The Like That Lands: What 17 Million Posts Revealed About Depression
For decades, one of the most reliable findings in the study of depression has been that it flattens a person's response to pleasure. Clinicians call the symptom anhedonia — a reduced capacity to feel reward — and in laboratory tests it shows up as blunted reinforcement learning. Give a depressed person a small win and, on average, they update their behavior less than someone who isn't depressed. By that logic, the people least likely to keep chasing a hit of approval online should be exactly the people who are most depressed.
A study published in JAMA Psychiatry on July 8, 2026, found the reverse. A team led by Dan-Mircea Mirea at Princeton, working with colleagues at Trinity College Dublin, examined more than 17 million posts from 7,736 users of X, formerly Twitter. They wanted to know how strongly one day's likes shaped the next day's behavior — and whether depression made that pull weaker, as the lab would predict, or something else entirely.
The measure at the center of the work is deceptively plain. For each person, the researchers calculated a "reinforcement tendency": the degree to which a higher average of likes per post on a given day predicted more posting the day after. It is a behavioral fingerprint, not a survey answer — a record of what people actually did, scraped from years of public activity rather than recalled in a clinician's office. And here the surprise arrived. People with more depressive symptoms were more reinforced by likes, not less.
What makes the result hard to wave away is that it held up three separate times. The datasets differed in almost every respect that ought to matter — one drew on users who had publicly disclosed a depression diagnosis, another on people who self-reported symptom severity, a third on participants who completed a standardized depression scale. Different populations, different ways of defining depression, different sample compositions. The effect appeared in each one, across ages and genders. In research, that kind of stubborn replication is worth more than any single dramatic number — which is just as well, because the numbers are small. The standardized effect sizes land around 0.01 to 0.02, and for a substantial share of individual users the relationship ran the other way, with a day of heavy approval followed by less posting rather than more. What the study establishes is that the effect is real and points consistently in one direction across very different populations, not that likes govern anyone's behavior. It also describes only people who post regularly to begin with.
So why would a mind that shrugs at rewards in a lab lean harder into them on a phone? The likeliest answer is that the lab and the timeline are measuring two different things. A poker chip or a dollar in a controlled experiment is an abstraction; a like from another human being is a small, genuine social signal — the digital equivalent of someone nodding at you across a room. The researchers raise the possibility that social media functions as a kind of substitute economy for people who aren't getting much reward offline, or that it taps into long-noted patterns in which depressed individuals actively seek reassurance from others. Their own framing is careful about what the finding does to the textbook: "Our findings add nuance to the existing literature, which generally links depression to blunted reinforcement learning," the authors wrote.
There's a second twist buried in the specifics, and it's the part that keeps the study from being a tidy headline. When the team used a broader mental-health survey to sort symptoms into clusters that cut across the usual diagnostic lines, the amplified pull of likes didn't attach itself to depression in general. It clustered in one subtype the researchers labeled anxious depression — low mood and apathy braided together with long-standing anxiety. Compulsivity and intrusive thinking produced a different behavioral signature altogether. In other words, the effect isn't a vague fog hanging over anyone who feels low. It's specific, and specificity is what makes a finding useful rather than merely arresting.
None of this proves the direction of the arrow, and the authors are the first to say so. Because the study captures behavior across time but only measures mood at a single point, it can't tell us whether depression makes people more responsive to likes, or whether being trained so heavily by likes gradually deepens depression — or whether both feed each other in a loop. What it does deliver is a rare look at the machinery running underneath everyday online life, and a reminder that the tidy story from the lab doesn't always survive contact with how people actually behave. As Mirea put it, understanding mental health means watching not just what happens under fluorescent lights, but "the behavior people show in their actual lives." The loop, it turns out, runs whether or not the person inside it ever notices it's there — and knowing it exists may be the first quiet step toward loosening its grip.
This piece discusses depression and mental health. If you're struggling, a licensed professional can offer support tailored to your situation.