Why Doom Scrolling Feels Impossible to Stop

The session starts as a quick check. You pick up your phone with something specific in mind — a message to send, a notification to clear — and then you're in the feed. Twenty minutes later you're still there, not quite sure when it became twenty minutes. This isn't unusual and it isn't specific to any particular app. It's a structural property of how feeds are designed.

If you want a number for what those sessions add up to across a year, the Doom Scroll Cost Calculator will show you in about 30 seconds. What this article covers is the mechanism underneath it: why the sessions run longer than you intended them to in the first place.

Feeds that don't end

Most content formats have natural stopping points. A printed article ends at the bottom of the page. A TV episode has credits. A podcast finishes. Even a standard webpage has a bottom. The format itself tells you when you're done.

Feeds don't work that way. There's always more content below whatever you're currently looking at. Scroll to the end of what's loaded and the app pulls in more. The feed has no floor.

Tristan Harris, testifying before the U.S. Senate Commerce Committee in 2019, described interfaces like this as "bottomless bowl" designs — formats that remove the stopping cues most content experiences provide.^[1] When there's no natural endpoint, stopping requires the user to make an active decision against a stream that doesn't stop on its own. That's a structurally harder exit than simply reaching the end of something.

Why unpredictable content keeps people scrolling

Most posts in a feed aren't particularly interesting. But some are — and you can't predict which ones they'll be.

That unpredictability matters. When outcomes are predictable, stepping away is easier — you have a sense of what's coming and you've calibrated whether it's worth your time. When they're not, there's always a reason to check one more. The University of Michigan Institute for Healthcare Policy and Innovation has noted that social media platforms apply reinforcement patterns similar to those found in gambling behavior, where unpredictable rewards sustain engagement more reliably than consistent ones.^[2]

One scroll produces something worth reading. The next five don't. The eleventh does. The ratio doesn't need to be high — the irregularity itself is what keeps the session going.

Pull-to-refresh and what it's modeled on

The pull-to-refresh gesture — dragging down from the top of a feed to check for new posts — is a specific design choice. It produces a variable outcome each time: sometimes new content appears, sometimes nothing does.

In his Senate testimony, Harris compared this gesture to pulling the lever on a slot machine.^[1] The physical action is similar, and so is the uncertainty of what it returns. The comparison is structural, not rhetorical — both involve a repeated physical action that yields an unpredictable outcome and prompts the user to try again.

Whether or not you use pull-to-refresh directly, the same pattern — uncertain outcomes from a repeated action — runs through most feed behavior regardless of how you navigate it.

Why sessions feel shorter than they are

Time perception depends partly on the structure of what you're doing. During a sustained activity with a clear arc — a film, a long conversation, a task with a defined endpoint — elapsed time is relatively easy to track. The activity has internal structure that maps onto time.

A feed works differently. Content arrives as short, rapidly cycling, discrete pieces: one post, then another, then a video, then a comment thread. None of it connects to what came before in a way that builds a sense of progression. There's no arc — just more content.

When a session is made up of many short units, each feeling brief, the total is hard to estimate from inside it. The elapsed time doesn't register the way it does during a coherent activity. Most people underestimate how long they've been scrolling — not because they're inattentive, but because the structure of the session makes the total genuinely difficult to track.

Continuous-feed design and the absent exit

In most content formats, stopping is easy because the content itself signals where to stop. The episode ends. The article concludes. The page has a bottom. You reach it and that's the natural moment to decide what to do next.

Feeds don't produce that moment. The transition from one piece of content to the next is seamless — a photo becomes a video, a video becomes a suggested account, a suggested account's content leads to more suggested content. Each transition continues the session. None of them function as a pause point.

Stopping requires actively interrupting something that has no built-in stopping point. That's not impossible — but it's structurally harder than simply reaching the end of something. The feed doesn't offer a natural exit. You have to decide to stop within a stream that does not naturally end.

How 15 minutes becomes an hour without a decision

No single moment of a scrolling session feels long. That's part of why the annual total is usually uncomfortable to look at directly.

Four short sessions across a typical day — a quick morning check, a few minutes at lunch, some time in the evening, a look before bed — can add up to 90 minutes without any individual session feeling like much. Over a week that's 10.5 hours. Over a year, it's 546 hours. In full calendar days, that's 22.8 of them. In eight-hour workdays, it's 68.

The math isn't complicated. What makes it invisible is that no individual session registers as the thing producing the annual total. The cost accumulates across moments that felt too small to count — the same dynamic that makes small impulse purchases hard to track until you add them up. If you want to see your own number, the Doom Scroll Cost Calculator shows the full picture in about 30 seconds.

When overuse becomes the baseline

The first time a session runs unexpectedly long, it might register. The fortieth time, it doesn't. The behavior has become the reference point — the amount of time you scroll is just what you do, not something that needs explaining.

That normalization process runs through a lot of recurring costs. Subscriptions that process on autopay stop feeling like active spending after a few billing cycles — the charge appears, the balance adjusts, and no decision is required. The pattern behind normalized subscription costs is the same one at work here: a cost that continues without re-decision eventually stops being visible as a cost at all. None of this is meant to create guilt. It's just context. The normalization doesn't change the annual total.