Tiny Feed Changes on X Trigger Massive Political Division Surge

Minor modifications to social media algorithms can produce dramatic increases in political polarization at unprecedented speed, according to groundbreaking new research. Scientists found that barely perceptible adjustments to X users’ feeds generated as much political animosity in seven days as historically developed over three years, demonstrating the platform’s extraordinary influence on democratic society.
The research team conducted a carefully controlled experiment involving more than 1,000 users during the 2024 presidential election campaign. Using artificial intelligence, they analyzed posts for divisive content in real-time and then manipulated what appeared in users’ “for you” feeds. Some participants saw marginally more posts expressing antidemocratic views and partisan hatred, while others saw fewer such posts. Crucially, these changes were so subtle that most users didn’t realize their feeds had been altered.
The platform has faced scrutiny for its role in spreading politically divisive content, including fake images and AI-generated propaganda that garnered millions of views during the election. Since its acquisition and rebranding, X introduced algorithmic curation that prioritizes engagement-maximizing content over posts from accounts users actively follow. This shift has intensified concerns about the platform’s impact on political discourse.
Repeated exposure to posts expressing antidemocratic attitudes and partisan animosity significantly influenced users’ feelings of polarization and increased their levels of sadness and anger. The research team measured these effects using a novel approach that asked participants to rate their feelings toward political opponents on a 0 to 100 degree scale. Those exposed to more divisive content showed increased hostility of more than two degrees—equivalent to the polarization increase that occurred over four decades in American society.
The findings carry important implications for addressing political division. While social media platforms have long been suspected of promoting divisive content to boost engagement and advertising revenue, this study proves they could just as easily reduce polarization through algorithmic adjustments. The research found that while down-ranking divisive content slightly reduced some engagement metrics, users actually showed higher rates of meaningful interaction. This suggests platforms face a practical trade-off between maximizing short-term engagement volume and mitigating harmful societal consequences.