Biases and effectsPeter Wason, 1960

Confirmation Bias

People go looking for proof they are right and never for proof they are wrong. The first page of search results, and you are sure.

What it says

The classic study: you are shown "2, 4, 6" and asked to work out the rule that generated it. You may propose as many triples as you like and be told whether each fits.

Most people try "8, 10, 12", then "20, 22, 24". Both fit. They answer: "goes up by two."

Wrong. The rule was "any increasing sequence." "1, 2, 3" would have fit. So would "5, 100, 1000". They lost because they only tested examples that agreed with their hypothesis and never one that would break it.

Why it happens

Being told "yes, that fits" feels good. Being told "no" is a small loss. So people only ask questions they can win.

And once you think "goes up by two", incoming information rearranges itself around it. Counter-evidence looks like an exception; supporting evidence looks like "see?". Same facts, more certainty.

It has nothing to do with intelligence. If anything, clever people build better defences for their own theory.

The fix

One move: ask "if I am wrong about this, what would I expect to see?" — and go look for that.

In the number game, one try of "1, 2, 3" would have done it. Mixing in a single question designed to break your idea removes most of the trap.

When asking for opinions, show it to the person most likely to disagree first.

Where you have seen this

  • The first page of results

    Search "is X bad for you" and you get articles saying it is bad. Search "is X good for you" and you get the opposite. The answer is in the query, so every search adds certainty.

    It feels like research. It is confirmation.

  • The person you decided you dislike

    Once you decide someone is difficult, their ordinary tone sounds cold and their ordinary silence looks like a snub. The same behaviour from someone you like reads as "busy day".

    Same behaviour, different lens.

  • Reading your quiz result

    It happens with the quizzes on this very site. Reading a result, the lines that fit jump out and the ones that do not get skipped — and the verdict is "so accurate".

    Try finding three lines that are wrong about you. If you can, that is normal, and fine.

If you take one thing

"I looked into it" and "I confirmed it" are different jobs. It only counts as the first once you have gone looking for one piece of evidence against yourself.

Related

Reading