Groundbreaking scientific research has quantified the extraordinary power of social media algorithms to generate political polarization at accelerated rates. An experiment demonstrated that minimal changes to X users’ feeds produced as much political division in seven days as would have naturally taken three years to develop, revealing platforms’ unprecedented influence over democratic societies and political culture.
The study brought together researchers from Stanford, Johns Hopkins, Northeastern, and the University of Washington to conduct a carefully controlled experiment. They developed a sophisticated system using artificial intelligence to classify posts based on divisiveness, then adjusted what appeared in the feeds of more than 1,000 participants during the 2024 presidential election. Some users received slightly more posts containing antidemocratic sentiments, partisan hostility, opposition to bipartisan cooperation, and biased political information, while others saw fewer such posts, all while keeping modifications imperceptible.
Assistant professor Martin Saveski emphasized that the algorithm’s power lies in its subtlety—participants reported significant differences in how they felt about political opponents despite barely perceptible changes to their feeds. Co-author Tiziano Piccardi noted that the shift corresponds to approximately three years of polarization based on trends observed in American political attitudes over recent decades. The platform has faced scrutiny for viral spread of manipulated content during the campaign, including fake and AI-generated images that received millions of views.
The measurement methodology involved participants rating their feelings toward opposing political parties on a “feeling thermometer” ranging from 0 to 100 degrees, assessing warmth or coldness, favorability or unfavorability. Those exposed to more divisive content showed increased hostility of more than two degrees on this scale—matching the polarization increase that accumulated across four decades from 1978 to 2020. The research also demonstrated that reducing exposure to posts with antidemocratic attitudes and partisan animosity decreased political division by a similar amount.
The implications extend far beyond academic interest. Survey research indicates that overwhelming majorities in democratic nations believe political opponents cannot agree even on basic facts, with many viewing current division levels as dangerous to society. This study proves that platforms possess the technical capacity to address this crisis through algorithmic redesign. While social media companies have been accused of promoting divisive content to maximize engagement and advertising revenue, the research found that down-ranking such content resulted in only slight decreases in overall engagement volume, while users actually engaged more meaningfully through likes and reposts. This suggests platforms face a practical trade-off between short-term engagement maximization and mitigating harmful societal consequences, but that pursuing social responsibility may be more compatible with viable business operations than commonly assumed.
