In 2021, leaked internal documents revealed that Facebook's own researchers had flagged a problem they couldn't quite name. More than 350 million users — roughly one in eight — reported that their use of the platform had crossed a line. It was affecting their sleep, their work, their parenting, their relationships. The researchers had proposed studying whether specific features — video autoplay, like-count notifications, the endless feed — were creating what they carefully called "frequent, automatic, undesired behaviors." Habits users didn't choose and couldn't stop.
Facebook's response was not to run the study. The team was reshuffled twice, given half the resources it requested, and eventually disbanded entirely. The question they wanted to answer — are we building something addictive? — remained strategically unanswered.
Here's what makes that decision revealing. Facebook didn't need the study to know the answer. The behavioral data was already in their servers. They could see the sessions that started at 10 PM and ended at 1 AM. They could see the engagement metrics climbing while satisfaction surveys flatlined. They could see the same pattern that neuroscience had already documented in a different context — the pattern of a system that makes you reach for something without making you enjoy what you find.
The Slot Machine in Your Pocket
The reason you can't stop scrolling at 10 PM is the same reason a gambler can't leave a slot machine at 2 AM. The mechanism is identical. It has a name: variable-ratio reinforcement.
In a variable-ratio schedule, rewards arrive unpredictably. Sometimes three scrolls produce nothing interesting. Then the fourth delivers something genuinely funny, or outrageous, or validating. Then seven scrolls produce nothing. Then two good ones in a row. The pattern is random, and randomness is precisely what the dopamine system is designed to lock onto.
Research on reward variability has demonstrated that random-ratio reward schedules maximize dopamine release in the midbrain and basal ganglia — more than predictable rewards of the same magnitude. The unpredictability isn't a bug in the reward system. It's the feature that drives it hardest. A social media feed is a variable-ratio reinforcement schedule delivered at the speed of a thumb swipe. Each scroll is a pull of the lever. Each post is a potential jackpot. The dopamine system doesn't care that the jackpot is a meme instead of money. It cares that the pattern is unpredictable, and it will keep you pulling.
This is the wanting system at full power. The dopamine circuit that generates the pull toward an activity — the craving, the "one more scroll," the inability to put the phone down — is not the circuit that generates enjoyment. It's the same wanting-liking dissociation that drives founder burnout — chasing the next milestone without enjoying the work. Kent Berridge at the University of Michigan has spent decades mapping this dissociation. Dopamine mediates wanting. It never enhances liking. Even when you stimulate dopamine directly in the brain's hedonic hotspots, pleasure doesn't increase. Drive does.
Your phone at 10 PM is a wanting machine running on a liking deficit.
The Satisfaction Gap
A 2025 study by Lakatos and colleagues applied Berridge's wanting-liking framework directly to Facebook use, surveying 1,436 people. The results confirmed exactly what Facebook's internal researchers suspected but weren't allowed to study. Wanting — the dopamine-driven craving — positively correlated with usage time, frequency, and problematic use patterns. Liking — actual enjoyment — showed only marginal and inconsistent correlation with any of them.
The researchers' conclusion: the dynamics are "strikingly similar to substance use." People don't keep scrolling because they're having a good time. They keep scrolling because the wanting system has been sensitized to crave the next variable reward, and the liking system has nothing to do with it.
You've felt this. You open your phone to check one thing. Forty-five minutes later, you're watching a video you don't care about, posted by someone you don't follow, about a topic you won't remember tomorrow. You don't feel good. You don't feel entertained. You feel the particular emptiness of a craving that keeps renewing without ever being satisfied. That's the wanting-liking gap, and it's widest at 10 PM.
Why 10 PM Is the Worst Time
The wanting-liking gap exists all day. But it widens dramatically after dark, and the reason has nothing to do with willpower.
By 10 PM, the prefrontal cortex — the region responsible for executive control, impulse suppression, and strategic decision-making — has been running sustained computation for fifteen or sixteen hours. Its connectivity with the emotional and motivational systems it's supposed to regulate has degraded. The precision-weighting system that keeps goal-relevant signals prioritized over immediate-reward signals has shifted toward comfort and habit — the same cognitive decline curve that dictates your morning routine. The part of your brain that would normally evaluate the scroll and conclude "this isn't worth my time" is running at reduced capacity.
Meanwhile, the wanting system fatigues far less than the executive system. Dopamine circuits don't need the prefrontal cortex to fire. They run on subcortical loops that operate independently of executive function. At 10 AM, the wanting signal says "check your phone" and the prefrontal cortex evaluates the signal, weighs it against your goals, and often vetoes it. At 10 PM, the same signal fires and the veto system is running on fumes. The impulse passes through unchecked.
Then the blue light makes it worse. Research on evening smartphone use has found that short-wavelength light from screens before bedtime suppresses melatonin secretion and disrupts sleep architecture. But the critical finding wasn't about the light. Night-mode filters and blue-light-blocking settings made little measurable difference to actual sleep outcomes, despite partially protecting melatonin levels. The problem isn't the wavelength. It's the cognitive and emotional arousal from the content itself. The variable-ratio reinforcement keeps cortisol elevated, preventing the natural decline the brain needs to initiate sleep.
You're not just wasting time. You're degrading the hardware that will make tomorrow's decisions.
The Numbers
The scale of this is worth pausing on. Eighty-six percent of Americans use their phone in bed before sleeping. The average bedtime scroll lasts 38 minutes. Americans lose an estimated 231 hours of sleep per year to bedtime scrolling — nearly ten full days. Twenty-eight percent stay up scrolling past 2 AM on work nights.
And the satisfaction data is damning. Thirty-eight percent of adults say bedtime scrolling makes their sleep "slightly or significantly worse" — rising to 46 percent among 18-to-24-year-olds. Research suggests that roughly 45 minutes of daily social media use predicts the lowest negative affect. The average for U.S. adults is now roughly two hours and fifteen minutes — about three times the optimum. Not because people are enjoying those hours. Because the wanting system has no off switch, and by 10 PM, the system that would provide one has clocked out.
Try This: The Scroll Audit
This protocol makes the wanting-liking gap visible — which is the first step to closing it.
-
The satisfaction log. For five days, every time you put your phone down after a scroll session longer than ten minutes, rate two things: time spent (in minutes) and satisfaction (1-10, how much you actually enjoyed the experience). Don't rate the wanting — rate the liking. "Did I have a good time?" The gap between time spent and satisfaction is the wanting-liking gap made measurable. Most people find it's widest for sessions that start after 9 PM.
-
The 10 PM phone curfew. At 10 PM, the phone goes in a drawer in a different room. Not on the nightstand. Not on airplane mode in your pocket. In a drawer, in another room, where retrieving it requires standing up, walking, and opening something. The friction doesn't need to be large — the same environment design principle that explains why your office is making you stupid works here. It needs to be larger than the impulse, and at 10 PM, the impulse is running on a degraded veto system. A closed drawer in another room is usually enough.
-
The replacement prediction error. The wanting system fires hardest when the alternative is boredom — when there's no other source of prediction error available. Replace the scroll with something that generates mild novelty without a screen: a book you haven't read, a conversation, a walk. The wanting system doesn't care what produces the prediction error. It just needs one. Give it something that the liking system can also enjoy.
-
The weekly audit. At the end of each week, look at your satisfaction log. Calculate the average satisfaction rating for pre-9 PM sessions and post-9 PM sessions. If the evening scores are consistently lower — and they will be — you've now quantified what the wanting system is costing you. The number makes the abstract concrete, and concrete evidence is what the prefrontal cortex needs to override a habit.
Facebook's researchers wanted to study whether their product was addictive. They already had the answer in 350 million data points. The wanting signal was high, the liking signal was flat, and the gap between them was widest in the hours when the brain's executive system was least equipped to notice.
Your phone does the same thing to you every night. The variable-ratio reinforcement keeps the dopamine flowing. The degraded prefrontal cortex lets the impulse through. The blue light and cognitive arousal suppress the sleep signals. And in the morning, you wake up tired, wondering where the evening went, telling yourself you'll do better tonight — with the same brain that will run the same computation at 10 PM.
The fix isn't discipline. The discipline system is the one that's offline. The fix is architecture: remove the phone from the environment before the wanting system and the weakened veto system negotiate a deal you didn't agree to. An if-then implementation intention — "If it's 10 PM, then I put the phone in the kitchen drawer" — can automate the decision before the veto system clocks out.
Chapter 6 of Wired covers the full neuroscience of how your phone hijacks the prediction engine — including the inventor who removed the "done" signal from the internet and now regrets it, why your brain generates phantom vibrations for notifications that aren't there, and how curated feeds distort the brain's model of reality in ways you can't detect from the inside. If you've ever put your phone down and immediately picked it back up without knowing why, that chapter explains the machinery.
FAQ
Why can't you stop scrolling your phone at night? Late-night phone scrolling is driven by the collision of two systems: the wanting system (dopamine-driven, activated by variable-ratio reinforcement from social media notifications and feeds) and a degraded prefrontal cortex (the veto system that catches impulsive behavior runs at reduced capacity after a full day of decisions). The wanting signal stays strong while the braking system is offline, creating a gap where compulsive behavior runs unchecked.
How does your phone affect sleep quality? Phones disrupt sleep through three mechanisms: blue light suppresses melatonin production (the hormone that signals sleep readiness), cognitive arousal from content keeps the brain in an alert state, and variable-ratio reinforcement from notifications creates a wanting signal that competes with sleep signals. The combination delays sleep onset, reduces sleep quality, and degrades next-day cognitive function.
What is the wanting-liking gap in phone use? The wanting-liking gap describes using your phone compulsively (high wanting) without enjoying the experience (low liking). Facebook's own internal research found that 350 million users reported their platform use had crossed into unwanted automatic behavior — affecting sleep, work, and relationships. The wanting system, driven by dopamine and variable reinforcement, doesn't track satisfaction. It tracks prediction error and novelty.
How do you break the late-night scrolling habit? The fix is environmental architecture, not discipline — because the discipline system is the one that's offline at night. Remove the phone from the bedroom before the wanting system activates. Use a physical alarm clock instead of a phone alarm. Set the phone to charge in another room by 9 PM. You're not fighting the impulse; you're deleting the cue that triggers it, the same principle that worked for Google's M&M experiment.
Works Cited
- Haggin, P. (2021). "Facebook Knew Its Products Were Habit-Forming." The Wall Street Journal. https://www.wsj.com/articles/facebook-knew-its-products-were-habit-forming-11632931201
- Ward, A. F., Duke, K., Gneezy, A., & Bos, M. W. (2017). "Brain Drain: The Mere Presence of One's Own Smartphone Reduces Available Cognitive Capacity." Journal of the Association for Consumer Research, 2(2), 140–154. https://doi.org/10.1086/691462
- Berridge, K. C., & Robinson, T. E. (1998). "What Is the Role of Dopamine in Reward: Hedonic Impact, Reward Learning, or Incentive Salience?" Brain Research Reviews, 28(3), 309–369.
- Chang, A.-M., Aeschbach, D., Duffy, J. F., & Czeisler, C. A. (2015). "Evening Use of Light-Emitting eReaders Negatively Affects Sleep, Circadian Timing, and Next-Morning Alertness." Proceedings of the National Academy of Sciences, 112(4), 1232–1237. https://doi.org/10.1073/pnas.1418490112