X's Algorithm Feeds Off Ragebait and Impacts Democrats More, Study Finds

A new study published in the Proceedings of the National Academy of Sciences (PNAS) reveals that X's algorithm tends to serve more "ragebait" content to people who identify as Democrats. According to the paper titled "Value misalignment of X's feed algorithm is a reflection of value tensions in engagement," the site's algorithm prioritizes engagement above all else when generating a user's "For You Page."

While the exact reason for this disparity is unclear, researchers believe it may be due to Democrats engaging with conflicting content more often or because there is more right-wing content on X overall. The study, conducted by Ziv Epstein, a postdoctoral researcher at Stanford University, found that replies accounted for just 6.8% of interactions but carried disproportionate weight.

Epstein explained that the algorithm learns that outrage leads to more engagement and continues to serve similar content. This feedback loop creates a cycle of outrage baiting, where the algorithm learns and adapts based on user reactions. X's former product chief, Nikita Bier, confirmed that the site was set to favor replies, but this has since been changed.

Bier noted that the largest contributor to seeing ragebait was the reply predictor, which was given a 15x boost for friend posts and reduced ragebait by an order of magnitude.