Newcomb’s Paradox Explained: One Box or Two?

Quick answer: In Newcomb’s paradox, a highly reliable predictor sets up two boxes: box A holds 1,000 dollars; box B holds 1,000,000 dollars only if the predictor expected you to take box B alone. You may take just box B or both boxes. Expected-value reasoning, given the predictor’s track record, says take one box. Dominance reasoning says the money is already there, so take both. The puzzle, devised by physicist William Newcomb and published by philosopher Robert Nozick in 1969, still splits experts.

Some puzzles have a right answer that people struggle to see. Newcomb’s paradox is different: careful thinkers split into two camps, and each camp is sure the other is making an obvious mistake. It is a test of what “rational choice” really means.

The Setup

  • Box A is transparent and always contains 1,000 dollars.
  • Box B is opaque. It contains either 1,000,000 dollars or nothing.
  • You may take only box B, or both boxes.
  • Yesterday, a Predictor with an almost perfect track record predicted your choice. If it predicted you would take only B, it put 1,000,000 dollars in B. If it predicted you would take both, it left B empty.
Predictor said “one box”Predictor said “two boxes”
You take only B1,000,000 dollars0 dollars
You take both1,001,000 dollars1,000 dollars

The Case for One Box

Suppose the Predictor is right 99% of the time for each kind of chooser (99% of one-boxers and 99% of two-boxers are predicted correctly). Compare expected dollar payouts (expected value):

\[ \text{One box: } 0.99 \times 1{,}000{,}000 = 990{,}000 \text{ dollars} \]

\[ \text{Two boxes: } 0.99 \times 1{,}000 + 0.01 \times 1{,}001{,}000 = 11{,}000 \text{ dollars} \]

One-boxers point out that, in practice, people who take one box walk away rich and people who take two walk away with 1,000 dollars. This reasoning treats your choice as evidence about what the Predictor did (often called evidential decision theory). Note that an overall 99% accuracy alone is not enough: a predictor that always says “one box” is 99% accurate if 99% of people one-box, yet it is wrong about every two-boxer.

The Case for Two Boxes

Two-boxers reply: the Predictor has already acted, and the money is in the boxes or it is not. Your choice now cannot change the past. Look at each column of the table: whatever is in box B, taking both boxes gives you exactly 1,000 dollars more. This is the dominance principle, and it underlies causal decision theory: choose the action that causes the best outcome, holding fixed what you cannot affect.

ApproachComparison being madeRecommendation
Evidential decision theoryWhich outcomes are likely, given that I take each action?Take box B alone
Causal decision theoryWhat does each action cause, holding fixed what is already in the boxes?Take both boxes

Why It Is Still Debated

  • It pits two respected principles, maximizing expected value and dominance, against each other.
  • It raises questions about free will and prediction: could a reliable predictor of your choice exist?
  • It matters in game theory, economics and AI design, where agents may be predicted by others, or by copies of themselves.

Martin Gardner introduced the puzzle to Scientific American readers in July 1973 and returned to it in March 1974. In the 2020 PhilPapers Survey, 39.0% of philosophers accepted or leaned toward two boxes and 31.2% toward one box; among decision theory specialists the split was much sharper, 72.9% two boxes to 20.8% one box (respondents could endorse more than one answer). The debate continues: an August 2026 technical report by statistician Radford Neal argues that the concrete way the prediction is made, for example by simulating you, changes the right answer.

Common Misconceptions

“The Predictor cheats by changing the boxes after you choose.” No. The contents are fixed before you choose; that is the whole point.

“One side just made a calculation error.” No. Each side’s arithmetic is correct for the comparison it makes; the disagreement is about which comparison should guide a rational choice, and that question remains open.

Watch: Newcomb’s Paradox in Video

If you learn better by watching, these videos are worth your time.

This Paradox Splits Smart People 50/50 (Veritasium). A thorough, entertaining look at both camps.
Newcomb’s Paradox – What Would You Choose? (Smart by Design). A short animated introduction.

Check Your Understanding

Five quick questions to see how much of this has stuck. Pick an answer and press Submit; you will see a short explanation either way.

Answer each question and press Submit to see your result right away.

1. What does box A always contain?
Found a bug? Let us know!
2. If the Predictor said “two boxes” and you take both, how much do you get?
Found a bug? Let us know!
3. With a 99% accurate Predictor, what is the expected value of taking one box?
Found a bug? Let us know!
4. Which principle supports taking both boxes?
Found a bug? Let us know!
5. Who first published Newcomb’s paradox?
Found a bug? Let us know!

Frequently Asked Questions

What is Newcomb’s paradox?

A decision puzzle in which a reliable predictor fills an opaque box with 1,000,000 dollars only if it predicts you will take that box alone; you must choose one box or both.

Should you take one box or two in Newcomb’s problem?

There is no consensus. Expected-value (evidential) reasoning favors one box; dominance (causal) reasoning favors two boxes.

Who invented Newcomb’s paradox?

Physicist William Newcomb devised it around 1960, and philosopher Robert Nozick published it in 1969.

Why does Newcomb’s paradox matter?

It tests competing theories of rational choice, which matter in economics, game theory and the design of AI systems that may be predicted by others.

Keep Exploring

Sources and Further Reading

Related to This Article

What people say about "Newcomb’s Paradox Explained: One Box or Two? - Effortless Math"?

No one replied yet.

Leave a Reply