> adversaries already know what they did wrong so telling them isn't going to make much difference
I'm not sure that's true. Adversaries know they are intentionally trying to game Google's system, but that is not the same as knowing all of the internal parameters of Google's system. Telling them what they did wrong in specific cases might well give them useful additional information about those internal parameters that Google does not want to give them.
> False positives do not know what they did wrong and telling them will make a tremendous difference.
While this is true, it is also not actionable, since the whole point is that Google does not know which positives are false positives and has no cost-effective way of finding out (since finding out would require actual humans and the scale of its ad business is too large to make the number of humans that would be required affordable).
Not the number of humans it would take to replace their automated fraud detection algorithms to dramatically reduce false positives at scale, no. The whole point of their business model is that all of those processes have to be automated, otherwise they aren't profitable.
I'm not sure that's true. Adversaries know they are intentionally trying to game Google's system, but that is not the same as knowing all of the internal parameters of Google's system. Telling them what they did wrong in specific cases might well give them useful additional information about those internal parameters that Google does not want to give them.
> False positives do not know what they did wrong and telling them will make a tremendous difference.
While this is true, it is also not actionable, since the whole point is that Google does not know which positives are false positives and has no cost-effective way of finding out (since finding out would require actual humans and the scale of its ad business is too large to make the number of humans that would be required affordable).