A new breakdown every week — read the newsletter →
PeopleBusinessesTools
Strategies Mental ModelsDecision Tools Business ModelsFrameworksMoats
Learn Book SummariesReading Lists GuidesQuote CollectionsLearning Paths
Artificial IntelligenceNewsletter AboutContact
Cognitive bias

Confirmation Bias

We seek and favour evidence that confirms what we already believe.

Overview

Confirmation bias is the tendency to notice, seek, and remember information that supports our existing beliefs, while overlooking or discounting information that contradicts them. It operates quietly, filtering the world so that we keep finding 'proof' we were right — which is why strong opinions can feel so justified even when they're wrong.

The model matters because it corrupts judgement from the inside. You can't simply try harder to be objective; you have to actively go looking for the evidence you'd rather not find.

When to use it

Evaluating evidence about anything you already have an opinion on.

How to apply it

Step 1

Notice your prior belief

Name what you already think about the question.

Step 2

Actively seek disconfirming evidence

Deliberately look for what would prove you wrong, not just right.

Step 3

Steelman the other side

State the opposing view in its strongest form before judging it.

Step 4

Weigh evidence by quality, not comfort

Judge information on its merits, not on whether it flatters your view.

Common pitfalls

  • Believing you're immune — everyone is subject to it, including experts.
  • Mistaking a pile of confirming evidence for proof when you never sought the counter-evidence.
  • Surrounding yourself with people who only agree.

Frequently asked questions

Can you eliminate confirmation bias?

No — but you can counter it by actively seeking disconfirming evidence and steelmanning opposing views, rather than trusting your sense of objectivity.

Why is it so dangerous?

Because it quietly manufactures confidence — you keep finding 'proof' you're right precisely because you're not looking for proof you're wrong.

Related models