Skip to main content
computer-science economics psychology sociology

​
Common knowledge

​
Description

Common knowledge is the recursive epistemic condition in which every participant knows a proposition, every participant knows that every participant knows it, every participant knows that every participant knows that every participant knows it, and so on. The recursion is not decorative philosophical machinery. In coordination problems, an actor may need assurance not only that others possess the relevant fact but that they can safely expect everyone else to act on the same assurance. The structural shape is participants + public proposition + publicity mechanism + recursive closure + coordination dependence. A public announcement heard in one another’s presence differs from identical private messages because the first event also supplies evidence of reception. Each participant sees the same basis on which the others know, and sees that the others see it. One event supports the epistemic ladder without requiring an endless chain of acknowledgments. The ideal formal definition is infinitely recursive; practical systems approximate it. Public ceremonies, courtroom announcements, signed releases, synchronized clocks, and broadcasts with visible audiences differ in how strongly they warrant higher-order belief. The useful diagnostic is therefore not “can these humans literally represent infinity?” but what public evidence licenses each actor to stop worrying that the others may not know that they know? This makes common knowledge a coordination primitive, not a synonym for notoriety. “It is common knowledge that the cafeteria closes at five” often means only that the fact is widely known. The concept earns its keep when behavior depends on the recursive part: a protest, simultaneous cutover, cooperative investment, or maintained social frame fails if each actor fears being the only one to proceed.

​
Triggers

User-initiated: “Does everyone know that everyone knows?”, “Why wasn’t telling each person enough?”, “How do we make this public?”, “Can they safely coordinate on the announcement?”, or “Why do we need one more acknowledgment?” Agent-initiated: The agent sees multiple actors who possess the same information but still rationally hesitate because reception was private, unreliable, or invisible. Candidate inference: create a public event or shared artifact whose observation is itself observable, and identify how much recursive assurance the action actually requires. Situation-shape signals: Simultaneous launches and cutovers; collective action; public commitments; ceremonies; coordination games; distributed protocols over lossy links; announcements whose audience visibility matters; situations where identical facts delivered privately produce less action than one public broadcast.

​
Exclusions

  • Everyone separately knows the same fact — identical private information is mutual or shared knowledge. Common knowledge additionally requires a public basis for each participant to model the others as knowing, recursively.
  • Agreement on a value — consensus is a decision outcome produced by a protocol; common knowledge is an epistemic condition. Participants can commonly know that they disagree, or reach consensus without certainty about who knows the result.
  • Independent estimates aggregated for accuracy — wisdom-of-crowds benefits from keeping estimates independent. Common knowledge deliberately makes reception mutually observable and is useful for coordination rather than error cancellation.
  • A collectively sustained fiction — kayfabe requires common knowledge that a frame is fictional plus a protocol to act as if it were real. Common knowledge is content-neutral and does not itself require fiction, performance, or pretense.
  • Selecting which of several equally-good options is obvious — focal-point picks one equilibrium because it stands out to everyone; common knowledge is the recursive-awareness condition a focal point relies on, not the salience that selects among options. Fire common-knowledge when the load-bearing question is whether reception is recursively assured; fire focal-point when it is which salient option everyone lands on.
  • Coordination that tolerates unilateral or asynchronous action — common knowledge is most diagnostic when success depends on others acting conditionally together. If each actor can succeed independently, the infinite epistemic ladder is unnecessary.

​
Examples

See examples/ for source-grounded instances from game theory, distributed systems, psychology, and public ritual.

​
Structure

Internal structure of common-knowledge: a table of its component slots and the concepts that fill them.

​
Relationships

Relationship neighborhood of common-knowledge: a graph of the concepts it connects to and the concepts it is a part of.
  • consensus — shared decision outcome versus recursive awareness; either can occur without the other.
  • wisdom-of-crowds — preserves informational independence for aggregation, while common knowledge creates mutual observability for coordination.
  • kayfabe — uses common knowledge as substrate, then adds a known fiction and an act-as-if protocol.
  • focal-point — supplies the recursive awareness a focal point needs; focal-point adds salience-based selection among equilibria on top of that substrate.
  • network-effect — creates interdependent value that public knowledge of expected participation can unlock.
  • group-mind — ongoing ensemble responsiveness rather than a particular epistemic state.

​
Examples

Rubinstein, A. (1989). "The Electronic Mail Game: Strategic Behavior Under Almost Common Knowledge." American Economic Review 79(3): 385-391. https://www.jstor.org/stable/1806854 · economics

Rubinstein’s electronic-mail game gives two players a coordination problem whose appropriate action depends on the state of the world. One player learns that the state has changed and automated messages then acknowledge one another, but every message has a positive probability of being lost. Any finite exchange creates another level of “I know that you know” while leaving uncertainty about whether the final message arrived.The striking equilibrium result is that arbitrarily many successful messages do not reproduce the strategic effect of common knowledge: in the paper’s setup, players still choose the safe action rather than coordinate on the state-dependent risky action. The example isolates recursive closure as load-bearing. It is a deliberately discontinuous theoretical game with particular payoff and equilibrium assumptions, not a claim that one lost-message possibility prevents all useful real-world coordination.

Halpern, J. Y., and Moses, Y. (1990). "Knowledge and Common Knowledge in a Distributed Environment." Journal of the ACM 37(3): 549-587. https://doi.org/10.1145/79147.79161 · computer-science

Halpern and Moses formalize the relation between knowledge and coordinated action in distributed systems. Their coordinated-attack analysis considers parties who must act together while communicating across a channel on which messages may be lost. A message can establish knowledge of the plan; an acknowledgment can establish knowledge that the message was received; but any finite final acknowledgment may itself be lost, leaving the recursive ladder incomplete.For simultaneous coordinated action, the needed fact must become common knowledge, while unreliable communication cannot guarantee that state. The result explains why adding “just one more acknowledgment” cannot solve coordinated attack over a lossy channel. Its scope matters: weaker timing requirements, bounded uncertainty, probabilistic guarantees, and weaker knowledge notions can support useful practical protocols even though ideal simultaneous coordination remains impossible under the paper’s assumptions.

Thomas, K. A., DeScioli, P., Haque, O. S., and Pinker, S. (2014). "The Psychology of Coordination and Common Knowledge." Journal of Personality and Social Psychology 107(4): 657-676. https://doi.org/10.1037/a0037037 · psychology

Thomas, DeScioli, Haque, and Pinker experimentally varied whether information relevant to a coordination problem was private, shared through finite higher-order knowledge, or public in a way intended to establish common knowledge. Participants faced coordination choices in which the higher-return action depended on the other person choosing it too.Participants were more willing to attempt coordination in the common-knowledge conditions than in private or merely shared-knowledge conditions. The result supports the claim that people treat publicity as a distinctive coordination license rather than simply adding up explicit levels of “I know that you know.” The experiments use controlled games and vignettes, so the finding does not imply that public announcements override trust, history, payoff asymmetry, or coercion in real collective action.
Chwe analyzes public rituals, ceremonies, and mass-media events as communication structures that create common knowledge for coordination. A message received privately tells each person its content; a public event with a mutually visible audience also tells participants that others received it and that others can make the same inference. The ritual’s topology is therefore part of what it communicates.The analysis explains why organizers may pay a premium for conspicuously public occasions rather than merely reaching the same number of people separately: when adoption or collective action depends on expectations about others, audience visibility can change behavior. Chwe’s account is a game-theoretic interpretation of culture, not a claim that coordination exhausts the emotional, historical, sacred, or coercive functions of ritual.