X algorithm prioritizes content that provokes anger, study finds

A study published in the Proceedings of the National Academy of Sciences (PNAS) reveals that the X (formerly Twitter) recommendation algorithm tends to prioritize content that clashes with users' personal values. By analyzing the feeds of 715 U.S.-based users, researchers from institutions including Stanford University found that the algorithm is optimized for engagement, which often translates to amplifying content that provokes anger or disagreement.

The mechanism relies on the fact that users are more likely to reply to posts they oppose than those they agree with. Because the algorithm treats replies as a high-value engagement signal, it creates a feedback loop where users are served more “ragebait” to drive further interaction. The study noted that while this effect impacts both political sides, it was more pronounced among Democratic users.

Researchers observed that the algorithm was more likely to promote content regarding tradition and social rules while demoting content related to caring for others or protecting nature. While X has not officially responded to the study, former product lead Nikita Bier indicated that the platform has since adjusted its recommendation algorithms to reduce the volume of such content.