< Back to all clusters
[TECHNOLOGY] · United States · 2 sources

started · updated

X algorithm amplifies content contradicting user values, study finds

A study published in the Proceedings of the National Academy of Sciences (PNAS) reveals that the recommendation algorithm for the “For You” feed on X, formerly Twitter, tends to amplify content that contradicts the expressed values of its users.

Led by researchers from the Massachusetts Institute of Technology (MIT) and Stanford University, the study analyzed the feeds of 715 American users. By comparing algorithmic recommendations with the accounts users actually chose to follow, researchers found a negative correlation of -0.150 between user values and the content amplified by the feed.

The discrepancy is attributed to the platform's reliance on engagement as a primary metric. Because users often interact—through comments, shares, or replies—with content they disagree with, the algorithm interprets this high engagement as interest. This creates a cycle where the system prioritizes provocative or indignating content to maintain user attention, even when it conflicts with the users' genuine preferences and values.

Entities

Massachusetts Institute of Technology · Michael S. Bernstein · Stanford University · Ziv Epstein