Blog 5 min read
Why you cannot stop scrolling
The feed is not winning because you are weak. It is winning because four specific design decisions removed every place a session could naturally end.
The question is usually asked as a confession. It is better read as a design question, because the answer is mostly in the product rather than in the person asking.
Four decisions, none of them yours
Infinite scroll. Before it existed, a page ended and you clicked for more — a small moment where continuing was an action you took. Aza Raskin built the pattern in 2006 and has spent much of the time since publicly regretting it, for exactly that reason: it deletes the pause in which a person might ask whether they still want to be here. Every other medium hands you stopping cues for free. The last page of a chapter. The credits. The bottom of the article. A feed has no bottom, so the only available stopping cue is one you have to generate yourself, against the grain, while the thing is still working.
Variable-ratio reward. Most of what passes is unremarkable. Occasionally something is genuinely funny, genuinely useful, or genuinely enraging. Reinforcement delivered on an unpredictable schedule produces the most persistent behaviour of any pattern in the behavioural literature — more persistent than reliable reward, which is the counter-intuitive part. If every tenth post were good, you would learn to check every tenth post. Because you cannot know, you keep checking. This is the slot machine result, and it was not adapted into consumer software by accident.
Ranking that learns you specifically. The feed is not showing you what is popular. It is showing you what has held people like you, adjusted continuously by what has held you. Every session is training data. The system's estimate of what will stop your thumb improves every time you use it, which means the honest description of "I have less self-control than I used to" is often "the thing has more information about me than it used to".
Negativity bias, on the other side of that ranking. Attention goes to threat before it goes to anything else, because that is what kept the species alive. A ranking system optimising for attention discovers this within a few sessions and does not need to understand it. The result is a feed slanted toward alarm — not from malice, but because alarm performs.
Stack the four: no natural ending, unpredictable reward, a model of your specific weaknesses, and content selected for the emotion hardest to look away from. The remarkable thing is not that people scroll for three hours. It is that anyone stops.
Is it "addiction"?
Worth answering directly, because the word does real work in both directions.
Formally, no. The DSM-5 does not recognise social media or smartphone addiction as a disorder. The ICD-11 added gaming disorder but not scrolling. What researchers actually study is problematic use or compulsive use, measured with scales — like the Doomscrolling Scale published by Satici and colleagues in 2022 — rather than diagnosed against criteria.
That distinction matters for two reasons and neither is the one people expect.
It matters because "addiction" implies a treatment model that mostly does not apply. Abstinence is not available: you cannot stop using a phone the way you can stop drinking. The realistic target is a changed relationship with a tool you will keep using daily, which is a different project with different methods.
And it matters because "not a real addiction" gets used to imply "so just stop", which does not follow either. The absence of a diagnostic category says something about the state of psychiatric nosology, not about whether the behaviour is hard to change. The measured associations with psychological distress, sleep loss and lower life satisfaction do not require a diagnosis to be real.
The useful framing sits between the two: a strongly reinforced habit, running on hardware you carry everywhere, maintained by systems specifically optimised to maintain it. Not a moral failure, not a disease. An engineering problem, mostly on the other side of the screen.
Why knowing this does not fix it
Here is the deflating part. Everything above is now widely known, and knowing it changes behaviour barely at all.
You can understand variable-ratio reinforcement in detail and still lose an evening. Insight is not the mechanism the habit runs on. The habit runs on cue and reflex, and cue and reflex do not read.
This is why awareness-based tools plateau so predictably. A weekly report telling you it was 4h 47m is a fact, and facts are interesting the first two times. A notification pointing out that you have opened Instagram eleven times today is information delivered to a person who is not, at that instant, in an information-processing mode.
What changes behaviour is changing the situation, not the understanding of it. That is the consistent finding across the self-control literature, and it is why the useful question is not "why can't I stop" but "what would have to be different for stopping to be the default".
What would have to be different
Three things, in order of how much they buy you.
An ending that comes from outside. The core defect is the missing stopping cue. Supply one that does not depend on you generating it mid-session: a hard limit with authentication behind it, a room the phone is not in, a person expecting you.
A cost that does not fall as your resistance falls. Friction made of seconds is cheap at 1am, which is precisely when it is tested. This is the asymmetry that defeats most tools in this category, and the reason we build around a price instead — the argument in full is under commitment devices.
A decision point where the design removed one. Infinite scroll deleted the moment where continuing was an act. Anything that reinstates it does real work, even crudely. The point of putting a price on going past your limit is not the money. It is that paying is unavoidably an act — a thing you do on purpose, at a number you can see, at the one moment a reflex was about to pass for a choice.
That is what Last Ferry is: one shared daily budget across the apps that eat your day, a real lock when it runs out, and a printed price to go past it that rises with each unlock — $0.50, then $1.00, then $2.50, resetting at midnight and never resetting because you paid.
Most days you close the app and keep the money. That is the app working.
Common questions
Is scrolling addiction a real diagnosis?
Not formally. The DSM-5 does not include social media or smartphone addiction as a disorder, and the ICD-11 recognises gaming disorder but not scrolling. Researchers generally study it as problematic or compulsive use, measured on scales rather than diagnosed.
Does that mean it is not a real problem?
No. It means the clinical category does not exist, not that the behaviour does not. The measurable associations with distress, sleep loss and reduced wellbeing hold regardless of what it is called.
Why is it harder to stop scrolling than to stop watching a series?
A series has episode boundaries and an ending. A feed has neither. Every stopping cue that other media provide for free has been deliberately removed, so the decision to stop must come entirely from you.
Who invented infinite scroll?
Aza Raskin, in 2006. He has spoken publicly and repeatedly since about regretting it, on the grounds that it removes the natural pause where a person would otherwise ask whether they want to continue.