Study: X's 'For You' feed shifted users toward conservatism

X's 'For You' feed pushed users toward conservatism, study finds

Using X's algorithmic "For You" feed shifted users' political views toward conservatism, according to a new study published in Nature. The research, a 2023 field experiment with 4,965 active US users, found the effects persisted even after participants returned to a chronological feed.

Participants were randomly assigned to use either the "For You" feed or the chronological "Following" feed for about seven weeks. Researchers monitored their feed content, online behavior, and political opinions through surveys and a custom browser extension.

Algorithmic feed increased engagement and conservative follows

Users on the "For You" algorithmic feed engaged more with the platform overall. They were also more likely to follow conservative political activist accounts compared to those on the chronological feed.

The study found these users adopted more conservative policy priorities. Specific shifts in opinion were noted regarding:

• Criminal investigations into Donald Trump
• The war in Ukraine
• General policy attitudes

No similar changes in views or follow behavior were observed for users assigned to the chronological "Following" feed.

Algorithm boosted conservative and activist content

Analysis of the feed content revealed the X algorithm's curation played a direct role. The "For You" feed systematically promoted certain types of content while demoting others.

The algorithm showed users more posts from conservative and activist accounts. It simultaneously demoted content from traditional news outlets.

This curation created a distinct online political environment that the researchers say meaningfully shaped attitudes. The effect was strong enough to endure after the seven-week experiment ended.

Contradicts earlier findings on algorithmic influence

The findings challenge previous large-scale studies, including collaborations with Meta, which found little evidence that feed algorithms changed political attitudes. A key difference in this experiment was the ability to measure the impact of initial algorithmic exposure.

Earlier research could not determine if users' views were already shaped by years of algorithmic curation before the studies began. This experiment started with a baseline measurement and observed changes from it.

The authors conclude that social media algorithms do more than just control what users see. They actively influence the political environment users inhabit and the opinions they form, with effects that can last.