A watershed study has quantified the extraordinary influence of social media algorithms on political polarization. Researchers successfully demonstrated that minimal adjustments to X users’ content feeds can compress three years’ worth of growing political division into a single week, revealing the platform’s enormous power to shape democratic discourse.
The experiment took place during the contentious 2024 presidential election, a period marked by viral spread of manipulated images and inflammatory content. Scientists from multiple prestigious universities collaborated to test whether algorithmic manipulation could measurably affect users’ political attitudes. They employed artificial intelligence to assess the divisiveness of posts in real-time, then adjusted what appeared in participants’ feeds accordingly.
Over 1,000 X users participated in the study without knowing their feeds were being modified. One group received slightly more posts containing antidemocratic sentiments, partisan hostility, opposition to bipartisan solutions, and distorted political information. Another group saw fewer such posts. The adjustments were subtle enough that participants generally didn’t notice them, yet the psychological impact was profound.
After just one week, researchers measured participants’ attitudes using a “feeling thermometer” that captured their warmth or coldness toward political opponents. Those exposed to more divisive content showed a significant increase in negative feelings—more than two degrees on the 100-point scale. This matches the amount of polarization that accumulated across American society between 1978 and 2020, demonstrating the algorithm’s power to accelerate societal trends.
The study also found that reducing exposure to divisive content decreased polarization by a similar magnitude, suggesting platforms could actively promote political harmony if they chose to do so. While overall engagement measured by time spent and posts viewed decreased slightly when divisive content was down-ranked, users actually engaged more deeply through likes and reposts. This trade-off poses a dilemma for platforms whose business models depend on maximizing engagement, but it also offers a practical pathway toward reducing the harmful societal consequences of politically divisive content.