On the first morning, the phone belonged to a stranger. We had wiped it back to its out-of-box self and made a pact: for one week, the desk would live as the user the product was actually designed for. Every prompt accepted. Notifications on. Personalisation on. Autoplay on. Location shared with a generosity we normally reserve for close friends. “Improve your experience” — yes, obviously, whatever that means; agreeing without reading was rather the point. By lunch the lock screen had become a noticeboard in a building we didn’t remember entering. Something wanted to tell us about a sale. Something else had assembled a slideshow of memories we had not asked it to keep.
The standard complaint about defaults is that they’re annoying — too many pings, too much autoplay, an inbox filling with things you never asked for. True, and entirely beside the point. What a week inside the untouched configuration makes clear is that a default is not the absence of a choice. It is somebody else’s choice, made in advance, wearing the costume of no choice at all. The person who made it has interests, and those interests are not identical to yours. Living on factory settings means living inside someone else’s argument about what you want — and the argument is winning, because you never noticed one was being made.
The laboratory version of this has a name. In a 2012 paper in Judgment and Decision Making, Nikhil Dhingra, Zach Gorn, Andrew Kener and Jason Dana at the University of Pennsylvania isolated what they called the “default pull”: even subtle defaults change what people come to prefer, partly by nudging them to consider whether they might prefer the preset. The effect held apart from the usual explanations — not laziness, not switching costs, not loss aversion — and in one experiment a default moved average giving by 17 per cent of the endowment. Their paper cites earlier work by McKenzie, Liersch and Finkelstein on “information leakage”: people read a default as a recommendation from whoever set it. The maker probably knows best, we assume. And so every default functions as a suggestion from an interested party, accepted at scale by people who didn’t notice a suggestion was being made.
Out in the world, the leverage has been documented for years, most readably in a 2016 ProPublica rundown of how defaults rule it. Under the American opt-in system for organ donation, about 40 per cent of the population is registered; in opt-out countries such as Spain, Portugal and Austria, registration runs above 99 per cent. A New York middle school moved its salad bar from against the wall to the centre of the cafeteria, in front of the cash registers, and salad sales more than tripled. Automatic enrolment does the same for retirement savings; in one company’s records, participation among new hires climbed from under 20 per cent. Richard Thaler and Cass Sunstein built Nudge on this material in 2008 and gave the practice its name: choice architecture. The quieter half of their argument tends to get lost. There is no neutral architecture. Some arrangement of the cafeteria is unavoidable. The only question is who arranges it, and toward what end.
Toward what end is not much of a mystery. A 2020 review by the UK’s Centre for Data Ethics and Innovation, working with the Behavioural Insights Team, found that online defaults are typically set in businesses’ interests rather than users’: personalised advertising switched on by default in Google Search and Bing, autoplay on by default on most major streaming services. The untouched configuration is not a natural resting state for software. It is a showroom route, laid out so that you pass the register.
So: what does the showroom feel like from the inside? Populated, mostly. The default phone is a shared building in which every app keeps office hours near yours. The interruptions accumulate not as emergencies but as a steady drizzle of almost-importance — a suggestion to reconnect with someone you last spoke to in another decade, a badge count that never reaches zero because zero is not the goal. What surprised us was the speed of acclimatisation. Within days, the drizzle stopped reading as noise and started reading as weather. That is the product working as designed, and it is the mechanism the research describes. The default pull doesn’t feel like a pull. It feels like settling in.
The recommendations drifted next. With personalisation on, everything half-watched counts as a vote; with autoplay on, the next episode begins while you are still deciding whether you wanted the last one, and indecision is recorded as consent. By midweek our feeds had assembled a portrait of a person who enjoys things none of us remember enjoying — a composite built entirely from our least attentive moments. The machine met us at our worst and took us at our word. The permissions, meanwhile, were granted in passing, each tap trivial on its own. Nothing about the week involved a dramatic handover of data. It was more like leaving small change in every room of a very large house, and only later realising you can no longer list the rooms.
Here is where it stops being a story about clutter. Somewhere, a dashboard recorded all of it. We kept the notifications on; we watched whatever autoplay served; we never once opened the settings screen. In the data, we are indistinguishable from satisfied. A preference expressed by ninety per cent of users through inaction is not a preference — it is a measurement of how much effort opting out costs, weighed against how much attention people have left to spend. But the dashboard cannot tell the difference, so the loop closes: the default manufactures the behaviour, and the behaviour is cited to justify the default. That is the Dhingra result at platform scale. The setting doesn’t reflect what users want. It produces what users do, then takes a bow for listening.
The industry even has its own default statistic, fittingly enough. Ask how many people ever change their settings and you will hear a figure in the mid-nineties, usually attributed to Microsoft research. We went looking for the source and found the citation equivalent of a friend of a friend — a number repeated for years because it arrived pre-installed and nobody bothered to reconfigure it. As demonstrations of the thesis go, you could hardly build a better one.
When the week ended, we changed everything back, and the asymmetry is the part worth keeping. Living as the default user required nothing of us; it happened the way a current happens to a swimmer who stops swimming. Leaving required menus, toggles, decisions — an afternoon of administrative self-respect. Every default is that asymmetry formalised: a bet that the cost of changing the setting will always outweigh the will to change it, multiplied by everyone. If the people building these products want an honest number for their dashboards, it isn’t how many users leave a setting on. It’s how many switch it off the moment the switch is shown to them — the one cohort on the chart actually telling you anything. The rest of the silence isn’t satisfaction. It’s just the sound of the current.