A new study published in the Proceedings of the National Academy of Sciences finds that social media feed algorithms can prioritize posts that conflict with users’ stated beliefs. The research links this pattern to how algorithms weigh user engagement, particularly when accounts reply to certain content.
Across outlets, the explanation centers on the idea that users comment more often on posts they disagree with than on posts they agree with. Because algorithms incorporate those interactions into ranking, they may repeatedly surface content that triggers replies. Phys.org reports this generally, focusing on the algorithm’s reliance on reply behavior. Macau News adds a specific example from X, saying the platform’s recommendations promote content contradicting Democratic and Republican users’ values, and that the effect occurs more strongly for Democrats than Republicans.
