Series hub: This post is one chapter in the critical thinking series. For the full map in one place, read the complete guide.
In one sentence
Correlation means two things move together; causation means one produces the other — headlines often sell the second while showing only the first.
Why this matters
Health tips, education hot takes, and productivity threads love paired charts: ice cream sales and drowning, screen time and grades, coffee and longevity. The pattern is real enough to feel like proof. Without a causal story that survives alternatives, acting on the pair can waste effort or blame the wrong lever.
This habit also shows up when you skim AI summaries of papers: models often upgrade “associated with” to “improves” or “causes.” Spotting that upgrade is a core media-literacy move for learners who read science-adjacent content on HiddenLogic.
Core ideas
Association ≠ mechanism. A study can report that A and B travel together without showing how A would change B if you intervened.
Reverse causation. Maybe B drives A (poor sleep → more late-night scrolling, not only scrolling → poor sleep).
Confounders (third factors). Wealth, age, prior skill, or school quality can push both variables. The link between A and B may be mostly that shared push.
Selection and who shows up. People who choose a method (Anki, a diet, a course) differ before the method starts. Observational “users do better” needs that caveat — as in many Anki and exam findings.
Chance and multiple looks. Enough comparisons and some pairs look striking by luck. Ask whether the link was predicted or mined.
Language tells. “Linked to,” “associated with,” “predicts” are softer than “leads to,” “boosts,” “because of.” Prefer the softer wording until design justifies more.
Practices you can try
- Three alternatives. For any causal headline, write: reverse cause, one confounder, and “coincidence / weak design.” Keep the original claim only if it beats these.
- Intervention test. Ask: “If we changed only A, would B move?” If the study never changed A, stay cautious.
- Verb audit. Highlight causal verbs in the piece; match each to methods (experiment, natural experiment, longitudinal control, cross-section).
- ThinkLens pass. Paste the causal paragraph into ThinkLens and check whether “potential issues” flags overclaim.
- Teach with a false pair. Explain ice-cream/drowning-style examples to a friend; the habit sticks better than a slogan alone.