Common knowledge (logic)

Common knowledge is a unique type of shared information within a group of agents. It is much stronger than simple shared information. In a group, common knowledge of a fact occurs when every agent knows the fact, every agent knows that they know it, and every agent knows that everyone else knows it, continuing infinitely.

This infinite loop of awareness is what separates common knowledge from mere mutual knowledge. While mutual knowledge involves everyone knowing a fact, common knowledge requires that the fact of everyone's knowledge is also known by everyone. This distinction is vital in fields like game theory, logic, and computer science.

To understand how common knowledge functions, consider how information is transmitted. If a truthful announcement is made publicly, the information becomes common knowledge. However, if the same information is sent to each person in a group through private messages, it only becomes mutual knowledge. Even if you tell every person privately that "everyone in the group knows the fact," it still does not reach the level of common knowledge.

To achieve common knowledge, an agent must publicly announce their knowledge. If every single agent in a group publicly announces that they know a fact, then that fact becomes common knowledge for the entire group.

A famous way to visualize this concept is through the "muddy children" or "blue-eyed islander" induction puzzles. Imagine an island where some people have blue eyes and others have green eyes. No one knows their own eye color, and there is no communication allowed. Everyone can see everyone else's eye color, but they cannot see themselves.

If an outsider makes a public, truthful announcement that "at least one person has blue eyes," the status of that fact changes. Before the announcement, the existence of blue eyes was mutual knowledge, but not common knowledge. The public announcement makes it common knowledge, which triggers a chain of logical deductions.

If there is exactly one person with blue eyes, they will see only green eyes and realize they must be the one. They will leave the island at the first dawn. If there are two blue-eyed people, they will each see one other blue-eyed person. They will wait for the first dawn to see if that person leaves. When no one leaves, they realize that the other person must have seen someone else with blue eyes. This realization leads both to leave on the second dawn.

This process follows an inductive pattern where, if there are *k* blue-eyed people, they will all leave on the *k*th dawn.

The history of this concept spans several disciplines. The philosopher David Kellogg Lewis first introduced the idea in his 1969 study, *Convention*. In that same year, sociologist Morris Friedell defined it through the lens of shared awareness. The philosopher Stephen Schiffer also developed a similar idea called "mutual knowledge" in his 1972 book, *Meaning*.

While these thinkers explored the philosophical roots, Robert Aumann provided the first mathematical formulation in 1976. He used a set-theoretical framework to define the concept, which earned him significant recognition in the field of game theory.

Mathematicians and computer scientists have since expanded these formalizations. In the 1980s, computer scientists began using epistemic logic to study common knowledge. This helps them reason about distributed systems, where multiple computers must coordinate.

In modal logic, common knowledge is often defined using a fixed-point approach. This is because a standard definition would require an infinite list of formulas, which is difficult for computers to process. By using a fixed-point definition, logicians can describe the limit of this infinite loop of knowledge within a manageable mathematical structure.

Common knowledge has profound implications for how we understand social and technical systems. In game theory, Robert Aumann's agreement theorem shows that if agents have common prior probabilities and their updated beliefs become common knowledge, they must eventually agree.

This concept also helps explain market efficiency and how traders interact. In the digital world, it allows engineers to design artificial agents that can function reliably in complex, multi-agent environments. Whether studying human language or computer networks, common knowledge provides the rules for how groups of individuals can truly act as a single, informed unit.