← All games

Dark Pattern Safari

A field guide to the tricks built to keep you. Every specimen defanged

The field guide

In 2010 a London UX specialist named Harry Brignull started cataloguing interface tricks that work against the user rather than for them, and gave them a name: dark patterns. Intent is what separates the species from plain bad design — bad design gets in your way by accident, a dark pattern by blueprint. The fauna thrives best where the money is not in the purchase but in the return visit: mobile games, casinos, and — less expectedly — the "brain-training" apps, which borrowed nearly the whole casino toolkit and changed only the signage.

H. Brignull, darkpatterns.org, 2010, now deceptive.design; his "Deceptive Patterns", 2023. The largest field survey: A. Mathur et al., "Dark Patterns at Scale", ACM CSCW 2019 — 1,818 sightings across 11,000 shopping websites.

Below are ten exhibits, each shown in an enclosure: mechanism armed, teeth removed. Nearly all of them feed on three weaknesses — fear of loss (a loss weighs about twice an equal gain), variable reward (an unpredictable prize holds tighter than a regular one), and sunk cost (it hurts to abandon what you have put in). Each enclosure carries a sign, "Here:" — what this site does instead.

1. The Streak

Habitat: Duolingo, step counters, "daily rewards" · Mechanism: loss aversion

A counter of consecutive days. It hatches quietly: on day three you learn you have a "streak", by day thirty you are afraid to lose it. The trick is precise because it rests on a measured asymmetry: a loss is felt roughly twice as strongly as an equal gain. A missed day takes nothing real from you — whatever you learned stays learned — but the counter shows zero, and the brain mourns that zero as a loss. The genre's crowning touch is selling insurance against itself: a "streak freeze" for in-app currency. The app first manufactures the fear, then sells the umbrella.

D. Kahneman, A. Tversky, "Prospect Theory: An Analysis of Decision under Risk", Econometrica 47, 1979; their estimate of the loss-aversion coefficient, about two: Journal of Risk and Uncertainty 5, 1992.
Here: There is nothing here to lose. No visit counter, no series, no freezes. Skip a week and the site greets you exactly as it did yesterday.

2. Variable Reward

Habitat: slot machines, loot boxes, pull-to-refresh · Mechanism: variable reinforcement

The discovery was made on pigeons. Ferster and Skinner compared schedules of reinforcement and found the stickiest one: a reward after an unpredictable number of tries. A bird fed every tenth peck works at a steady pace; a bird fed every tenth peck on average, but never knowing which, pecks continuously and is the last to give up when the feed is switched off. A slot machine is that schedule in pure form, a loot box is the same thing gift-wrapped, pull-to-refresh is the same thing free of charge. The field mark of the species: the outcome does not depend on skill, only on the next try.

C. B. Ferster, B. F. Skinner, "Schedules of Reinforcement", 1957; N. D. Schüll, "Addiction by Design: Machine Gambling in Las Vegas", 2012.
Here: Whatever chance exists here lies on the table — the backgammon dice are visible to both players — rather than hidden in a reward schedule. The rung grows from skill, not from the number of tries.

3. Fake Progress

Habitat: brain-training apps, mobile games · Mechanism: sunk cost

An experience bar, levels, a "performance index". They grow from attendance: show up, and it drips. The number climbs reliably and produces a feeling of motion, though what it measures is at best your practice at the exercise itself, and more often mere presence. The job of such progress is not to report but to attach: the taller the column, the harder to walk away. Test the enclosure below: the bar fills from the mere fact of clicking.

EXHIBIT · DEFANGED
Level 1 · 0%
The bar grows from clicks. Skill does not enter the formula. Reach the end and you get a new level and the same bar.
That practice improves the practised task and little around it is established: D. Simons et al., "Do “Brain-Training” Programs Work?", Psychological Science in the Public Interest 17, 2016.
Here: The rung on this site can fall. It does not grow from visits — only from beating a stronger opponent than before — and it drops when you stop managing that. This makes it duller than an experience bar. And honest.

4. Science-Flavoured Claims

Habitat: "trains your brain", "designed by neuroscientists" · Mechanism: deference to authority

"Trains memory, attention and intelligence." In January 2016 the US Federal Trade Commission priced that sentence at two million dollars: Lumosity paid for unfounded claims of staving off age-related cognitive decline — it had been buying ads against search queries about dementia and Alzheimer's, among others. The meta-analyses of Sala and Gobet, pooling hundreds of experiments, reach a stable verdict: training improves the trained task and almost nothing around it. Near transfer is weak; far transfer — to memory in general, attention in general, intelligence — fails to appear. You get better at sudoku. At sudoku.

FTC press release, 5 January 2016: Lumos Labs to pay $2 million. G. Sala, F. Gobet, "Does Far Transfer Exist?", Current Directions in Psychological Science 26, 2017; their "Cognitive Training Does Not Enhance General Cognition", Trends in Cognitive Sciences 23, 2019.
Here: The test pages here say it plainly: your score on this test will improve because you get used to the test. About the brain in general, nothing is promised at all.

5. The Leaderboard

Habitat: anything with a top-100 table · Mechanism: social comparison

A leaderboard turns a pastime into a tournament you never entered. It leans on what Festinger described back in 1954: lacking an objective measure, people measure themselves by their neighbours — preferably the ones slightly above. The top rows of any public table are soon colonised by bots, cheaters and people for whom the table has become a job; everyone else gets a seat in the tail and a conclusion about themselves. Both halves serve retention: the top has something to defend, the bottom somewhere to crawl.

L. Festinger, "A Theory of Social Comparison Processes", Human Relations 7, 1954.
Here: This site's global tally — humanity against the computer — adds everyone into a single number. It has no rows, names or ranks: it can tell you how people are doing, and it cannot tell you that you are worse than someone.

6. Artificial Scarcity

Habitat: shops, booking sites, energy and lives in games · Mechanism: fear of missing out

"Two rooms left at this price", "offer expires in 1:59", energy that runs out and lives that regenerate in forty minutes. Scarcity speeds up the decision by switching off comparison: no time to think — grab it. In games it also works in reverse: your lives ran out not so that you would leave, but so that you would return on a timer or pay to re-enter early. A Princeton field study of shopping sites found countdown timers that honestly reach zero — and start over. One of them grazes below.

EXHIBIT · DEFANGED
The −70% subscription discount expires in 0:06
There is no subscription and no discount. Sit it out to zero — the timer starts again.
A. Mathur et al., "Dark Patterns at Scale: Findings from a Crawl of 11K Shopping Websites", ACM CSCW 2019 — 1,818 sightings; fake countdowns and stock counters are described as distinct species.
Here: Nothing here burns down or runs out. The daily puzzle comes one per day — but that is a rhythm, not scarcity: a missed one is not "lost", tomorrow simply brings a new one. You never wait to be allowed to play.

7. Confirmshaming

Habitat: subscription pop-ups · Mechanism: shame

A trick with its own hall of fame. The refusal is worded so that saying it is unpleasant: consent is a large bright button — "Yes, I want to grow!" — refusal a small grey line, "No, I don't want to get smarter". The interface puts into your mouth a line you never said and asks you to sign it. It works on a one-second sting: most people know it is a trick, but the sting arrives before the knowledge. Press it and check.

EXHIBIT · DEFANGED
The herbarium of specimens was kept by confirmshaming.tumblr.com; the species classification is H. Brignull's, deceptive.design.
Here: The only refusal button on this site is closing the tab, and it carries no caption. The site does not comment on your leaving.

8. The Roach Motel

Habitat: subscriptions, accounts · Mechanism: sunk cost plus friction

Free to check in, no way to check out. Subscribing takes one click; unsubscribing takes a phone call during office hours, three screens of persuasion and a survey about your reasons. The asymmetry is not an accident: every step added to the exit is a share of people who gave up along the way and kept paying. In 2023 the US Federal Trade Commission sued Amazon over its Prime cancellation flow — the prosecution did not even have to invent the route's internal name, "Iliad". In 2025 Amazon settled for 2.5 billion dollars.

FTC v. Amazon.com, complaint filed 21 June 2023, US District Court, Western District of Washington; settlement of $2.5 billion, September 2025.
Here: You cannot unsubscribe from this site because you cannot subscribe to it. There is no account to delete and no data to ask back: the tab closes, the relationship ends.

9. The Infinite Feed

Habitat: social networks, video services, autoplay · Mechanism: absence of a stopping point

A book has pages; a feed has no floor. The next screen loads before this one runs out, the next episode starts by itself. The trick removes from the process the moment at which it would be natural to ask yourself whether to continue. The decision to have "a little more" is never presented explicitly — it is taken by default every few seconds. Aza Raskin, who invented infinite scroll in 2006, later regretted it publicly and compared the feed's bottomlessness to a behavioural drug.

A. Raskin on BBC Panorama, "Smartphones: The Dark Side", July 2018 — the inventor of infinite scroll on his own invention.
Here: Everything here has an end. A game finishes with a result; the daily puzzle is one per day, and once it is solved the site does not offer the next one — the next one comes tomorrow. The stopping point is built into the machinery instead of being left to your willpower.

10. Dopamine Notifications

Habitat: the phone in your pocket · Mechanism: variable reward, carried outdoors

A notification is a variable reward that reaches you when you were not playing. Sometimes the badge hides something that matters, usually nothing, and precisely that mixture makes you check: Skinner's schedule carried off the screen and into your pocket. Apps of the genre send "Your streak is in danger!" and "We miss you" not because something happened, but because you didn't. The owl mascot begging you to come back became a meme — which has not stopped it from working.

The manual is openly in print: N. Eyal, "Hooked: How to Build Habit-Forming Products", 2014 — the chapter on triggers.
Here: This site has no way to call you: no push notifications, no emails, no address of yours. It learns you exist when you open the tab and forgets when you close it.

Telling them apart in the wild

Three questions replace the whole guide. What do I lose if I don't open this tomorrow? If the answer is not "nothing", you are wearing a streak. Why can't I predict what I'll get? If there is no reason, you are facing a slot machine. What here makes stopping or leaving harder? If anything does, it is not an accident — it is a cost item engineered as revenue.

None of these devices makes a product evil by itself: a timer can be real, a level earned, a notification important. The diagnosis is made on the whole picture. Once a product measures its success in the time you spend inside it, the entire fauna gathers around it on its own. This site measures something else — whether you played better than the computer — which is why it can afford enclosures instead of hunting.

Where to go from here
To cite this page: “Dark Pattern Safari”, You vs Computer, https://youvscomputer.com/games/dark-patterns.html