Pinker explores common knowledge – the distinction between everyone knowing something versus everyone knowing that everyone knows it. This seemingly subtle difference drives coordination games, social relationships, and collective behavior. Language cheaply generates common knowledge that makes social coordination possible, while indirect speech and plausible deniability let us navigate situations where full transparency would be socially costly.
The book examines how recursive mentalizing ("I know that you know that I know") develops around age 8-10, with adults typically handling about 4 levels. Social phenomena like face, status hierarchies, and emotional expressions all function as common knowledge generators. Pinker connects this to cancel culture as the urge to prevent ideas from becoming common knowledge, and explains why we often don't say what we mean – the art lies in selecting the optimal level of vagueness.
Quotes
Language cheaply generates the common knowledge that makes social coordination possible.
Legal systems are not just coercive, they're expressive.
Eye contact is the ultimate common knowledge generator.
Key Takeaways
Common knowledge = everyone knows that everyone knows (infinite recursion)
Mutual knowledge = everyone knows, but not necessarily that others know
In practice, common knowledge is never 100% – more like "common 0.9 belief"
Nash equilibrium requires common knowledge of strategies
Self-signaling, self-committing speech acts
Examples: battle of the sexes, stag hunt, chicken game, hawk/dove
Theory of mind develops around 3-4yo; recursive mentalizing around 8-10yo
Adults have ~4 slots for recursive mentalizing
Solution: chunking related concepts into mental scripts with pre-packed beliefs about beliefs
We try to gain status by not caring about status
We rebel against conformity in the same way as everyone else
We show humility to prove we're better than others
We make anonymous donations to not care about getting credit
Communal sharing
Authority ranking
Equality matching (equity, fairness)
Market pricing / rational legal
Laughter: social, contagious, involuntary – shifts relational model from dominance to communality
Crying: surrender, neediness
Blushing: conspicuous shame that earns forgiveness and trust
Eye contact: the ultimate common knowledge generator
The art is selecting the optimal level of vagueness
Direct propositions are broadly interpretable, lossless (bystanders hear)
Even applies to self – viewing oneself as moral requires some ambiguity
The cancelling instinct: urge to prevent ideas from becoming common knowledge
Pluralistic ignorance: private rejection of norms everyone appears to accept
Social media enables millions to generate common knowledge simultaneously
Notes
1: Common knowledge + why it matters
Common knowledge: statement widely known to be true
Two features that make it key to understanding human life:
Instantly imparted by a conspicuous event
Can unite people in action, disrupt the social status quo
Learnings of book:
“Learning about something in public, even if everyone already knows it, can change everything”
“The main thing that common knowledge changes is the ability to coordinate”
We’re particularly aware of it since it’s so potent
This has empowered us to leverage in communities, economies, nations
Personal relationships are also common knowledge games
Common vs mutual knowledge
Eg. language is common knowledge
Fairness (eg. prisoners dilemma) vs coordination
Conventions: solution to coordination
↑ creates “non-physical realities”, not “shared fictions” (Harari)
Super Bowl ads uniquely generate common knowledge
Great for network externalities, companies reliant on brand image
Crypto speculation
Pluralistic ignorance:lack of common knowledge necessary for action
Cancel culture: intention vs perception
With social media / digital town squares, millions of people can now generate common knowledge
Punitive mobbing
Social relationships are ratified by common knowledge
Idioms for common knowledge:
The cat is out of the bag
It’s on the record
Spilled the beans
The closet
The bell can’t be un-rung
Idioms for avoiding common knowledge:
The emperor has no clothes
Look the other way
Buried their head in the sand
Elephant in the room
2: Common knowledge + common sense
Necessarily infinite (eg. general couriers)
Common knowledge is never really 100%, more like “common 0.9 belief”
Hub+spoke → distributed computing with handshake protocols
Rational agents cannot agree to disagree
Bayes’ rule + ↑: posteriors cannot be different with common knowledge, only shared knowledge
Rational argument should be a random walk (updating priors)
Rational markets: assumes common knowledge and priced in
Rational debate tips:
State confidence in probabilities
Steel man the other case
Agree to adversarial collabs
Bet on concrete predictions
3: Coordination games
Nash equilibrium
Thus, altruism > rationality
“Calls for a higher rationality in which actors submit to a social contract that ties their hands for their own good”
Self-signaling, self-committing
“Language cheaply generates the common knowledge that makes social coordination possible”
The battle Of the sexes: hypothetical scenario with 2 equilibrium, where each actor prefers the other
Need fair coordination
Hunting stag together or hare alone
The chicken game: worst if both defect
The hawk / dove game:
War when both mistakenly think they’re the hawk
The madman strategy: win or catastrophe play in hawk/dove game
“Legal systems are not just coercive, they’re expressive”
Ie. norms > law enforcement
Eg. no smoking in public spaces
4: How we get inside each other’s heads
recursive, 2) reflexive, 3) ↓
Self-evident → common knowledge
Theory of mind develops around 3-4yo
The curse of knowledge: learned knowledge from one is assumed upon other people → jargon
Recursive mentalizing around 8-10yo
The age of reason: ↑
Generally, adults have 4 slots for recursive mentalizing
Eg. “Of course I care about how you perceive I wanted you to feel”
Eg. “I know you think you understand what you thought I said, but I’m not sure you realize what you heard is not what I meant”
Humans have content-addressable memory vs computers with arbitrary address storage
Eg. only 1 slot for “X believes”
Solution: chunking of related concepts
Alt. macros, caching, memorization
↑ mental scripts with pre-packed beliefs about beliefs