Hate in the Time of Algorithms: Evidence on Online Behavior from a Large-Scale Experiment

Abstract

Social media algorithms are thought to amplify variation in user beliefs, thus contributing to radicalization. However, quantitative evidence on how algorithms and user preferences jointly shape harmful online engagement is limited. I conduct an individually randomized experiment with 8 million users of an Indian TikTok-like platform, replacing algorithmic ranking with random content delivery. Exposure to "toxic" posts decreases by 27%, mainly due to reduced platform usage by users with higher interest in such content. Strikingly, these users increase engagement with toxic posts they find. Survey evidence indicates shifts to other platforms. Model-based counterfactuals highlight the limitations of blanket algorithmic regulation.