This Paradox Splits Smart People 50/50

At a glance

You walk into a room. Two boxes. One is open with $1,000 you can see. The other is sealed. A supercomputer that has correctly predicted thousands of people before you has already decided what is inside the sealed box: $1 million if it predicted you would take only the sealed box, nothing if it predicted you would grab both. The boxes are set. The prediction is locked. Now choose. Take both, or take only the mystery box.

This is Newcomb's paradox, and the unsettling thing is not that it is hard. It is that the answer feels obvious to everyone, and yet smart people land on opposite answers in almost equal numbers, each convinced the other side is being silly. Veritasium runs the experiment on its own staff and on strangers on the street, watches them split right down the middle, then carefully lays out both arguments with the actual expected utility math. The one-boxers walk away millionaires. The two-boxers have an argument that is airtight. Both are right, and that contradiction is the whole point.

This is not a magic trick to be debunked. It is a clean wedge that pries open three of the deepest questions there are: does free will exist, what does it actually mean to be rational, and is there an ideal way to act in life. The video follows the puzzle out from a parlor game to nuclear deterrence and ends somewhere genuinely useful. Below is the entire argument, both camps, every number, in the order Veritasium builds it.

The setup, stated exactly

The cold open is a confession. One Veritasium staffer admits there is a problem he cannot bring up without starting a fight, that it has infiltrated every single Veritasium meeting for two months, that he even argued with Derek about it. Then the setup, delivered with no spin.

You enter a room. On the table sit a supercomputer and two boxes. One box is open and contains $1,000. There is no trick. You can see the cash. The other box is sealed and you cannot see inside. You know one more thing: this supercomputer is extremely good at predicting people. It has correctly called the choices of thousands of people facing the exact problem you are about to face. It has been right almost every time.

Your options are only two. Take both boxes, the sealed one plus the visible $1,000. Or take only the sealed mystery box and leave the $1,000 on the table.

What is in the mystery box depends on a prediction the computer already made, before you ever walked in:

• If it predicted you would take only the mystery box, it put $1 million inside.
• If it predicted you would take both boxes, it put nothing inside.

The prediction is finished. The money is already in the box or already not. The computer is not trying to trick you and not trying to cheat you out of anything. Its single goal is to predict correctly. And it does not matter at all how it predicts. Replace the supercomputer with a super intelligent alien, a cunning demon, or a team of the world's best psychologists. All that matters is that the predictor is extremely accurate and made the call before you entered the room. Pause and decide.

Pick one box

Veritasium asks staff and strangers, and the answers come back instantly and confidently in both directions. "I should just take two boxes, like, obviously." "I'm just gonna take the $1 million and go with it." "Of course you take two boxes." "I would not get the two boxes." One person works through it out loud and lands on: "This is seeming less paradoxical than I thought because I should just go in and take the mystery box only." The reply, off camera, is a baffled "No! What?"

So there are two tribes. One-boxers take only the sealed box. Two-boxers take both. The American philosopher Robert Nozick captured exactly why this is famous: "To almost everyone, it is perfectly clear and obvious what should be done. The difficulty is that these people seem to divide almost evenly on the problem, with large numbers thinking that the opposite half is just being silly."

The puzzle is named for its inventor, the physicist William Newcomb, and popularized by Robert Nozick in 1969. It splits people for real, not just in theory. When The Guardian polled over 31,000 readers in 2016, the result was 53.5% one-boxers to 46.5% two-boxers. A coin flip, give or take. The video then does the responsible thing and gives the strongest version of each argument rather than picking a winner.

The one-boxer case, in the words of a one-boxer: "I'm a reasonable guy and I like money, so I'm gonna do whatever gets me the most money." So weigh the two decisions properly. Call C the probability the computer predicted you correctly, which means 1 minus C is the probability it got you wrong. Now compute the expected utility of each choice.

Two-box:  EU = C × $1,000      + (1 − C) × $1,001,000
One-box:  EU = C × $1,000,000  + (1 − C) × $0   =  $1,000,000 × C

If you try to two-box, there is a C chance the computer saw it coming and left the mystery box empty, so you walk with $1,000, and a 1 minus C chance it guessed wrong, left the million in, and you scoop $1,001,000. If you one-box, there is a C chance the million is there and a 1 minus C chance the box is empty. Set the two expected values equal and solve for the break-even accuracy. The crossover sits at C = 0.5005, just a hair over 50%. In other words, the instant the computer is even slightly better than a coin flip, one-boxing has the higher expected payout. And this computer is far better than a coin flip; it nailed thousands of people before you. So the one-boxer rests the case, takes the sealed box, and says "Here's my $1 million. Take that, Casper." The math is not vibes. It is a weighted average that tips to one box for any predictor worth the name.

Pick both

Now the other side, and it is just as clean. A two-boxer hears the one-box reasoning, finds it surprising, and says the answer is also obvious, only it is the opposite. "Here's how I think about the problem in a way that actually makes sense."

The supercomputer has already set up the boxes. Whatever you decide now does not change whether there is zero or a million dollars in the sealed box. The contents are fixed. So lay out the four possible worlds you could be in and ask what your choice does inside each.

Read it column by column. If the sealed box is empty, two-boxing nets you $1,000 instead of nothing. If the sealed box holds the million, two-boxing nets you $1,001,000 instead of a flat million. Either way, whatever is already in that box, two-boxing leaves you exactly $1,000 richer. This is strategic dominance: one option beats the other in every possible state of the world, so you take it. "So, give me those boxes." Anyone in their right mind picks both.

Even the people who lean one-box feel the pull of this. One staffer admits "the two-boxer argument makes a lot of sense to me. Once you explain it, I'm like, okay, yeah, I can see why exactly you're right." But then he adds the thing that keeps him one-boxing: "I can also see that just having those thoughts in your brain are what might allow the computer to give you nothing. I think I'm grateful that I just don't have those thoughts." That single line is the whole tension. The dominance argument is correct, and being the kind of person who is moved by it is exactly what gets your box left empty.

What decision theory actually is

So you have two perfectly reasonable approaches that give two completely different answers. Veritasium's framing is the key that unlocks everything that follows: your choice reveals something fundamental about how you make decisions. It comes down to two statements, both of them true.

1. As far as you know, basically everyone who took one box walked away a millionaire, and everyone who took two walked away poor.
2. The supercomputer made its prediction before you even knew about the problem. The boxes are already set up, so your decision now cannot change whether the million is in there.

Both true. The split happens because each statement carries a hidden assumption, and which one you privilege quietly decides your camp. The two assumptions correspond to two formal schools of decision theory, the field that studies how to choose under uncertainty.

The one-boxers run on evidential decision theory, EDT. As one of them puts it, the expected utility calculations are based on probabilities that use the prior evidence of how accurate the computer is. The thousands of people it predicted correctly are evidence enough that when you reach for one box, the million is waiting. EDT asks: given that I choose this, what should I now expect to be true about the world? Choosing one box is strong evidence that the million is there, so choose it. Veritasium notes its own audience leans this way hard. They polled viewers and got more than 24,000 responses, and two thirds came back one-boxers.

The two-boxers run on causal decision theory, CDT. The motto: "I believe that whatever I do now can't influence and change the past. I only take into account things that I can actually influence." Since the prediction and the box contents were set before you learned about the problem, your choice cannot cause the million to appear. So CDT swaps in a different probability. Call P the probability, fixed back when the computer predicted, that you were going to one-box. Then:

One-box (CDT):  EU = $0      + $1,000,000 × P
Two-box (CDT):  EU = $1,000  + $1,000,000 × P

The $1,000,000 × P term is identical in both rows because your choice cannot move P now; that ship sailed. So two-boxing is just one-boxing plus a guaranteed extra $1,000, no matter what the computer predicted. Of course you two-box. The two camps are not doing the same math badly. They are doing different, internally correct math, because they disagree about which probability is the right one to plug in. That is the real paradox.

On the street, the arguments get heated and human. A two-boxer to a one-boxer: "You guys think that your decision, whatever you think now, is gonna change the past. That's called wishful thinking." A one-boxer holding firm: "Your little thought does not change God's mind, bro." And the cleanest distillation of the disagreement, from a defiant two-boxer: "I'm not losing $1 million. It was never in the room, man. You're gonna walk into a room and there's either money in the room or there isn't money in the room. Your question is, do you pick it up?" Veritasium's verdict is not a tiebreak. Because both assumptions are true, both camps have valid answers. Which raises the real worry: if there is no right answer, is this just a meaningless game?

What it says about free will

It is not meaningless, because the puzzle quietly forces you to take a stance on three of the biggest questions there are. The first is free will.

Notice the only way to truly win, to walk out with the full $1,001,000, is to be the kind of person the computer reads as a one-boxer and then two-box at the last second anyway. But if the predictor is good enough, that move is impossible. Push it to the limit: suppose the predictor is 100% accurate. Then there is genuinely nothing you can do between entering the room and choosing that changes what was already predicted. Your future is, in a real sense, readable in advance. Does a perfect predictor mean free will does not exist?

One of the Veritasium hosts goes there directly, and it is the most personal moment in the video: "Maybe free will doesn't exist. I come down in this point of like free will is an illusion, but our world operates in a way that is indistinguishable from free will being real, and therefore, you have to act as though it's real, as though it's 100% real." The argument then gets a hard practical edge. If free will is an illusion, you might be tempted to say a criminal is not at fault, so swap prison for gardening classes. But that changes the environment: now everyone knows you can kill someone and just do the gardening. The conclusion holds even if determinism is true: "Whether we do or don't have free will, you have to live as though it exists." The metaphysics may be open. The way you must act is not.

What it means to be rational

The second question is what it actually means to be rational, and here the video lands its sharpest point.

The two-boxer made the textbook rational choice, refusing to believe his thoughts can rewrite the past, and ended up with $1,000 while the one-boxers shop for private islands. The needling name for this is the Why Ain'cha Rich? argument: if you are so smart, then why ain'cha rich? If winning means money, one-boxers win, every time.

The two-boxers have a real answer, not a dodge. In their 1978 paper, philosophers Allan Gibbard and William Harper argue the rational choice is both boxes, while openly admitting two-boxers fare worse. Their move: the game is rigged. "If someone is very good at predicting behavior and rewards predicted irrationality richly, then irrationality will be richly rewarded." You are being punished for being rational, which does not make rationality wrong. Veritasium calls this a bit of a cop out, and offers the more interesting reading: Newcomb's paradox shows that sometimes, in order to be a rational person, you must act irrationally.

This is the crux, and the video sharpens it with a distinction borrowed from the prisoner's dilemma: there is a difference between a rational person and a rational act. Most of the time a rational person does rational acts. Sometimes they come apart.

In the prisoner's dilemma, two players each choose to cooperate or defect:

• Both cooperate: three coins each.
• You defect, they cooperate: you get five, they get nothing.
• Both defect: one coin each.

No matter what your opponent does, you are better off defecting, exactly the dominance logic of two-boxing. So defection is the rational act. But play the game repeatedly, as you do across a life or a society, and it flips: now you are better off cooperating. Hence the split-level insight: a rational society is full of cooperators, while a rational person, in the single shot, is a defector. You might expect a rational society to be made of rational people, but rationality at one level is not compatible with rationality at the other. Back to the boxes: the rational act is to two-box, but a rational society would be full of one-boxers.

The three ways to flip a two-boxer

The self-described "fervent two-boxer" then names the only three conditions that would convert him to one-boxing. They are worth keeping because each isolates exactly what the standard puzzle denies.

1. Backward causation. If your choice now could actually change the past. Suppose the computer predicts by opening a tiny wormhole to peek at the future. Then choosing one box literally causes the million to have been placed there. He one-boxes.
2. Multiple trials. If the game repeats. Every round, your choices build a reputation; one-box consistently and you get predicted as a one-boxer, so the million arrives this round or a later one.
3. Precommitment. If you can talk to the computer and make your case before it predicts. Then he 100% one-boxes, because staying true to his word matters to him, and the computer would know that. As another puts it, "If I put my word on it, I'll take the one box."

That third lever, precommitment, is where the video turns from a thought experiment into something with stakes.

Mutually assured destruction

The same structure, a precommitment to a worse option that produces the better outcome, runs the entire logic of the Cold War.

On 29 August 1949, the Soviet Union detonated the RDS-1 bomb in its first nuclear test, igniting a furious arms race. By the mid-1960s the US held over 30,000 warheads and the USSR just over 6,000, each side more than able to destroy the other. US Secretary of Defense Robert McNamara did not push for disarmament. He recommended a strategy of assured destruction: deter a deliberate attack by maintaining a highly reliable ability to inflict an unacceptable degree of damage upon any single aggressor. This became mutually assured destruction, MAD. Attack first and the other side surely retaliates, ending in total annihilation. The commitment to retaliate is what stops the attack in the first place.

Then the video puts you in the chair. You are the US president. You have publicly committed to retaliate if attacked. Word arrives that the Soviets have launched. It is not a system error; it is real. Launch now and at best everyone in the US and USSR dies, at worst a nuclear winter kills nearly everyone on Earth. Do you push the button? The honest answer, from nearly everyone asked: "Everyone in the whole world dies, then I probably don't launch." But you precommitted. "Yep. I don't like the outcome of everyone on Earth dying, so I'm gonna just not."

There is the Newcomb structure exactly. The rational person, in the moment, does not launch, because launching helps no one now. But a country full of leaders known to flinch invites the first strike. So which leader do you elect? The chilling answer: "You want someone who maintains the posture of always pushing that button, and then you want someone who secretly will not actually push that button." The risk is obvious: if anyone finds out, the deterrent collapses.

The pure form is the game of chicken. Two cars speed at each other. The worst outcome is neither swerving, you both die; you win if the other swerves and you do not. The dominant strategy is brutal and visible: rip your steering wheel out and throw it out the window where your opponent can see it. Now you cannot swerve, you are a mad dog headed straight, and the opponent's best move becomes swerving. Making yourself unable to back down is what wins.

Cinema nailed this in 1964's Dr. Strangelove. The Russians build a perfect doomsday device: detect a nuclear attack or any tampering and it automatically triggers an explosion large enough to kill everyone on the planet. The tamper switch is not to stop enemies disabling it, it is to stop the Russians themselves from having second thoughts, removing their own ability to flinch. The whole point is to be so automatic and devastating that the US would never even consider attacking. And it only works if everyone knows the device exists, which, in the film's bitter joke, is the one thing the Russians forgot to announce.

Precommitment is the ultimate strategy

In both Newcomb's paradox and MAD, the best outcome flows from a precommitment to a worse option. That is what secures at least $1 million in the one case and a tense but stable peace in the other. The commitment is the thing. So maybe being rational is not about choosing well in the moment. It is about choosing well which rules you will live by.

The video gives this its most philosophical voice: the question is not how to act, it is what rules one ought to follow, and how one even decides which rules to follow. Imagine you were a robot who could rewire your own programming to obey one set of rules rather than another. What rules would you wire in? You would make yourself the kind of creature that always acts in line with the commitments it would have been good to form, even before you knew the specific problem. Then, dropped into a Newcomb case, you think: if I could have precommitted, the good precommitment would have been to be a one-boxer, and since I already wired myself to live up to the commitments I would have made, I am in effect already committed to one-boxing, even though I did not realize it.

This reframes the one-shot puzzle as an iterated problem in life. Treat it not as a single isolated case but as one instance among every future predictor, every future case, the steady building of your own reputation. You always want to live up to the commitments you have made, so even facing a brand new dilemma you stick to those ideal precommitments and act as the best version of yourself. That is what flips a committed two-boxer, on camera, to one-boxing, prompting the wry aside that it is rare to convince anyone to switch on the Newcomb problem.

The closing logic ties straight back to the prisoner's dilemma. Even if you never meet another generous supercomputer, life does not end when you leave the room. You should always defect in a one-shot prisoner's dilemma, because betrayal only gains. But across the many rounds of life and society, it pays to cooperate. So being the kind of person who sticks to an ideal precommitment is simply beneficial. Maybe you were a one-boxer all along; it just took a reframe and a new perspective.

And the single cleanest statement of the whole thing comes near the end: the core of Newcomb's paradox is deciding whether a strong correlation that you know is not causal should still matter in your decision. Your one-boxing does not cause the million; it correlates with it through the predictor. The entire fight is whether a correlation you cannot trace to a cause deserves a seat at the table when you choose. That is not a parlor question. Teasing causation from correlation is exactly how you tell whether a drug really works or whether the benefit is random chance, a problem that reaches far beyond thought experiments.

Key takeaways

• Both answers are defensible, and that is the point. One-boxing and two-boxing each follow from a true premise. The puzzle does not have a hidden trick that breaks the tie; it exposes a real fork in how rational people reason.
• The split is a split between decision theories. Evidential decision theory treats your choice as evidence about the world and one-boxes. Causal decision theory only counts what your choice can cause and two-boxes. Same facts, different probability plugged into the same expected utility formula.
• The math favors one-boxing past a coin flip. Break-even predictor accuracy is C = 0.5005. Any predictor better than random makes one-boxing's expected payout higher.
• Dominance favors two-boxing in every fixed world. With the boxes already set, two-boxing nets exactly $1,000 more than one-boxing no matter what is inside.
• It cracks open free will. A sufficiently accurate predictor leaves nothing you can do to deviate from the prediction. The practical takeaway holds either way: you must live as if free will is real, even if it is an illusion.
• Rational person and rational act can diverge. As in the prisoner's dilemma, the rational single act (defect, two-box) is not what a rational society would be built from (cooperators, one-boxers).
• Precommitment is the unifying strategy. Newcomb, mutually assured destruction, chicken, and Dr. Strangelove all reward binding yourself in advance to a choice that looks worse in the moment. Throw the steering wheel out the window.
• It is really about correlation versus causation. One-boxing correlates with the million without causing it. Whether a non-causal correlation should sway your decision is the live question, and it is the same question behind every drug trial.

Chapters

Timestamps are clickable. Click one and the player jumps there and keeps playing while you read.

• 0:00 What is Newcomb's Paradox?
• 3:24 Pick 1 Box!
• 5:24 Pick Both!
• 6:38 What is decision theory?
• 11:27 What does Newcomb's Paradox say about free will?
• 13:25 What does it mean to be rational?
• 16:49 Mutually Assured Destruction
• 20:02 Precommitment Is The Ultimate Strategy

Notable quotes

To almost everyone, it is perfectly clear and obvious what should be done. The difficulty is that these people seem to divide almost evenly on the problem, with large numbers thinking that the opposite half is just being silly. Robert Nozick, quoted at 2:01

Here's my $1 million. Take that, Casper. Thank you very much. Gregor Cavlovic, 3:13

I'm always better off by picking both boxes. This is known as strategic dominance. Casper Mebius, 3:48

I can also see that just having those thoughts in your brain are what might allow the computer to give you nothing. I think I'm grateful that I just don't have those thoughts. Henry van Dyck, 3:55

Your little thought does not change God's mind, bro. Henry van Dyck, 6:44

Free will is an illusion, but our world operates in a way that is indistinguishable from free will being real, and therefore, you have to act as though it's real, as though it's 100% real. Derek Muller, 11:47

If you're so smart, then why ain'cha rich? Casper Mebius, 13:39

Sometimes in order to be a rational person, you must act irrationally. Casper Mebius, 14:25

The best strategy in this game is to visibly take the steering wheel out of your car and throw it out the window so that the opponent can see that you've done that. Derek Muller, 19:00

Maybe being rational isn't deciding about what to choose in the moment, but it's about deciding what rules you're going to live by. Casper Mebius, 20:30

The core of Newcomb's paradox is deciding if a strong correlation that you know isn't causal should matter in your decision. Derek Muller, 24:10

Resources mentioned

• Newcomb's paradox, invented by physicist William Newcomb and popularized by Robert Nozick in 1969.
• Robert Nozick, the philosopher whose framing of the even split made the problem famous.
• The Guardian's 2016 poll of over 31,000 readers: 53.5% one-boxers, 46.5% two-boxers.
• Evidential decision theory and causal decision theory, the two frameworks that divide the camps.
• Allan Gibbard and William Harper's 1978 paper arguing for two-boxing while conceding two-boxers fare worse.
• The Why Ain'cha Rich? objection to two-boxing.
• The prisoner's dilemma and the game of chicken, the game theory parallels.
• Mutually assured destruction, Robert McNamara, and the RDS-1 Soviet nuclear test of 1949.
• Dr. Strangelove (1964) and its doomsday device.
• The video credits Dr. Arif Ahmed, Dr. Adam Elga, Dr. Kenny Easwaran, Dr. Peter Slezak, Dr. David Wolpert, Dr. Scott Aaronson, and Dr. Michael Huemer for their expertise. The causal expected utility calculation is based on a post by Michael Huemer. Full references at ve42.co/NewcombRefs.
• Veritasium, Derek Muller's science channel, and the episode's sponsor Brilliant, whose Bayesian probability course covers the drug-trial version of correlation versus causation.

The one idea to walk away with

Newcomb's paradox is not asking you to find a clever escape. It is asking you to notice that two reasonable, well defined ways of reasoning, evidence and causation, can pull a perfectly smart person in opposite directions with total confidence. The trap is thinking your side is obvious and the other side is silly. The way out is to stop deciding the single act and start deciding the rule. Be the kind of person who keeps the precommitments worth keeping, and you tend to find that the money was in the box, the peace held, and the steering wheel was already out the window. Live the iterated game, not the one shot, and the paradox dissolves into a way to live.