Why Does a Rare Thing Feel Likely? Ignoring How Common It Is
By the BrainSnail editorial team. How these articles are written and checked, and how to tell us when one is wrong.
People judging how likely something is attend to how well it matches a description and ignore how common it is to begin with. That single habit produces errors in medicine, in courts and in everyday judgement.
What the error is
Judging the probability of something correctly requires combining two pieces of information, being how common that thing is in the relevant population and how well the available evidence fits it. The first is the base rate and the second is the specific evidence, and correct reasoning weighs both. The documented error is that people attend almost entirely to the second and treat the first as nearly irrelevant, so a description matching a rare category produces a confident judgement that the rare category applies. That is not a failure to know the base rate, since the effect persists when the number is stated explicitly in the problem, which is what makes it a genuine reasoning failure rather than an information problem.
Where it shows up
The error appears across professional and everyday judgement:
- •Interpreting a positive result from a test for a rare condition as meaning the condition is present
- •Judging somebody's occupation from a personality sketch while ignoring how many people hold each job
- •Treating a security alert as meaningful when the underlying event is extremely rare
- •Assessing whether an unusual event has an unusual cause
- •Evaluating forensic evidence without reference to how many people could have produced it
- •Estimating risk from vivid examples rather than from frequency
The experiment that named it
Daniel Kahneman and Amos Tversky demonstrated the effect in the 1970s with a series of problems, the best known describing a person in terms fitting a stereotype of a particular profession and asking participants to judge which of two professions they held, while stating the proportions in the sample. Participants' judgements tracked the description and were largely unaffected by the stated proportions, including when the proportions were reversed between conditions. Related work showed the same pattern in judgements by trained professionals about problems in their own field. The programme these experiments belonged to established a body of findings about systematic departures from probabilistic reasoning, and Kahneman received a Nobel Prize in economics in 2002 for the work, Tversky having died in 1996.
The related errors
Several other departures from probabilistic reasoning were documented in the same research programme and they interact. The conjunction fallacy is judging a specific detailed scenario more likely than the general case it falls inside, which cannot be true and is reported consistently. The availability heuristic estimates frequency from how readily examples come to mind, which is why vivid and reported risks are overestimated and common mundane ones are not. Anchoring lets an arbitrary number presented beforehand shift an estimate. Representativeness, which underlies base rate neglect, judges probability by resemblance to a prototype. What these share is substituting an easy question for a hard one, which is generally efficient and fails in identifiable ways.
What reduces it
Several interventions have measurable effects and one is much stronger than the others. Presenting the problem as natural frequencies, meaning counts of people rather than percentages, improves performance dramatically and consistently, with studies of doctors and of students showing large gains from nothing more than restating the same information as how many out of how many. Asking people to state how the sample was generated helps, since a base rate that plainly describes the population is harder to ignore than one that feels like background. Explicitly laying out a table of possibilities works. Simply warning people about the error does not. The practical lesson is that the fix lies in how the question is presented rather than in the reasoner.
The takeaway
Correct judgement combines how common something is with how well the evidence fits, and people attend almost entirely to the fit, even when the frequency is stated explicitly. A positive test for a rare condition is the standard case. Restating the problem as counts of people rather than percentages improves performance dramatically, while warning people about the error does not.