A Poverty of Attention

In 1971 a man who had helped build the first thinking machines sat down to write about how organizations should handle their mail.

Herbert Simon was fifty-five. He had already done the work that would win him a Nobel in economics: the demonstration that human beings satisfice. We take the first option good enough and stop looking, because we are working with a mind of finite size in a world of infinite detail. He had also, with Allen Newell, written what a fair reading calls the first artificial intelligence program, and predicted in 1957 that a machine would be world chess champion inside a decade. He was off by thirty years, close enough to embarrass a prophet and too close to call him wrong.

So when he contributed a chapter to a Johns Hopkins volume called Computers, Communications, and the Public Interest, he wrote as one of technology's parents. The chapter was called "Designing Organizations for an Information-Rich World," and it was, in the main, a piece of practical advice about institutional plumbing.

Buried in it is the most-quoted sentence in the history of the internet, written twenty years before there was one.

A wealth of information creates a poverty of attention.

He offered it calmly, as arithmetic.

Simon noticed that information consumes something. Whatever else it does, it eats the attention of whoever receives it. So in a world where information becomes abundant, the scarce good is the attention.

When an economist says a thing has become scarce, he is making a set of predictions that follow like a proof. A scarce good acquires a price. A price attracts a market. A market attracts people whose living is made by allocating the good, and those people, being people, will follow the price, because the price is what pays them. This is the ordinary machinery by which societies decide what to make more of.

Simon saw the first half of that clearly, and he wrote about the useful half: how an institution might spend its attention well, given that attention is what it is short of. He left out what happens when the market for a scarce good is someone other than the person who owns it, and in 1971 there was no reason to raise it.

That arrangement was, by then, already a hundred and forty years old.

In September 1833 a printer named Benjamin Day put out a New York paper called The Sun and priced it at a penny. Every serious paper in the city cost six cents and made its money from the people who read it. Day's cost less than it cost to print. He made his money somewhere else, from advertisers, and what he sold them was the readers.

Tim Wu, in The Attention Merchants, dates the industry to that decision, and the date matters more than it looks. The arrangement is a nineteenth-century business model that has been waiting patiently, through every new medium, for better equipment. The internet inherited it. So did the smartphone.

It took two years to produce its first masterpiece. In August 1835 the Sun ran six articles reporting the discovery of life on the moon (bat-winged people, unicorns, temples of sapphire), credited to a real astronomer who knew nothing about it. Circulation soared. The paper admitted the fabrication three weeks later and kept the readers. Nearly everything the next two centuries would argue about was present in the first prototype, including the part where being caught costs less than the attention was worth.

The equipment arrived in stages. The illustrated poster, with which advertisers learned that a street could be sold. The radio hour underwritten by a soap manufacturer, which gave us the word soap opera and, more importantly, the idea that a story could be a vehicle. The television program engineered to carry you across the break. Each generation performed the same trick with better tools, and each generation was met by people who found the whole thing vulgar and were, in the end, ignored.

What changed in the last twenty years was the instrumentation.

There is a line every advertising man used to quote: half his spending was wasted and he didn't know which half. It is attributed to John Wanamaker, and he probably never said it, which is somehow appropriate. For most of its history the attention industry was working blind, buying something it could not see and could not count.

The feed can see. It can count. And a thing that can be measured can be optimized, and a thing that is optimized long enough stops being designed and starts being grown: selected, release after release, test after test, for a single number. Nobody has to intend the result. Selection needs no conspiracy and little talent. It needs a metric and patience, and it will find things about human beings that no designer would have thought to look for and no focus group would admit to.

That is the honest description of what happened, and it is worse than the villain story, because there is no one to fire.

Running the whole way alongside it is a word.

Every one of these machines, at its introduction, was sold as a cure for human distance. The telegraph would end war; in 1858 two of its chroniclers wrote that old prejudices and hostilities could not survive it. Radio would bind the nations. Television would let us see each other's faces and, seeing them, behave. And then, in our own time, the mission statements: bringing the world closer together, giving everyone a voice, connecting every person on earth.

The uncomfortable part is that all of it was partly true. You can speak to your mother from a bus in another hemisphere for the price of nothing. A person who is ill can find, in an afternoon, forty people with the same rare diagnosis. That is not a small thing and it deserves better than a sneer. Anyone who writes about this subject as though the last century were simply a mistake is asking you to give up something real, and you should not listen to them.

The connection was the feature. The attention was the business. And when the two pull apart (they always eventually pull apart, because one is what a thing is for and the other is what it is paid for), the business wins, on a schedule you could set a watch by.

We built machines for attention and called them connection.

The word was a description of the product, accurate enough to be repeated sincerely by the people building it, and inaccurate in precisely the direction that made it profitable. Those are the most durable kinds of words. They cannot be argued with, because everyone involved means them.

This subject has a bad habit: it reaches for the strongest available finding and then leans on it harder than it can bear.

Consider the most charming study in the field. In 2017 Adrian Ward and his colleagues published work in the Journal of the Association for Consumer Research showing that the mere presence of your own phone (face-down, silenced, untouched) measurably reduced working memory and fluid intelligence, and that the effect was strongest in the people most attached to the device. It is an irresistible result. It has been cited thousands of times and repeated at a thousand dinner tables, including by me.

A pre-registered replication published in 2022 did not find it.

Two meta-analyses since have split. One, pooling twenty-two studies, reports a real but small effect concentrated almost entirely in memory, and not statistically significant in North American or European samples. The other, covering fifty-six studies and seven thousand participants, finds one surviving effect out of six and calls the support limited and smaller than theorized.

That does not make the original wrong. Psychology is hard, effects are small, and none of these settles anything alone. But it means the correct statement is that one of the most repeated facts about phones is currently contested, and anyone who tells you otherwise is selling something.

The same discipline has to be applied all the way down. The 2023 U.S. Surgeon General's advisory on loneliness is a serious document about a measured thing: social isolation is widespread and it tracks with mortality. Its most quoted line (that the risk is comparable to smoking fifteen cigarettes a day) comes from a 2010 meta-analysis that measured social connection: network size, social integration, perceived support. Loneliness was never the variable. The finding is real and it has held for fifteen years across millions of participants; the word attached to it changed somewhere in the retelling. The isolation is measured. The cigarette is a metaphor wearing a lab coat.

Jonathan Haidt's The Anxious Generation argues that a rewiring of childhood by smartphones and social media caused the collapse in adolescent mental health after 2012. It is the most consequential version of this argument anyone has made. It was also reviewed in Nature by Candice Odgers, who has spent her career on exactly this data and concluded that the causal claim outruns the evidence. She has argued, pointedly, that blaming the phone may be a way of not looking at causes we could actually do something about. Amy Orben and Andrew Przybylski, working across datasets covering more than three hundred thousand adolescents, found that technology use accounts for at most 0.4 per cent of the variation in adolescent wellbeing, an association about the size of the one with eating potatoes, and smaller than the one with wearing glasses. They also found something more unsettling about the field itself. Across those three datasets there were over six hundred million defensible ways to run the analysis, and the choice among them could deliver a harm, a benefit, or nothing at all.

The science is not settled, and there is nothing to be gained by pretending it is.

Some of it requires no settling at all.

Nobody disputes what the machines are for. Findings come out of studies. This one comes out of a filing: disclosed quarterly, audited by an accounting firm, and read aloud on earnings calls to people who would sue if it were misstated. The revenue line says, in plain language, which side of the transaction you are on. You do not need a p-value to read an income statement.

The case rests on purpose. A thing built to hold your eyes will get better at holding your eyes, year over year, forever, because in that architecture "better" has only one available meaning. Whether or not it is damaging you is a live empirical question. What it is trying to do is a matter of public record.

Simon said poverty, and we heard overwhelm. Those are different words.

Overwhelm is a feeling. It arrives, it is unpleasant, it passes; you close the laptop and go for a walk and it lifts. Poverty is a condition of the account. It sets what you can afford.

Whatever the studies eventually say about calm and happiness, what the arrangement has taken is a specific and unglamorous capacity: the ability to stay with one thing long enough for it to become strange.

That capacity is the entry fee for nearly everything worth having. A difficult page only opens on the third reading. A person only becomes interesting after the part where they are not being interesting yet. A question only turns useful somewhere past the point where holding it is uncomfortable. Grief, prayer, mathematics, love, and every craft anyone has ever been good at run on the same faculty, and it is exactly the faculty that a market for attention is structurally obliged to break down into smaller and more liquid units.

Nobody lost this by being weak.

We trained. Every one of us has spent between ten and twenty years in continuous, well-funded, superbly designed instruction in a single skill: the rapid evaluation and discard of the present thing in favour of the next one. The curriculum was excellent and the feedback was immediate. The practice was distributed across every idle moment of the day, which is precisely how expertise is built.

We are all very good at it now. It is the most widely held expert skill on earth and not one of us chose to learn it.

You cannot be ashamed of a skill. You can only decide whether you want to learn a different one.

The loneliness is the strangest part, because on paper it should not be there.

Never in the history of the species have so many people been so continuously available to so many others. Every person you have ever met is, at this moment, one thumb-motion away. By every measure the nineteenth century would have used, we have solved distance. And the measured experience of the last decade has not been closeness.

Sherry Turkle saw the shape of it in 2011, before the smartphone was universal, and put it in nine words: we expect more from technology and less from each other.

The economics of that are simple and merciless. A substitute does not have to be as good as the real thing. It only has to be good enough, and cheap enough, and available at the moment the real thing is not. A message is a decent fraction of a conversation at a tiny fraction of the cost, and it is available at eleven at night, and the person on the other end never has to be met, or travelled to, or forgiven. So the fraction wins. Nobody was ever offered that choice in a single sitting. The price changed, one convenience at a time, and the market for the rest of it quietly thinned.

Nobody decided to stop having friends.

The usual response is discipline. Delete the applications. Grey out the screen. Put the phone in a wooden box during dinner. Some of it genuinely works, and none of it should be mocked by anyone who has tried it.

It is also a private tax, paid by individuals, against a system with a research department. Willpower makes a poor policy. Asking several billion people to out-concentrate an optimizer assigns the blame downward.

Simon's arithmetic points at a different response, and it is a question of design.

If attention became a commodity because information became abundant, then the interesting question is what could be sitting on the other side of the transaction.

Every machine in the history above answers the same question: how do I keep you here. It is possible to build one that answers a different question. Hard, of unproven profit, and possible.

What would a machine be, if the measure of its success were how you were doing an hour after you closed it?

That is a specification, and it has consequences, most of them inconvenient. It means the thing has to be able to end: a real ending, taken deliberately, while you are still enjoying it. It means the metric that would make it money and the metric that would make it good point in opposite directions, and somebody has to decide, in advance and in writing, which one loses.

A thing that ends has stopped harvesting. Those are the same sentence read from either side.

Simon named the poverty and then went back to work on the machines. He saw no contradiction in that, and there was none. The bill came due, and the arithmetic that produced it is a sequence of decisions, taken by people, mostly for money, in rooms that still exist.

And poverty, unlike loss, is a description of an account at a moment.

Accounts can be refilled.

You've been quite busy.
— Gnoia