Representativeness Bias

A quiet, methodical man who loves puzzles — is he more likely a librarian or a farmer? Most answer librarian. But farmers outnumber librarians vastly, so the quiet puzzle-lover is statistically far more likely to farm. The description fit the stereotype so well that the base rate was never consulted.

This is representativeness: judging probability by resemblance. How much does this thing look like my picture of the category? It is a powerful shortcut — and it ignores the most important number in every judgment: how many of each kind there are.

The error fills the world: the company that "looks like" the last success gets funded though successes are rare; the patient who matches the textbook disease gets that diagnosis though common diseases are commoner; the profile that resembles the criminal type gets the suspicion.

The corrective question is always the same: how common is each possibility, before I saw the details? Resemblance is evidence about the match; base rates are evidence about the field. The wise judgment multiplies the two. The mind, lazy, uses only the first. Be less lazy.