The four-card puzzle most people get wrong

In 1966, British psychologist Peter Wason described a puzzle that became one of the most studied tasks in the psychology of reasoning. He wrote it up in detail in a 1968 paper called Reasoning about a Rule.

In a common version, four cards lie on a table showing A, D, 4 and 7. Each card has a letter on one side and a number on the other. The rule is: if a card has a vowel on one side, it has an even number on the other. Which cards must you turn over to test whether the rule is true?

The answer is A and 7. The A could have an odd number on the back, which would break the rule. The 7 could have a vowel on the back, which would also break it. The 4 does not matter: whatever is behind it, the rule survives. Yet most people choose A and 4, or A alone. Only a small minority, often reported as under one in ten, get it right.

Later studies found that people do much better when the same logic comes in a familiar setting. In 1982, Richard Griggs and James Cox framed it as checking who is drinking beer in a bar against the legal age, and about three in four people solved it.

Wason saw the errors as a sign that people look for evidence that fits a rule, not evidence that could break it. Other researchers argue that people simply pick the cards the rule mentions. Either way, the result is the same: people rarely go looking for the case that would prove them wrong.

This matters every time you ask AI to check your thinking. If you ask "Find evidence that my plan will work", a chatbot will usually oblige. It is very good at finding or writing support for whatever you ask about. The search for disproof has to come from you.

The useful question is the 7 card: what result would show I am wrong, and has anyone looked for it?

The idea behind this caseLesson: Check what AI tells you →

1. 1Ask AI "What evidence would show this is wrong?" before you ask what supports it.
2. 2Ask it to argue the opposite case as strongly as it can, then judge both sides yourself.
3. 3When you test an idea, list the results that would make you change your mind, and look for those first.
4. 4Be wary when AI only gives reasons you already agree with. Ask what it left out.