. Effects varied substantially by algorithm and platform, underscoring the importance of design choices and tailoring interventions to specific platforms. Our findings demonstrate that alternative algorithmic designs can improve societal outcomes without undermining user engagement.