Common Knowledge and Mutual Knowledge

Core Idea

Mutual knowledge that p means each agent knows p; common knowledge that p means each agent knows p, each agent knows that each knows p, and so on infinitely. Formally, common knowledge is the limit of an infinite sequence of nested operators: everyone knows p, everyone knows everyone knows p, etc. Common knowledge is crucial for coordinating behavior and understanding discourse, yet is surprisingly difficult to achieve.

Explainer

You've studied knowledge and belief operators — the formal tools for reasoning about what agents know: K_i(p) means agent i knows p. Mutual knowledge now extends this to groups. If agents A and B both know that it will rain, then we have mutual knowledge that it will rain: K_A(rain) ∧ K_B(rain). This seems like enough for coordination — if both people know to bring umbrellas, they'll both bring umbrellas. But a classic puzzle shows that mutual knowledge often is not enough.

Consider the coordinated attack problem: two generals, A and B, plan to attack simultaneously at dawn. General A sends a messenger to B confirming the attack. But A cannot attack until B confirms receipt, because if the messenger is lost, A attacks alone and loses. So B sends a confirmation. But now B can't be sure A got *that* confirmation, so A must confirm the confirmation — and so on infinitely. Each round of messaging adds one layer: "I know you know," "I know you know I know," etc. No finite number of confirmation rounds ever achieves genuine coordination certainty. What the generals need is common knowledge — an infinite iteration of nested knowledge that the attack is on — and that is precisely what a finite sequence of fallible messages cannot guarantee.

Formally, common knowledge that p (written CK(p)) is defined as: everyone knows p, *and* everyone knows that everyone knows p, *and* everyone knows that everyone knows that everyone knows p, *and* so on without end. Using the knowledge operator K, if we let E(p) mean "everyone knows p," then common knowledge is E(p) ∧ E(E(p)) ∧ E(E(E(p))) ∧ ... — the infinite conjunction. This is not just philosophical abstraction: common knowledge is the epistemic condition required for genuine convention. A word means what it means, a traffic light works as it does, money has value — all because everyone knows the convention, everyone knows everyone knows it, and so on. Without that infinite-iteration structure, coordination is fragile.

Common knowledge is also surprisingly rare in practice. You and a friend may both know that a party was awkward — but do you both know that you both know? And do you both know that? A public announcement — something heard simultaneously by all parties with no private uncertainty — is one of the few mechanisms that generates genuine common knowledge instantly. This is why rituals, public ceremonies, and formal declarations have such social power: they produce common knowledge by design. The mutual-to-common knowledge gap explains a host of social phenomena, from why whispered agreements are less binding than public ones to why scientific consensus requires public publication rather than private circulation of findings.