[selridge]

The issue is that those hypothetical scenarios do not have to look like how patients actually interact with the tool.

Real life use is full of ill posed questions open ended statements inaccurate assessment of symptoms, and conclusory remarks sprinkled in between. Real use of chat bots for Health by non-clinicians looks very different than scenario based evaluation.