No. Youre missing the UX forest for the pedantry trees here. Ive worked on a team that did similar change detection with little to no ML magic. It matters how its presented as a hint (“top five suggested”) and not THE ANSWER. In addition its VERY common to do things like present confidence or weight to the user. And why theres a huge need for explainability.
And this is just part of the diagnosis process. The system should still be providing breadcrumbs or short cuts for the user to test the suggested hypothesis.
Which is why any responsible system like this will include feedback loops and evaluation of false positive/negative outcomes and tune for sensitivity & specificity over time.