A Principle ·
Nothing Left but Wanting
Half a century of psychology says the reasons we are given decide what we come to love. AI is about to take the reasons away.
More reason outside, less wanting inside.
You do not read your own motives. You infer them, the way a stranger would, by watching what you do and looking for the reason. Find a reason outside, and the mind credits the outside. Find none, and it concludes: I must want this.
The assistant has already offered twice. Dinner for four, the usual place, arriving in twenty-five minutes; one word and it is done. You say no, and keep chopping the onion.
Your eyes sting and the kitchen smells right, and you notice you cannot say why you said no. Nobody is making you. Nobody is grading it. That small blank where a reason used to be is the most interesting thing that happened to you today.
Twenty dollars and one dollar
In 1959, at Stanford, Leon Festinger and James Carlsmith gave students an hour of deliberately dull work: turning pegs on a board, emptying and refilling a tray of spools. Afterward, each student was asked to tell the next participant that the task had been enjoyable. Some were paid twenty dollars to tell that lie. Others were paid one dollar. Later, when asked how much they had actually liked the task, the one-dollar students rated it higher. The twenty-dollar students had an explanation for what they had done: the money. The one-dollar students did not. A dollar is not a reason to lie. So their minds supplied the only reason left. Maybe it was not so bad.
Fourteen years later, also at Stanford, Mark Lepper, David Greene and Richard Nisbett ran the same law in the other direction. They found preschoolers who already loved drawing with felt-tipped markers. Some were promised a Good Player award, a certificate with a gold seal and a ribbon, for drawing. Others drew with no promise at all. Weeks later, with the markers back out during free play and no award in sight, the children who had drawn for the certificate drew less than the children who had drawn for nothing. They had been given an outside reason, and their own reason quietly moved out.
Daryl Bem gave the pattern its plainest form: we come to know our own attitudes partly by watching our own behavior, the way an observer would. Pay a person to do a thing and the observer inside concludes it was for the pay. Give them no reason and the observer concludes it was for love. Psychologists call one half the overjustification effect and the other insufficient justification. They are the same law. The mind keeps a ledger of reasons, and outside reasons crowd the inside ones off the page.
Paid twenty dollars
outsideinside“I did it for the money.”Paid one dollar
outsideinside“I must have liked it.”
The mind does not read its own motives. It infers them, like a stranger watching through a window.
The largest reward was always necessity
Here is what the experiments never had to account for. For almost all of history, the biggest outside reason in a human life was not money or gold seals. It was necessity. You cooked because people had to eat and no one else would cook. You wrote the letter because the letter would not write itself. You drove because the car would not drive. Necessity was the twenty dollars of ordinary life, and it was paid out on nearly everything we did. Which means that for most of what we did, the observer inside never had to ask whether we wanted to. The reason was standing right there.
AI is withdrawing necessity, task by task. Not all at once and not everywhere, but steadily: the meal that can be ordered by a sentence, the email that can be drafted before you finish wanting it drafted, the route, the form, the follow-up, the first draft of almost anything. Every time an agent could have done the thing and you do it anyway, you have run Festinger’s experiment on yourself, at the one-dollar rate. There is no outside reason left to file the act under. So the mind files it under the only heading remaining. You must want this.
That produces a sorting nobody has lived through before. When necessity leaves an activity, one of two things happens. Either the activity collapses, and you discover that you never wanted it, that you were doing it for the twenty dollars all along. Or you keep doing it, and keeping it with no reason makes you want it more. Some things fall away. The things that remain get heavier. Your loves stop being an accident of what you were forced to do and become a record of what you kept doing when nothing forced you.
And the law runs in its other direction at the same time. The same machines that remove necessity are superb at adding rewards: streaks, scores, badges, progress rings, a cheerful voice telling you how well you did. That is the gold-seal certificate, delivered by the billion. Anything an AI scores you on is at risk of becoming something you do for the score. The quiet danger in a personal AI is not that it will do your loves for you. It is that it will praise you for them.
Your teenager has spent the whole afternoon in the garage rebuilding an old bicycle that a repair service would have fixed by Tuesday, collected and returned without anyone lifting a wrench. Nobody assigned it, nobody is tracking it, and it will not count for anything.
You ask why. A shrug: “I just like it.” Festinger would tell you that shrug is the sound of a value being made.
Five predictions
The great sorting of chores from loves.
As agents take the necessity out of cooking, writing, planning and fixing, people will find out, often to their surprise, which of these they loved. Many will drop activities they assumed were part of who they were. A smaller set will become central to their identity, held more tightly than before.
“Unnecessary” becomes how people describe themselves.
“What do you do?” will be answered less by the job and more by the things a person does that a machine could have done. The sentence “I still do that myself” will carry the weight that a job title used to carry.
The best personal AIs will learn to withhold praise.
The design lesson from the marker study will reach the stack: do not score, streak or reward the things a person does for love. The assistants people trust most will be the ones that notice what you keep doing by hand and leave it alone, with no badge and no applause.
The people who stay in a craft will be the ones who never needed to.
When the work no longer requires a person, the people who remain in it will be the ones for whom pay was never the whole reason. Crafts will thin out and then deepen, and their practitioners will be harder to pay off and harder to discourage.
Children raised beside capable agents will have fewer, stronger loves.
A child who could have had anything done for them, and did some things anyway, will have interests that were chosen rather than imposed. Expect narrower sets of passions, held with unusual conviction, and expect it to read to older adults as stubbornness.
I am the experiment
I run my working life on a stack of agents, and I have already told them: you do it, you decide. So I am not observing this law from a distance. Everything I still do by my own hand is now, by definition, insufficiently justified. There is no outside reason left for it. I find that clarifying. Festinger says I will not be told what I love. I will find out by watching what survives. I have written before that you are what you still do. This is the reason underneath it: what you still do, when nothing makes you, is where the wanting goes.
- Already true
- The core findings are old and widely repeated: Festinger and Carlsmith’s one-dollar students, Lepper’s marker-drawing children, and a large 1999 meta-analysis by Deci, Koestner and Ryan concluding that expected, tangible rewards tend to undermine interest in tasks people already find interesting. And agents already remove necessity from meals, messages, scheduling and drafts for anyone who asks them to.
- What has to happen
- People have to feel the alternative as real. Insufficient justification only works if the mind registers “I could have had this done” at the moment of acting. If the agent is invisible, or the option never occurs to anyone, there is no missing reason and no sorting. And the machines must not refill the gap with rewards faster than they empty it.
- Where I am probably wrong
- The overjustification effect is contested. Some researchers, Judy Cameron and David Pierce among them, argued the undermining is narrow and that rewards tied to doing well can raise interest. More seriously, necessity may not compete with love at all. It may be the scaffold love grows on, and without it people may simply drift and keep nothing. If, ten years from now, people with full AI help are no more attached to the activities they kept than people without it, or have kept almost none, I was wrong.
The onion
Back in the kitchen, the assistant has gone quiet. It offered, you declined, and it did not try again. The onion did not need you. Dinner did not need you. That is exactly the point. For the first time, nothing about this was required, so nothing outside you can take the credit for it.
You are not cooking because you have to. You are finding out that you want to. Take away every reason the world used to hand you, and what remains is the most honest thing you will ever learn about yourself.
Nothing left but wanting.
Background
- Festinger, L. & Carlsmith, J. M. (1959). Cognitive consequences of forced compliance. Journal of Abnormal and Social Psychology. pubmed.ncbi.nlm.nih.gov/13640824
- Lepper, M. R., Greene, D. & Nisbett, R. E. (1973). Undermining children’s intrinsic interest with extrinsic reward: A test of the “overjustification” hypothesis. Journal of Personality and Social Psychology.
- Bem, D. J. (1972). Self-perception theory. Overview: Self-perception theory, Wikipedia
- Overview of the effect and its critics, including Deci, Koestner & Ryan (1999) and Cameron & Pierce (1994): Overjustification effect, Wikipedia
- Rector, J. (2026). You Are What You Still Do
- Rector, J. (2026). Unnecessary Obstacles · The Safety Will Be Spent