We fuss over models being confidently wrong. Reasonable to apply the same standard to people.