Scientists Prove Social Media Can Engineer Political Hostility in Days

Researchers have demonstrated that social media platforms possess unprecedented power to manufacture political animosity among their users. An innovative experiment showed that imperceptible changes to the content fed to users of X can generate levels of political polarization in one week that previously required three years to develop organically.
The academic team developed a novel methodology using artificial intelligence to analyze and manipulate posts in real-time. They created experimental conditions where some users saw marginally more content featuring antidemocratic attitudes and partisan aggression, while others saw less. The vast majority of the more than 1,000 participants remained unaware that their feeds had been modified, yet they exhibited measurable shifts in their political attitudes.
Martin Saveski, an assistant professor who led the research, emphasized the profound implications of these findings. The algorithm’s power to influence users’ feelings toward political opponents is particularly concerning given that the changes to their feeds were barely noticeable. Co-author Tiziano Piccardi noted that the documented shift in political attitudes corresponds to approximately three years of polarization based on historical trends in American society.
The measurement approach involved asking participants to rate their feelings toward political opponents on a scale from warm and favorable to cold and unfavorable. Those exposed to more divisive content showed increased hostility of more than two degrees on the 100-point scale—matching the polarization increase that occurred across four decades ending in 2020. Importantly, the reverse was also true: reducing divisive content decreased polarization by a comparable amount.
This research arrives at a critical moment when democratic societies face unprecedented political division. The study confirms long-standing suspicions that social media platforms amplify divisive content to boost engagement and advertising revenue. However, the findings also reveal a path forward: platforms could redesign their algorithms to promote political harmony. While this might modestly reduce some engagement metrics, users in the less-divisive condition actually showed higher rates of meaningful interaction through likes and reposts.