A new study shows that even small changes to what users see on X can quickly boost political polarization. In just a week, it can reach levels that used to take years. This highlights the power—and risk—of algorithms that shape public conversations.
The experiment involved over 1,000 American users during the 2024 presidential elections, a time full of misinformation and political memes. Researchers slightly tweaked the algorithm to show more posts with anti-democratic views or aggression towards the other party. Most participants didn’t notice any changes in their feed, but their hostility towards political opponents grew a lot.
The changes seen in the ’emotion thermometer’—which measures warm or cold feelings towards the other party—were similar to the increase in polarization in the U.S. over about 30 years. The findings, published in Science, show that X’s algorithm greatly affects how users view political reality.
At the same time, the study found that reducing exposure to divisive posts had the opposite effect: users reported much milder and less aggressive feelings towards the opposing party. Even though users spent a little less time on the platform, their interactions with posts (likes and reposts) went up.
Researchers believe that the platform can both increase and limit polarization if it wants to. However, such actions might clash with business models that depend on high-emotion content, which keeps users engaged.
