New research exposes the alarming speed at which social media algorithms can intensify political animosity between opposing groups. A controlled experiment on X demonstrated that exposing users to slightly more partisan and antidemocratic content for just seven days created as much political division as naturally accumulated over three years in past decades.
The study, published in the journal Science, involved researchers from leading universities including Stanford, Johns Hopkins, and Northeastern. They collaborated to measure how algorithmic content curation affects political attitudes. More than 1,000 participants were divided into groups, with some receiving feeds containing more divisive political posts and others seeing fewer such posts, all while the adjustments remained virtually imperceptible.
During the 2024 presidential campaign, X became a battleground for viral political content, including manipulated images and AI-generated propaganda. The platform’s “for you” feed, introduced after the company’s acquisition and rebranding, prioritizes content designed to maximize engagement rather than showing only posts from accounts users follow. This algorithmic approach has raised concerns about its societal impact.
The experimental intervention specifically targeted posts expressing hostility toward political opponents, support for undemocratic actions, opposition to bipartisan cooperation, and distorted presentations of political facts. After one week, participants exposed to higher levels of such content reported significantly increased negative feelings toward the opposing political party, along with elevated levels of sadness and anger.
According to research from polling organizations, the vast majority of Americans believe that political divisions now extend beyond policy disagreements to fundamental disagreements about reality itself. More than half of people in other democratic nations share similar concerns about dangerous levels of political division. The study’s revelation that platforms could reduce this polarization through algorithmic adjustments offers hope, though it remains unclear whether companies will prioritize social harmony over engagement metrics that drive advertising revenue.