The Danger of Mixing Up Causality and Correlation

Ionica Smeets's TEDx talk is a short primer on correlation vs causality: one of the most common reasoning mistakes is treating correlation as if it were already proof of causation. Her method is simple and memorable. Start with a joke about ice cream causing drownings, then keep widening the pattern until the audience can see that the same mistake survives in science, parenting advice, and public debate.

Core Idea

Smeets makes three distinct causal mistakes easier to tell apart. Sometimes two things rise together because a third factor caused both, as with warm weather increasing both ice-cream sales and swimming deaths. Sometimes the causal arrow runs backward, as in the claim that marriage makes men live longer when the better explanation is that healthier, wealthier, better educated men are more likely to get married in the first place. And sometimes people spot a real correlation but still move too quickly from pattern to advice, which is how the 1999 study linking sleeping with the lights on to later short-sightedness turned into anxious parenting guidance before the genetic explanation was accounted for.

What gives the talk its force is that each example corrects a slightly different habit of mind. The point is not only "correlation does not imply causation." It is that there are multiple ways to misread a correlation, and each one demands a different correction.

The Serious Example

Her sharpest case is the self-esteem story from the 1970s. Researchers noticed that children with high self-esteem also tended to do well in school, and many adults turned that into a child-rearing lesson: build self-esteem first and performance will follow. Later work suggested the direction was mostly the reverse. Success was building confidence, not confidence producing success. Smeets uses this example to show why causal mistakes matter. A bad theory here does not just produce a bad headline; it can shape how parents and teachers intervene.

Main Lesson

Smeets ends with a practical test: when someone points to a strong association and calls it proof, ask what the mechanism is and how the causal story is supposed to work. Correlation is a clue. It is not yet an explanation.

Worth linking: this talk is broader than false-cause-fallacy. It is really about the whole boundary between association and explanation, with the false-cause fallacy as one recurring mistake inside that boundary.