[BadHumans]

You would expect the reviewer to do some research and even maybe, I don't know, use the app?

[dhosek]

I’m guessing reviewers have very limited time for each review. I don’t know what the internal process looks like, but I’m guessing an automatically generated screen listing the app’s permission requests, the justifications for each and other stuff? I would be surprised if there would be time for them to actually use any of the apps that they review or if they even have a phone to try apps on.

[exe34]

It sounds like they are looking for plausible excuses, not necessarily what's going on or using common-sense. This seems ripe for LLM automation.

[dhosek]

This seems exactly the kind of result I would expect from LLM automation. I would never trust any system that used LLM output without human review. Actually, I don’t think I would trust any system that used LLM output even with human review.