Arguing with people on X may be teaching the platform to show you more posts you hate, according to a new study that found the platform's recommendation algorithm amplifies content misaligned with users' stated values.

Research published Monday in the Proceedings of the National Academy of Sciences found that X's "For You" algorithm amplified content that clashed with users' stated values, particularly among Democrats. The correlation between amplified content and stated values was -0.229 for Democrats, compared with -0.041 for Republicans, with more negative numbers indicating greater misalignment.

"Social media feed algorithms rank content that is purported to be preferred by users, but the engagement behaviors that drive these algorithms are (at best) indirect proxies for users' explicitly self-stated values," the researchers wrote. Ranking by predicted engagement is the standard architecture across major platforms — TikTok, Instagram, and YouTube similarly build their feeds around models that forecast which content users will interact with — so the gap the researchers identify between engagement signals and stated values is not unique to X's implementation.

According to the report, researchers from MIT, the University of Michigan, Johns Hopkins, Stanford, and Yale studied 715 active U.S. X users in September and October 2024. Participants installed a browser extension that collected their "For You" and Following feeds.

X reportedly showed Democrats more content that ran counter to their stated values, while the effect was much weaker for Republicans. Across all users, amplified content scored -0.150, while posts from accounts they chose to follow generally aligned with their values. The contrast matters in practice: "For You" is the default timeline on X, and users must manually switch to the chronological Following feed, so amplified content is what the platform serves first.

"However, beyond this alignment between self-stated values and inventory, we observe participants' values to be negatively correlated with the likelihood of content amplification," the researchers said, adding that the finding suggests "the algorithm is actually amplifying content that is misaligned with users' explicit values."

A replication produced a similar partisan split, at -0.199 for Democrats and 0.0372 for Republicans.

Replies may account for some of the difference, according to the study. Researchers said users who reply to people posting content that clashes with their values could inadvertently signal to X that they want more of it.

"Taken together, these findings offer a glimpse into the tensions between the values that people hold and those that provoke reactions, and how these value tensions can produce misaligned outcomes," the researchers wrote.

The researchers did not examine whether the accounts behind posts that prompted value-misaligned replies were human or automated, leaving the role of bots unanswered.

How feed algorithms shape what users see has also drawn regulatory scrutiny in Europe. The EU's Digital Services Act requires very large platforms including X to offer at least one recommender option that is not based on profiling, and the European Commission opened formal proceedings against X in December 2023 over suspected violations including its handling of illegal content and its "blue check" verification labels.

The findings come after X open-sourced its Grok-powered "For You" algorithm in January, revealing a recommendation system that ranks posts by predicting user actions, including likes and replies. It was the platform's second such release: X, then still named Twitter, first published parts of its ranking code in March 2023, months after Musk completed his roughly $44 billion acquisition of the company. Because the latest code is public, outside researchers can now inspect how signals like replies are weighted — the same mechanism the PNAS study links to value misalignment. Elon Musk said at the time that the algorithm "needs massive improvements."

"We know the algorithm is dumb and needs massive improvements, but at least you can see us struggle to make it better in real-time and with transparency," Musk tweeted following the release. "No other social media companies do this."