An algorithm driving the social media platform X actively amplifies content that directly conflicts with users’ core moral and political values, according to a study published in the Proceedings of the National Academy of Sciences.<\/p>\n
Researchers discovered that the platform’s feed recommendation systems generate an outrage feedback loop by disproportionately prioritizing high-friction interactions such as replies, despite those responses accounting for a small fraction of overall user activity.<\/p>\n<\/p>\n
The study tracked 715 American X users who agreed to install a browser extension monitoring their “For You” and “Following” feeds. Participants completed a political alignment disclosure and a values inventory using the Schwartz Theory of Basic Values, an analytical model evaluating belief systems across 19 distinct traits, including tolerance and dominance.<\/p>\n
By comparing served feed content against each user’s declared values, researchers observed that X’s automated distribution systems were consistently more likely to push material that provoked opposition.<\/p>\n
“X’s feed algorithm, like a lot of these social media algorithms, is optimized for engagement, but it turns out that not all types of engagement are considered equally,” Stanford researcher and study co-author Ziv Epstein told 404 Media.<\/p>\n
A core finding centers on how different interaction types shape distribution. While direct replies make up under seven percent of total interactions on X, the system weights them much more heavily than passive signals like likes.<\/p>\n
“Replying is only a fraction of engagement, but there does seem to be some evidence that these algorithms are prioritizing and learning more from this kind of rarer form of engagement,” Epstein said. “So it’s this feedback loop of outrage baiting. The algorithm learns that you get outraged and then continues to serve more content in that direction.”<\/p>\n
The data also revealed a distinct political disparity, with outrage-inducing content delivered at higher rates to users identifying as Democrats. Researchers noted that this skew could stem from a broader prevalence of right-wing content across X, or from a higher propensity among Democratic users to engage directly with opposing viewpoints.<\/p>\n
Addressing the broader impact of closed recommendation systems, Epstein emphasized that these automated curation mechanisms exercise substantial influence over public discourse with minimal accountability.<\/p>\n
“There are these social media algorithms that have enormous amounts of power in our lives, they shape the information that we consume and we have very little transparency into how they operate and what their implications are,” Epstein said, noting that public insight is necessary for civil society and countering the platform control over algorithmic distribution.<\/p>\n
X has not issued a formal corporate response to the research. Former X head of product Nikita Bier<\/a> stated that the platform’s underlying algorithm has been modified since the data collection took place, asserting that ragebait content was subsequently reduced by an order of magnitude.<\/p>\n