Correlation versus Causation Checklist
Test a causal claim before believing it.
The Open buttons send the prompt text as currently shown below to that site only when you click. Fill in the placeholders first.
Examine the causal claim below. The claim: [CAUSAL_CLAIM]. The evidence offered: [EVIDENCE]. The data source and design (observational, experiment): [DESIGN]. Go through this checklist and mark each item supported, not supported, or unknown, with the reason: 1. Is there a real association, and how large is it (effect size and interval)? 2. Does the cause come before the effect? 3. Could the relationship run the other way? 4. Could a third factor cause both (confounder)? Name at least three candidates. 5. Could the association come from selection into the groups? 6. Is there a plausible mechanism, and a dose-response pattern? 7. Does it hold in different groups and settings? 8. Has anyone run or could anyone run an experiment, natural experiment, or before-and-after comparison with a control? Output a table, then a verdict (strong, plausible, weak, or unsupported causal evidence), and the cheapest next piece of evidence that would most reduce doubt. Rewrite the original claim in wording that matches the evidence, for example "is associated with" instead of "causes".
Fill in before use
Optional: type your details and the prompt above updates. Nothing is saved or sent.
Replace each bracketed placeholder with your own details, then paste into Any AI or another assistant.