AI Tool Cuts Political Polarization on Social Media

What if you could scroll through social media and feel less angry, more understanding, and actually warmer toward people who disagree with you? Scientists just proved it's possible.

Researchers from the University of Washington and Northeastern University created a browser extension that tackles one of social media's biggest problems: algorithms that push extreme content and fuel political division. The tool uses artificial intelligence to analyze and reorder posts in real time, moving polarizing content down in users' feeds.

They tested it on more than 1,200 volunteers using X (formerly Twitter) during the 10 days before the 2024 U.S. election. Half saw feeds where polarizing posts were buried deeper in their timelines. The other half saw slightly more polarizing content than usual.

The results surprised even the researchers. People who saw fewer polarizing posts felt warmer toward opposing political groups, while those who saw more felt colder. The shift measured two to three degrees on a 100-point scale, which sounds small but equals three years of typical political attitude change in America.

The tool also reduced feelings of sadness and anger while scrolling. And it worked equally well for people across the political spectrum, proving that reducing exposure to extreme content helps everyone feel better.

The Bright Side

For years, critics have blamed social media algorithms for tearing society apart, but only the platforms themselves could access and modify those algorithms. This breakthrough changes everything.

"Only the platforms have had the power to shape and understand these algorithms," says study co-author Martin Saveski. "This tool gives that power to independent researchers." The team made their code freely available so other scientists can build on their work.

The extension bypasses the need for platform approval entirely, working by reordering what users already see rather than removing posts. That means researchers can now study and improve social media without waiting for cooperation from companies that rarely grant it.

University of Toronto psychologist Victoria Oldemburgo de Mello called the approach clever and unprecedented. The researchers combined rigorous scientific control with real-world conditions in a way no one had managed before.

The team plans to expand beyond browser-based sites to apps, though that presents technical challenges. They're also exploring how the same framework could address mental health, well-being, and other concerns tied to social media use.

The tool proves what many suspected: social media algorithms really do create political division. But more importantly, it shows we can fix the problem without sacrificing free speech or waiting for big tech companies to act.