Vivid case detail displaces the underlying prevalence, inflating confidence in rare conclusions.
Given a description of a shy, tidy man, people judge him more likely a librarian than a farmer, ignoring that farmers vastly outnumber librarians. In medicine it produces the standard error of reading a 95%-accurate test on a rare disease as 95% likely to be right.
In practice
- A screening test 99% accurate for a condition affecting one in ten thousand returns a hundred false positives for every true one, and every patient hears "99% accurate".
- A fraud model flags an account as "highly likely" fraudulent, and the analyst forgets that almost no accounts are fraudulent.
- Further reading
- Wikipedia
- Subject
- Decision-making & Reasoning
- Essay
- The catalogue of predictable errors