A Stanford study of X finds “For You” feed algorithms are more likely to show outrage-inducing posts than content matching your values.
Your “For You” feed is supposed to know you. A new Stanford study suggests it might actually be working against you.
Researchers studied X, formerly known as Twitter. They found something counterintuitive: the smarter the For You feed algorithm gets at reading your behavior, the further it drifts from what you actually care about.
The Core Problem: Engagement Isn’t the Same as Values
Social platforms rank content using a metric called engagement. That means likes, replies, reposts, and time spent watching or reading.
The logic seems reasonable at first. Platforms treat high engagement as a signal that you’re interested in more of the same content. But that assumption turns out to be flawed.
“Engagement is easy to measure, so the platforms use it as the primary metric to assess user interest in the content,” said Ziv Epstein, the study’s first author. He was a postdoc at Stanford when he did the work and is now at MIT’s Schwarzman College of Computing. “They are saying, in effect, engagement equals values, but it’s not that simple.”
Here’s why. People often engage most with content they disagree with — replying in anger, arguing in the comments, or reposting something that upsets them. The For You feed algorithm can’t tell the difference between passionate agreement and pure outrage. It just sees engagement, and serves up more of it.
How the Study Worked
The research team designed a simple, elegant test. First, they randomly selected 715 X users and asked them to complete a values survey.
They scored each response using Schwartz’s Hierarchy of Basic Human Values — a well-established academic framework covering 19 distinct values, from humility to hedonism.
Next, the researchers compared two different feeds from each person’s account. The “Following” feed showed content only from accounts the user chose to follow. The “For You” feed showed whatever the algorithm decided to surface.
They scored the content in both feeds against each user’s personal value profile. The result was clear: posts that clashed with a user’s stated values were more likely to appear in their For You feed than posts that actually matched those values.
What This Means for Platforms -and for Us
Michael Bernstein, the study’s senior author and a Stanford professor of computer science, sees a clear takeaway. “The message we take away is that the values that we hold — what we really care about — should be just as informative to the platforms as engagement,” he said.
He points out the platforms are misreading a basic human contradiction. “We engage with all sorts of content — sometimes because we agree with it, and sometimes because we’re angry at it,” Bernstein said. “But the platforms misinterpret it all as a kind of desire for more of that kind of content.”
The study, published in the Proceedings of the National Academy of Sciences, was framed by its authors as a starting point for dialogue, not a takedown of any one platform. It highlights a tension that lives inside users too: we say we want content aligned with our values, but we often react most strongly to content that provokes us.
Can Anything Fix This?
The research team already has one experiment underway. They built a Google Chrome browser extension that re-ranks feeds based on each user’s self-stated values, rather than raw engagement.
Epstein is upfront about its limits, though. A values-only feed could just become an echo chamber — showing you only what you already agree with, which brings its own problems.
Other possible fixes include pro-social algorithms designed to surface content that bridges divides instead of deepening them, or simply giving users more direct control over how their feeds get built.
Bernstein frames the goal as balance, not perfection. “Our feeds shouldn’t value confrontational engagement over complementary content, nor should they obsequiously feed us only content we agree with,” he said. “They should strike a balance and mend social disagreements rather than make them worse.”
His closing point is optimistic. “This work shows we can measure the space between what people value and what the algorithms are amplifying,” Bernstein said, “and in doing that, we can use science to narrow the gaps that divide us.”