I've been thinking about this post on Slate Star Codex for a while, in particular, this part:
Here's my pitch, because this seems like it would minimize frustration in that graph: we use the 51%-ish "preponderance of evidence" standard used in civil litigation. These are all trained lawyers, after all, so it seems thematically appropriate. And we've collectively decided this is appropriate for non-criminal investigations – and I would be astonished if you could allege something more serious than what's already being decided in civil courts.
That's probably also the standard I would use when hiring (and while the public sector is not precisely the same thing, it's still hiring). I could see myself saying, look, he's a really good accountant, and he's probably not a murderer, so leave him alone.
Of course, if the crime is directly related to their job responsibilities (for public office: corruption or misuse of power), I'd probably use a more actuarial calculation – a 10% chance of hiring an accountant with a history of accounting fraud is obviously insanely high.
It's worth taking a moment to think about this, I think, because there are a lot of people making very confident statements, who won't even be humiliated as a hypocrite in the future – their own standards are so unclear you can't hold them to anything. This doesn't diminish the intensity of their feelings, though.