A new small AI sentiment study, as summarized in Becker’s: “Patients undergoing mammography preferred having their images interpreted by a radiologist before and after AI review.”
Having their imaging interpreted by AI alone was acceptable to 4.44% participants.
Having their imaging interpreted by AI after radiologist interpretation was acceptable to 71% of participants.
The finding that 4% of people were already (incorrectly) comfortable is the most interesting thing about this study. We are so early in the game here, and the current tools so unrobust, that almost all the opining about AI diagnosis is only meaningful as a snapshot for the historical record and a ward against the hindsight bias when inevitably most predictions are wrong most of the time.