Hacker News Reader

Comment by dragonsngoblins | Hacker News Reader

> Most accusations are real

See, I want to believe that, but I'm not sure there is a way to know whether or not that is actually true. In much the same way that there is rarely evidence of guilt when an accusation is true there isn't likely to be evidence when an accusation is false. I keep seeing people say some variant of "X% of rape accusations are false" but none of the sources I have ever seen have had a particularly accurate way of gathering data... because one doesn't exist. The proportion of unproven accusations that happen to be false is unknowable in practical terms, and I get frustrated by people variously flat out stating it is high/low.

View on HN

You may borrow priors that are personal but calibrated based on reading statistics: accusations are real at a 3/4 rate, about 1/3 of people have been sexually assaulted ever, and in 9/10 of sexual assaults it is not the assaulter's first time.

I personally would not say that 3/4 is "most", but I don't think any kind of blanket statement can occur below 7/8, and usually I prefer to think in nines or other logarithmic scales very close to certainty or uncertainty.

We need to have evidence before we can make certain reasoned decisions. Without evidence, we need to listen, consider, and keep ourselves open to possibilities.

Do you have sources for those numbers?

Thanks to Aumann's Agreement, I cannot convince you that the priors are well-calibrated priors without infinite regress. Do you have sources for other numbers that you would prefer be used instead?

I'm gonna take that as a no.

Sure. Take a step back and look at this with some perspective:

* Neither of us really know what's going on

* Both of us have done private research

* We came up with some numbers

* I shared my numbers, you didn't share yours

Your complaint is that I'm not delegating the numbers to some authority. My complaint is that, regardless of which authority you choose, you're still choosing an authority.

I'm not saying that I'm right; in fact, part of the point of choosing priors is to be wrong. However, when we publicize the process of selecting priors, we of course embarrass each other: If the priors were already well-known and well-agreed-upon, then by Aumann's Agreement, we would have nothing to discuss.

You are correct that we cannot know. I've made some assumptions I can't prove. But I think they're likely.

Reply on news.ycombinator.com