← Home
Choiceology · March 25, 2024 · 30m
It's Not What It Looks Like: Confusing Correlation with Causation
Why we instinctively infer causation from correlation — and the costly real-world consequences. Explores how spurious correlations fool intuition, why randomized experiments are essential, and practical tests for distinguishing causation from coincidence.
Highlights
•
Humans are wired to infer causation from correlation — the brain is a pattern-completion machine
Milkman: the brain automatically constructs causal narratives from correlated events. If A and B co-occur, the brain assumes A caused B (or vice versa), even when both are caused by an unobserved third variable or are entirely coincidental.•
Randomized controlled trials are the gold standard because they isolate causation from correlation
Milkman: the only reliable way to distinguish causation from correlation is the randomized controlled trial (RCT), which randomly assigns people to conditions so that all confounding variables are equally distributed. Without randomization, observational data can mislead indefinitely.