In 2016, a team of UCLA researchers built a working replica of Instagram inside an fMRI scanner. Teenagers lay in the machine and looked at photographs, some of them their own, each stamped with a like count the researchers had secretly assigned rather than left to chance. The teens were more likely to like a photo themselves when it already carried a high like count, whether the photo showed something ordinary or something risky, like drinking or smoking. When the researchers looked at the brain scans, viewing photos with more likes lit up regions tied to reward processing, a pattern described in a paper published in the journal Psychological Science.
This is one study, run on adolescents specifically, using a simulated version of one platform, so it does not prove that every person of every age responds to a validation signal the same way an app is designed to trigger. What it does show, in a controlled and directly measurable way, is that a number attached to a piece of content, not the content itself, was enough on its own to shift both what teenagers chose to like and how their brains responded while doing it.
Why the number does more work than the content
The mechanism the UCLA team was tapping into is not unique to Instagram or to teenagers. What makes a like count, a notification badge, or a follower number so effective at holding attention is not that any single one of them is especially rewarding. It is that they arrive unpredictably, and unpredictable rewards are, according to decades of behavioral research going back well before social media existed, considerably more effective at sustaining repeated behavior than a reward delivered every single time. A person checking a phone does not know in advance whether this check will produce something validating, and that uncertainty is precisely what keeps the checking going.
This gives the people who design these systems a specific kind of power that has little to do with the actual quality of what is being shared. A platform does not need every post to perform well. It needs enough unpredictability in which posts get noticed, liked, or amplified that users keep returning to find out. The attention a platform captures, in this sense, is less a reward for good content and more a byproduct of a variable-reward structure that would work almost as well on a much weaker piece of content.
What decades of tracking actual attention shows
Gloria Mark, a professor of informatics at UC Irvine, has spent roughly two decades directly measuring how long people stay on a single screen before switching to another one, tracking, rather than surveying, real computer use among working adults. Her original study on the subject, conducted with colleagues and described in her later public research summaries, found that in 2004, people spent an average of about two and a half minutes on a given screen before moving to another task or window. That figure has fallen sharply since, down to roughly a minute or less in her more recent tracking.
Mark’s research measures something narrower than “attention span” as a broad, general trait, and it is worth being precise about that distinction: it tracks time spent on a screen before switching, in mostly workplace computer-use settings, not a validated psychological measure of someone’s overall capacity to concentrate. It is a real and consistently observed behavioral pattern, not a diagnosis of a shrinking human faculty. But the trend itself, a shortening window before attention moves elsewhere, lines up with a media environment increasingly built around exactly the kind of unpredictable, quick-hit validation signal the UCLA study measured directly in the brain.
Where the power sits
Put together, these two pieces of research point toward a specific and less flattering account of where the power actually sits in an attention-based platform. It is not primarily in persuading anyone that a particular post is good. It is in controlling the timing and unpredictability of a validation signal that the brain responds to somewhat automatically, regardless of the underlying content’s merit. That is a structural form of influence, built into how a feed is designed and how notifications are timed, rather than a simple byproduct of interesting content winning attention on its own.
What this means for anyone building an attention-based product
For a founder or product team building anything that competes for attention, whether a social app, a productivity tool with engagement metrics, or a feature that surfaces notifications, the research points to an uncomfortable design choice hiding inside a familiar decision: whether a feedback signal, a badge, a count, a streak, is delivered predictably or unpredictably. A predictable signal, shown every time and in the same way, produces a weaker and more easily ignored response than one whose timing or size is variable. That difference is not a minor implementation detail. It is close to the entire mechanism the UCLA study measured in the brain, and it means two products with near-identical features can produce very different levels of compulsive return depending on how that one variable is tuned.
That puts a real ethical weight on a decision that often gets made for purely engagement-metric reasons. A team optimizing for daily active users by making a reward signal less predictable is, whether or not this language is used internally, building toward the same mechanism that produces compulsive checking in the research literature. None of this requires a company to abandon feedback signals altogether. It does mean that “increase engagement” is not a neutral goal sitting apart from questions about how that engagement is actually produced.
What this does not prove
None of this means every person is equally susceptible, and the UCLA study specifically examined adolescents, a group whose sensitivity to peer feedback is well documented and plausibly higher than in most adults. It also does not mean platform design is the sole cause of shorter attention windows. Gloria Mark’s own research points to multiple contributing factors, including the sheer number of communication tools most people now juggle at once, not social media validation loops alone. What the research does support is a narrower, still meaningful claim: a simple, unpredictable number attached to content can measurably influence both behavior and brain activity, independent of whether the content underneath it deserves the attention it gets.
