Hacker News Reader

Comment by JoshTriplett | Hacker News Reader

Absolutely true (civil cases are "preponderance of the evidence"), but the same principle still applies, just with lower numbers. 51% likely might be enough for a verdict against someone in a civil case and a legally enforceable monetary judgement. But if you have someone 30% likely to have done something (where the baseline should be close to 0%), that's still enough that reasonable risk-averse organizations may reasonably choose to not associate with them.

A court that can render legally enforceable judgements against someone should absolutely have a higher bar of certainty than an organization that is simply excluding someone from participation in that organization.

View on HN

The problem with that is that if the accused is only 30% likely to have done something, then there may be an equal or greater likelihood that the accuser is maliciously lying (allowing for some likelihood of genuine misunderstanding).

Not necessarily. Unless the presence of the accusation already was used to get that 30% number, you also need to take into account the a priori probability of someone maliciously lying (or better yet, the particular accuser to be lying in this case) and apply Bayes's theorem.

I should have expected Simpson's Paradox here. Nevertheless, I think the possibility of witch-hunting is high.

The problem is that 30% is not a good estimate in the first place, because the organization doesn't have the same skills/resources/judgements to make an accurate assessment.

It might well have substantial uncertainty, in either direction. Suppose it's 30% plus or minus 20%, a huge amount of uncertainty. That's still absolutely a reasonable threshold someone could use as the basis for a decision.

The exact numbers here are examples, not the determining factor here, nor do most organizations express them quantitatively.

The important point is that organizations can and do have a different threshold for action. You could absolutely argue about whether an organization has the right threshold, but I don't think there's a case for a deontological requirement that every organization's threshold for taking any action at all must be the same as the threshold for conviction in a court. (Leaving aside that that would utterly invalidate freedom of association.)

Reply on news.ycombinator.com