C.W.K.
Stream
Lesson 04 of 05 · published

Correlation vs causation — the trap 95% fall in

~30 min · statistics, causation, everyday-trap

Level 0Curious Beginner
0 XP0/43 lessons0/12 achievements
0/100 XP to next level100 XP to go0% complete

The most expensive confusion in modern life

Two things move together. Ice cream sales rise. Drowning deaths rise. Both happen in summer. Did ice cream cause drownings? Obviously not — both are caused by hot weather. They are correlated, not causally linked. There's a hidden third thing (the season) driving both.

Most science headlines are about correlations dressed up as causes. "People who drink coffee live longer." Maybe coffee extends life. Maybe people with steady jobs and disposable income drink coffee, and people with steady jobs live longer for other reasons. Maybe healthy people enjoy coffee because they can metabolize it well, and the cause runs the other way. The headline doesn't tell you which.

The four shapes a correlation can hide

  1. A causes B. Smoking → lung cancer. Verified through massive studies plus a known mechanism (tar damages DNA).
  2. B causes A. Reverse causation. "People who exercise are happier" — maybe happy people exercise more, not the other way around.
  3. C causes both. Confounder. Ice cream and drownings, both caused by summer.
  4. Pure coincidence. With enough variables, some will line up by chance. The number of pirates and global temperature correlate inversely. Pirates didn't cool the planet.

The 5% move

Whenever you see a study saying "X is linked to Y," mentally run the four shapes. Which one does the headline actually establish? Almost always the answer is shape 1 (proper causation) is hard, expensive, and rare — needs a randomized controlled trial. The other three shapes can be dismissed as "linked" in casual writing.

This single habit — which of the four is this? — is the difference between being persuaded by every health article you read and being able to read them with calm distance. It's also the lens behind asking "what does the AI benchmark actually measure?" A correlation between benchmark score and real-world skill isn't the same as the benchmark causing the skill.

External links

Exercise

Find a recent health, productivity, or AI article using the phrase "linked to" or "associated with." Run the four shapes. Which one does the article actually establish? You'll usually find the article doesn't say.

Progress

Progress is local-only — sign in to sync across devices.
Spotted a bug or have feedback on this page?Report an Issue

Comments 0

🔔 Reply notifications (sign in)
Sign inPlease sign in to comment.

No comments yet — be the first.