The representativeness heuristic is a mental shortcut: judging how likely something is by how well it matches a mental stereotype, rather than by the actual statistical base rates involved. Given a personality sketch describing someone as meek, tidy, and detail-oriented, most people judge them more likely to be a librarian than a farmer — even though farmers vastly outnumber librarians in the overall population, which alone should make farmer the statistically safer guess regardless of the description.
What makes this heuristic so persistent is that it isn't really a failure of intelligence — it's a failure to even notice that base-rate information exists and matters. The description feels so relevant and vivid that it crowds out the comparatively dry, abstract fact of how common each occupation actually is, and that crowding-out happens automatically, before any deliberate reasoning gets a chance to weigh in.
In the Steve experiment, participants given a sketch of someone described as meek, tidy, and detail-oriented reliably guessed 'librarian' over 'farmer,' ignoring that farmers outnumber librarians by a wide margin in the general population. Kahneman's related Tom W experiment went further: participants were given a personality sketch deliberately written to be an 'anti-base-rate' description — orderly and detail-oriented but lacking real creativity or empathy — and asked which graduate specialty Tom W most likely pursued. They confidently ranked computer science and engineering as most likely and humanities or social work as least likely, almost entirely ignoring the actual, much larger enrollment numbers in humanities and social science programs.