When you judge what someone is “like,” you rely on resemblance to a stereotype. A description that fits a familiar type can overpower statistics.
The mind treats representativeness as probability. If a profile sounds like one category, you assign that category—even when base rates make it unlikely.
This error is not stupidity; it is the fast system doing what it does best: matching patterns. The slow system must step in to ask, “How common is this category?”
Without base rates, predictions become stories. You start with the portrait and ignore the population it was drawn from.
The correction is disciplined humility: never treat a vivid description as sufficient. Combine resemblance with frequency, and let the dull numbers have a vote at all times.