Assuming that a bad-people-filter is necessary, what would you use to recognize them?

2 that I can think of : 1) An AI badness pattern recognizer. 2) Vetting by trusted people.

It wouldn't have to be perfect. Would it?

That's a big assuming, yes.