Archive · 19 Aug 2026
X Algorithm May Amplify Outrage, New Study Finds
Researchers studied how X decides which posts to show people.
They found that the platform strongly values user engagement.
Replies were uncommon, making up 6.8% of interactions, but they influenced the feed a lot.
Angry replies could signal that a user was interested in similar content.
The system might then show that person more provocative posts.
This sometimes meant users saw content that did not match their stated values.
The pattern appeared among people from different political groups, but it was reported more often among Democratic participants.
Researchers do not yet know why this difference exists.
They warned that such algorithms could create isolated content bubbles, but said more research is needed.
A study found X’s feed algorithm prioritizes engagement when selecting content for users.
Researchers observed that replies, though only 6.8% of interactions, had disproportionate algorithmic influence.
Users who replied angrily to provocative posts were often shown more content in the same direction.
The study involved 715 active American X users during September and October 2024.
Democratic participants encountered value-misaligned content through replies more frequently, though researchers said the cause remains uncertain.
- Who
- Researchers, including Stanford postdoctoral researcher Ziv Epstein, studied 715 active American X users.
- What
- The study examined how X’s feed algorithm responds to engagement and whether recommended content aligns with users’ stated values.
- When
- Participants were studied during September and October 2024; the findings were published recently in the Proceedings of the National Academy of Sciences.
- Where
- The research involved users in the United States on X, including their For You and Following feeds.
- Why
- Researchers sought to understand how engagement-maximizing recommendations shape users’ information consumption and potentially produce value-misaligned content.
- Study title
- “Value misalignment of X's feed algorithm”
- Participants
- 715 active American X users
- Study period
- September and October 2024
- Interaction finding
- Replies accounted for 6.8% of interactions overall.
- Algorithmic pattern
- Replies had disproportionate influence over subsequent recommendations.
- Political pattern
- Democratic participants encountered misaligned content through replies more frequently.
- Research method
- Participants installed a browser extension that collected their For You and Following feeds.