Common Knowledge

Common knowledge is a concept from game theory and epistemology describing a state in which not only does everyone know something, but everyone knows that everyone knows it, and everyone knows that everyone knows that everyone knows it — and so on, infinitely. The dots are load-bearing: it is not enough for every individual to know X; what matters is that this knowing is mutually transparent at every level of recursion.

Formally, a proposition P is common knowledge in a group if: every member knows P; every member knows that every member knows P; every member knows that every member knows that every member knows P; and this holds for arbitrarily many iterations.

Why the infinity matters

In ordinary contexts, finite mutual knowledge suffices for coordination. But in fragile or contested situations — a currency run, a political uprising, a social norm under stress — the difference between ‘everyone knows’ and ‘everyone knows that everyone knows’ is decisive.

Bank run example. If every depositor privately worries that the bank is insolvent, no run occurs. If it becomes common knowledge that everyone is worried, depositors race to the door simultaneously. The cascade is not driven by new information about the bank’s balance sheet; it is driven by the move from private belief to common knowledge.

Political authority. No government enforces compliance 24/7 by force. Authority rests on the common knowledge that others will comply — she gets her way because she knows I’ll obey; I’ll obey because I know she gets her way. Revolutions happen when public demonstrations make it common knowledge that enough people are opposed; the authority of the regime evaporates not when it is defeated by force but when its common-knowledge support structure collapses.

Generating common knowledge

Because infinite recursive belief cannot literally be computed, real systems generate common knowledge through conspicuous public signals that are witnessed simultaneously.

Coral spawning is the limiting case: no beliefs, no brains, but the full moon functions as a public simultaneous signal that solves the coordination problem at the level of evolution rather than cognition.

Schelling points and common salience

A Schelling point is a closely related but distinct concept: a coordination equilibrium selected not by explicit common knowledge but by common salience — what pops into everyone’s mind simultaneously. Schelling’s example was meeting at the clock at Grand Central Station when separated in New York: no explicit agreement, but a focal point both parties could infer the other would infer.

At the Schelling point, game theory leaves off and, as Pinker puts it, ‘human whimsy, caprice, even poetry take over.’ There is no logical derivation of a Schelling point; its only qualification is that it is likely to pop into the relevant people’s minds.

Benign hypocrisy and strategic ambiguity

Common knowledge is not always desirable. Institutions sometimes survive precisely because certain things are not common knowledge — the common pretence that something is not what everyone knows it to be.

Examples (following Robin Hanson’s framework, extended by Pinker):

• The paper bag — police decline to arrest someone drinking in a park if the bottle is in a paper bag. The bag does not fool anyone, but it preserves the pretence that the rule of law applies without requiring enforcement of every infraction.
• Taiwan’s sovereignty — universally recognised in practice, officially denied by almost every government. Explicit recognition would require China to respond to a challenge to its claim; the ambiguity allows China to maintain the pretence.
• Israel’s nuclear arsenal — ‘nuclear opacity’ (ambiguity denied by Israel since the 1960s). If it were officially common knowledge, Arab states would face domestic pressure to match it. The pretence that it is not common knowledge preserves deterrence without triggering proliferation.

Preference falsification: the private-lies mechanism for the same cascade

Marc Andreessen, drawing on the economist Timur Kuran’s Private Truths, Public Lies (1995) and Vaclav Havel’s The Power of the Powerless (1978), describes the same collapse-of-consensus phenomenon from the opposite direction: not mutual awareness forming, but private lying breaking down. Kuran’s term, preference falsification, names the act of stating a belief in public that contradicts what one holds in private — whether concealing dissent or performing enthusiasm one does not feel.

Havel’s parable of the Prague greengrocer makes the mechanism concrete: a shop sign reading ‘Workers of the world, unite!’ that both the greengrocer and every passer-by know to be an empty, forced slogan. The sign enforces compliance, not belief — everyone can see through it, but no one can safely be the first to take it down. Because falsification runs in both directions across a whole population, the society loses any accurate reading of how many people actually agree: dissenters cannot tell whether they are a fringe or a silent majority, since the public signal (universal apparent assent) is manufactured rather than informative.

The unravelling, on this account, follows the same shape as a common-knowledge cascade but through a different trigger: one person tests the falsified consensus in public: if they are not punished, a second follows, and — as with a bank run or a revolution — the visible minority can flip rapidly into a visible majority once enough people discover, all at once, that they were never alone. Andreessen (echoing Eric Hoffer’s The True Believer) adds an elite/mass distinction Kuran’s framework does not itself require: a small, organised elite typically does the work of naming the heretical idea, while a much larger, less ideological middle simply follows whichever side currently looks socially safe.

Common knowledge vs preference falsification — same collapse, different mechanism. The frameworks describe the identical structural moment (a false public consensus that can invert abruptly) but locate the fragility differently. Common knowledge treats the fragile quantity as mutual awareness — everyone privately doubting, no one aware the doubt is shared. Preference falsification treats it as active, individual lying — people not merely silent but performing a belief they reject. The two are compatible rather than competing: a regime of preference falsification is, definitionally, a state in which private doubt has not yet become common knowledge; the moment it does, via a public test that survives, is exactly the trigger both frameworks describe.

Aumann’s agreement theorem

A formal consequence of common knowledge: rational agents with the same priors whose posteriors are common knowledge cannot agree to disagree. If two rational agents announce their probability estimates of a hypothesis to each other, those estimates must converge — not because they have exchanged evidence, but because the mere announcement of the estimates, under common-knowledge conditions, is itself informative.

Corollary: rational agents should not engage in speculative trade. In a public market, prices are common knowledge; any private signal that a stock is undervalued is almost certainly already reflected in the price. The case for index funds over active management rests on this corollary.

Common knowledge in social relationships

Pinker extends the framework to human relationships: friendship, authority, romance, and business are all coordination games resolved by common knowledge of the relationship type. This explains the social function of:

• Rituals and ceremonies — they generate the common knowledge that a relationship has changed status (marriage, promotion, naturalization)
• Innuendo and euphemism — allow a message to be conveyed without generating common knowledge of the message having been conveyed, thereby preserving a relationship that an explicit overture would threaten
• Tact and plausible deniability — the same mechanism; the relationship survives because neither party is committed by what was said

In the wiki

• Steven Pinker on Coordination, Common Knowledge, and the Retreat of Liberal Enlightenment — primary source; Pinker’s 2025 conversation with Tyler Cowen on his book of the same theme
• Steven Pinker — speaker page with broader bibliography
• Marc Andreessen on Preference Falsification, Institutional Power, and the AI Race — the preference-falsification account of the same cascade, via Kuran and Havel
• Marc Andreessen — speaker page
• Iron Law of Oligarchy — a related but distinct institutional-power mechanism: an organised minority’s structural control, rather than a false public consensus
• Shared Fictions — a distinct, non-game-theoretic account of large-scale coordination: collective belief in a construct with no underlying verifiable fact (money, nationhood), rather than mutual awareness of one