👁️

A group of children have muddy foreheads. Each can see the others but not themselves. A visitor announces something everyone can already see, that at least one is muddy, and somehow that changes what they can work out. A known fact, said aloud, becomes actionable.

Epistemic foundations

Game theory usually assumes rationality is common knowledge: everyone is rational, everyone knows that, everyone knows that everyone knows it, without end. That infinite tower is not a technicality. The muddy-children puzzle shows a public announcement of an already-visible fact can unlock a deduction that private knowledge alone could not.

From earlier lessons

Two quick ones before we start.

Common knowledge of a fact requires...

In the beauty contest, reasoning stalls because players are unsure...

Predict, then play

Set how many children are muddy. Each sees the others but not their own forehead, and the oracle has announced that at least one is muddy. Nobody speaks; they only act (leave) once they are sure. Step through the nights.

Why did that happen?

Take a guess, then reveal the reason

For two or more children the announcement states nothing anyone could not see: everyone already sees a muddy face. What it adds is higher-order knowledge. Before, you knew at least one was muddy, and you knew the others knew it, but the chain gave out at some level. The public statement makes “at least one is muddy” common knowledge to every order at once, and that is what lets the count-down induction run. Each night that passes without anyone leaving is itself informative, ruling out one possibility, until the muddy ones are forced to conclude.

The math

Argue by induction on the number of muddy children $k$ . If $k=1$ , the one muddy child sees no other muddy face, so the announcement (“at least one”) tells it directly, and it leaves on night $1$ . Suppose $k$ muddy children would leave on night $k$ . With $k+1$ , each muddy child sees $k$ muddy others and reasons: “if I were clean, there would be $k$ muddy children, who by the inductive step leave on night $k$ .” When night $k$ passes with no departure, each muddy child concludes it is muddy too, and all $k+1$ leave on night $k+1$ .

Definition, common knowledge and no-trade

A fact is common knowledge if everyone knows it, everyone knows that everyone knows it, and so on to every order. The public announcement creates exactly this for “at least one is muddy”. Aumann’s agreement theorem uses the same idea: two people with a common prior who commonly know each other’s posteriors cannot disagree. Milgrom and Stokey (1982) drew the no-trade corollary: rational traders with common priors will not take opposite sides of a purely speculative bet, because the willingness to trade would itself reveal information that erases the disagreement.

Work it out

Why the announcement matters

Take $k=3$ . Each muddy child sees $2$ muddy foreheads. It reasons: “if only those $2$ were muddy, then by the two-child argument they would leave on night $2$ .” Night $2$ comes and goes with no one leaving. That silence tells every muddy child there must be a third muddy face, which can only be its own, so all three leave on night $3$ .

Why does the seemingly redundant announcement matter? With $k \ge 2$ , everyone already sees a muddy face, so “at least one is muddy” is not new first-order knowledge. But the induction needs the base case ( $k=1$ ) to be common knowledge to all orders, and only the public announcement supplies that. Remove it and the countdown never starts.

Key idea

Higher-order knowledge, what you know others know others know, can be decisive. A public announcement of an already-visible fact matters because it creates common knowledge to every order, and rational agents with common priors cannot simply agree to disagree.

Check yourself

Check yourselfapply

With 4 muddy children, on which night do they leave, and what does each night of silence convey?

Check yourselfanalyse

For 3 muddy children, everyone already sees a muddy face. So what does the oracle's announcement add?

Check yourselfbrings back: common knowledge

The no-trade theorem says two rational traders with a common prior will not take opposite sides of a speculative bet. Why?

Sources

Aumann, R. J. (1976). Agreeing to disagree. Annals of Statistics, 4(6), 1236-1239.
Milgrom, P., and Stokey, N. (1982). Information, trade and common knowledge. Journal of Economic Theory, 26(1), 17-27.
Rubinstein, A. (1989). The electronic mail game: strategic behavior under almost common knowledge. American Economic Review, 79(3), 385-391.