Newcomb's Paradox
You are shown two boxes. One is glass and holds $1,000. The other is sealed. A predictor that is almost never wrong has already decided what goes in the sealed one, based on what it thinks you will do. Do you take one box, or both? Three steps: play it, watch a crowd play it, then see why clever people still argue about it.
Your turn
Before you walked in, the predictor studied you and made its call. If it predicted you'd take only box B, it put $1,000,000 inside. If it predicted you'd take both, it left B empty. It is right 90% of the time. The boxes are sealed now. What do you take?
Pick one. There is no trick in the boxes themselves.
Now watch a thousand people play
One round is luck. So send in a crowd: 500 people who always take only box B, and 500 who always take both. Each dot is one person. Gold means they walked out a millionaire.
So why is this a paradox?
Because there is a second argument, just as solid, that says the exact opposite. Read both and notice which one you believe.
Take only box B
Your choice is a clue to what the predictor foresaw. People who take one box usually find the million. Slide the accuracy and watch the average payout of each choice.
Take both boxes
The money was placed before you arrived. Whatever is in box B is already there, and nothing you do now can change it. So compare your two options row by row:
| If box B holds… | Only B | Both |
|---|---|---|
| $1,000,000 | $1,000,000 | $1,001,000 |
| nothing | $0 | $1,000 |
Whatever is inside, taking both always gets you $1,000 more.
Worth knowing
A physicist's puzzle
William Newcomb, a physicist, came up with it around 1960, and the philosopher Robert Nozick made it famous in 1969. Nozick wrote that almost everyone finds the answer obvious, and that people split almost evenly over which answer that is.
Going together is not causing
Taking one box goes together with getting rich, the same way carrying an umbrella goes together with rain. Whether that should change what you do depends on whether your choice can make any difference to the outcome, and that is a question the odds alone cannot answer.
It is not really about boxes
The same split shows up whenever a decision is also a signal: a firm that knows competitors will read its price as a clue, a voter who wonders whether people like them will vote the same way, or an AI system deciding how far to trust a model of its own behaviour.
The predictor here is a simple simulation: it guesses your choice correctly with the probability on the slider, and everything runs in your browser.
Try another puzzle →