I've heard a quip from my wife, "When you hear hoofbeats behind you, don't expect to see a zebra." Essentially, you need to know that booth horses and zebras make hoofbeats, but zebras are rare (in the US). It'd be foolish to look for a zebra until you've ruled out the possibility of a horse.
I agree that happens, my point is about the intersection of speciality and expected.
Aka it doesn't always happen but that is the goal.
Honestly disease diagnosis is the one area I could see AI being super helpful in which might lower this burden from extreme memorization to facilitating collecting data for analysis and being a guard on false positives.
AI has had the opposite effect on false positives. Statistically, most patients aren't (that) sick and don't have zebras.
Taking radiology as an example (because that's my specialty) ~90% of studies are normal and some types (e.g. CT for pulmonary embolism, CT for transient ischemic attack/vertigo) are closer to 98-99% normal.
Every diagnostic AI application I've seen implemented as of 2023 that merely replicates the work of a human has done nothing but increase false positives.
The extreme class imbalance makes this a non-trivial problem.
That makes more sense. I interpreted the "catch" quote to mean that doctors are expected to literally come down with any contagious disease within their specialty.
From my experience they don't. If you have any problem off the beaten path you may spend years and multiple doctors to figure out what's going on.