Chapter 6: Norms, Surprises, and Causes

Type: Chapter 6

Category: Psychology

Difficulty: Intermediate

Reading time: 6 min

Core idea

System 1 maintains a running model of what is normal in your world — a collection of expected patterns for people, objects, situations, and sequences. When something violates a norm, System 1 registers surprise. This is the alarm system that tells System 2 to pay attention.

But System 1 does not just detect surprises — it immediately explains them. And the explanations are causal, not statistical. When we see two events in sequence, we automatically assume the first caused the second. We do not ask “what is the base rate probability of this sequence?” — we construct a story. This bias toward causal explanation over statistical thinking is one of the deepest and most consequential features of human cognition.

Why it matters

Norms are narratives, not averages

When System 1 maintains a “model of the normal,” it means something richer than statistical expectation: a structured representation of what things are like in context. A normal professor reads, wears glasses, and drinks coffee. A tall woman is not surprising; a tall woman standing next to an extremely short man is. Normativity is narrative expectation, not just statistical expectation.

Causal attribution is automatic and prior to statistics

Humans are prediction machines for causality. Present two adjacent events and people generate a causal story. Show a ball rolling into another and the second ball jumping — “the first ball caused the second to move” is irresistible, even if the sequence is artificial. This is System 1’s correct default for a world where most adjacent events are causally connected. The problem is that it extends to domains — finance, sports, social policy — where the pattern is noise.

Hot hand and cold hand: causality where there is none

The “hot hand” in basketball is the belief that a player who has made several shots is more likely to make the next one. Statistical analysis consistently shows this is false — shot sequences are essentially random. But the causal story (momentum, groove, confidence) is so compelling that even players adjust their behavior accordingly. Causal narrative suppresses the statistical signal.

Key takeaways

Key takeaways

• System 1 maintains a model of normality — violations register as surprise and flag items for System 2 attention.
• Causality is automatic and prior to statistics. Two adjacent events generate a causal story whether or not causality is present.
• Causal stories feel more satisfying and are better remembered than statistical accounts. We prefer explaining to accounting.
• The hot hand fallacy: perceived causal patterns in random sequences persist even among experts who know intellectually that the sequences are random.
• Causal attribution bias leads to over-rewarding and over-punishing based on outcomes that may be random — and to false confidence in identified causes of complex events.
• System 1 infers causality from temporal contiguity: events that happen near each other in time are automatically linked causally, even without evidence of a mechanism.

Mental model

click to zoom

System 1's norm model and causal attribution — how surprise triggers explanation and causal stories replace statistical reasoning.

Read it as: System 1 continuously compares events to its norm model. Matches pass through unnoticed. Surprises trigger System 2 and trigger causal attribution — an automatic attempt to explain the violation. The causal story that emerges feels like understanding, but it is often a narrative imposed on random variation.

Practical application

The antidote to automatic causal attribution is deliberate statistical thinking — a System 2 task. Before concluding that a manager caused great results or a strategy caused poor ones, ask: what is the base rate probability of these results by chance? How many managers or strategies produced similar results without the same attributed cause?

Causes trump statistics — the downstream problem

When a compelling causal story is available, people largely ignore statistical base rate information. A vivid anecdote about a successful entrepreneur will override statistics about startup failure rates. Chapter 16 develops this as “causes trump statistics” — one of the most consequential biases in medical diagnosis, management, and public policy.

Example

A mutual fund posts three consecutive years of above-market returns. Financial journalists profile the manager: his contrarian philosophy, his risk management, his team culture. Readers form a causal story — this manager has edge. The statistical reality: given the number of funds in existence, several will outperform for three years by chance, and the same funds regress to the mean in years four and five. The causal story explains an artifact of randomness.

Related lessons

Related

Chapter 7: A Machine for Jumping to Conclusions WYSIATI — how System 1 builds causal narratives from available information while suppressing what is absent. Chapter 16: Causes Trump Statistics The consequence of causal attribution bias — statistical base rates are ignored when a compelling causal narrative is available. Heuristics The mental shortcuts System 1 uses — causal attribution is one of the most powerful and pervasive.