base rate neglect termbias
ignoring background frequency in favor of vivid specifics
When a vivid specific case is presented, people forget the base rate of the broader population. The classic medical-test example: a test 99% accurate for a disease 1 in 10,000 people have produces ~99 false positives for every true positive — but most people, told they tested positive, assume they're sick. Engineering version: 'this code path is hit by 0.001% of requests' should change how you triage.