X Algorithm Amplifies Content That Contradicts Users’ Values, Study Finds

A new study published in the Proceedings of the National Academy of Sciences examined 715 U. S. users on the X platform and found that the feed algorithm is more likely to surface posts that clash with a user’s stated values than those that align. The researchers traced this effect to the algorithm’s heavy weighting of replies—a type of interaction that users give more often to content they disagree with—over the more common “likes.” The bias is stronger for Democratic users, who tend to reply more to objectionable posts, creating a feedback loop that amplifies misaligned content. The authors suggest that allowing users to specify their values and then sorting feeds accordingly could reduce echo chambers and potentially lower polarization.