First principles · Attention
The light was never the point
The Hawthorne effect, free observation, and why only attention that could have gone elsewhere changes anyone.
Being regarded is an input. But only regard that costs someone something moves a person.
Your manager stops at your desk on her way to a meeting she is already late for. She reads the first page of the thing you made, says “this part is good,” and keeps walking. You work differently for the rest of the week.
Nothing about your job changed in that moment. Same desk, same tools, same deadline. She did not give you information you lacked; you already knew the part was good. What changed is that someone with forty other places to be spent ninety seconds on you, and you felt it the way you feel a hand on your shoulder. Everyone has had this week. Most of us have also had the opposite: months of careful work that no one opened.
There is a name for what happened at your desk, and it comes from a factory floor a hundred years ago.
The lights at Hawthorne
In the mid-1920s, engineers at Western Electric’s Hawthorne Works outside Chicago set out to answer a practical question: does better lighting make workers more productive? They raised the light in a test room, and output went up. They lowered it, and output went up again. In the version of the story that became famous, they dimmed it almost to moonlight and the women assembling relays kept working as well as ever. Later studies in the same plant, on rest breaks and pay and hours, kept turning up the same pattern. Whatever the researchers changed, the workers improved.
The conclusion people drew, and the one a sociologist named Henry Landsberger gave a name to in 1958, was that the light was never the point. The workers were responding to being studied. People in the test room had been chosen. Someone from management was asking how they felt, writing down what they did, coming back the next day to see. Being regarded was itself the intervention.
I should say plainly what is record and what is legend. When economists Steven Levitt and John List went back to the surviving illumination data in 2011, they found the famous effect far weaker and messier than the story. A large review of the research in 2014 concluded that the effect of being studied is real but varies a great deal, and that the word “Hawthorne” has been stretched over too many things. The legend is cleaner than the data. But the principle underneath it has outlived the correction, because every manager, teacher, coach and parent has watched it work.
Here is the principle, stated so you could explain it at dinner. Attention changes people, but not all attention. The kind that changes people has three properties.
- It comes from someone who matters. A stranger’s glance does little. The regard of a person whose judgment you care about does a great deal.
- It is news. It arrives as a change. The first visit to the test room moves people; the hundredth is furniture.
- It was scarce. It came out of a finite supply. Someone could have spent that hour anywhere and spent it on you.
The third property is the one people forget, and it is about to become the only one that matters.
- Scarce regard: a person chooses to look
- Constant observation: a system that always looks
What happens when watching is free
Artificial intelligence makes observation nearly free. A machine can read every email you send, every call you take, every page you draft, and comment on all of it, all day, without getting tired or bored or late for a meeting. If you believe the Hawthorne story is about being watched, this looks like the most powerful productivity tool ever built. Watch everyone, always, and everyone improves.
I think the opposite happens. The lift at Hawthorne was never about being seen. It was about being chosen. And observation that costs nothing carries no information about whether you were worth choosing.
A look that could not have gone anywhere else tells you nothing about yourself.
My Reality Equation gives the mechanism a precise shape. Reality is the ratio of what actually arrives to what you expected, and the surprise you feel is its logarithm.
R = A / E · S = ln R
The first time someone regards you, the Actual outruns the Expectation, and you feel it. Nobody can reach into your Expectation and set it; it moves only one way, as the things that actually happen to you pile up into history and retrain what you predict. So when a system watches you every hour of every day, that watching becomes part of what you expect within weeks. On that dimension the ratio settles at one, and the logarithm of one is zero. Nothing left to feel. Constant observation does not fail because it is cruel or because it is wrong. It fails because it is expected.
The human visit keeps working for the reason the machine’s stops. It stays scarce, so it stays news. The lift does not disappear when watching becomes free. It moves to the one kind of attention that cannot be copied: an hour from a person who had other hours.
Your delegate has read every page you wrote this year and left careful, accurate notes on each one. You stopped opening them in February. Then a woman you have met twice writes three lines: she read the whole thing on a plane, and the second section changed her mind. You rewrite the ending that night.
Five predictions
Monitoring will stop moving behavior within weeks of being switched on.
Workplaces that track everything with AI will see a bump when the system starts, then nothing. The bump was novelty, and novelty is the one thing a constant system cannot keep producing. People will not resist it so much as stop noticing it.
Management will shrink to the deliberate spending of regard.
When machines do the checking, the measuring and the reminding, a manager’s remaining lever is where she visibly spends her own hours. The good ones will plan their attention the way they once planned reviews: who has not been seen lately, whose work deserves a real read this week.
Automated praise will fall to zero and costly praise will rise.
“Great job” from a system that says it to everyone means nothing, the same way a form letter means nothing. A note that obviously took a person twenty minutes will be worth more than it has ever been, because effort is the only thing that proves the regard was real. I made the same argument about effort and identity in Proof of Person.
AI coaches and tutors will show a lift that fades, and the lasting ones will bring a human in.
The first weeks with a tireless tutor will look miraculous. Then the curve will flatten. The products that keep their results will be the ones that know when to bring in a person, and protect that person’s scarce time for the moment it counts.
People will stop performing for machines and start working for readers.
We behave differently for an audience that might look than for one that always looks. As machine observation becomes background, the work people care most about will be shaped for the few humans who might actually read it. The next audience that matters is small, specific and human.
Where I stand
I publish here every day, and I know machines read each post within minutes of it going live. That knowledge does not change a word I write. My own working life is built around delegates that watch everything, keep track of everything, and never tire. That is exactly why, if this principle is right, the thing that would change how I write is a single person telling me they read one all the way through. I wrote about the human beside the delegate in The Attender, and about attention as a gift in Undivided. This is the same idea seen from the other side: once the machines attend to everything, the attention that moves people is the attention that could have gone elsewhere.
The ledger
- Already true
- People change when they know they are being studied, even if the original lighting story was overstated. And we already ignore the automated “how did we do?” email while a handwritten card from a stranger stays on the fridge.
- What has to happen
- Machine observation has to become ambient and cheap, and people have to know it is cheap. The lift dies when everyone understands that the watching costs no one anything.
- Where I am probably wrong
- If people come to feel chosen by a machine, if an AI’s attention starts to feel like it could have gone elsewhere, the lift survives and this argument fails. Fear is the other trap: surveillance can force compliance that looks like improvement on a chart. That is not the Hawthorne effect, but it can hide inside the same numbers.
Go back to your desk on that Monday morning. Your manager did not tell you anything new. She did not measure you or correct you. She could have been anywhere, and for ninety seconds she was here. That was the whole effect, and no amount of free watching will reproduce it.
The light was never the point. She was.
Background
- F. J. Roethlisberger and William J. Dickson, Management and the Worker (Harvard University Press, 1939). The original account of the Hawthorne studies.
- Henry A. Landsberger, Hawthorne Revisited (Cornell University, 1958). Where the effect got its name.
- Steven D. Levitt and John A. List, “Was There Really a Hawthorne Effect at the Hawthorne Plant? An Analysis of the Original Illumination Experiments”, American Economic Journal: Applied Economics 3(1), 2011.
- Jim McCambridge, John Witton and Diana R. Elbourne, “Systematic review of the Hawthorne effect: New concepts are needed to study research participation effects”, Journal of Clinical Epidemiology, 2014.
- John Rector, The Reality Equation.
- John Rector, The Zero-Attention Economy.