Yup, that's basically what I'm getting at, thanks for the links!
So, I'm okay with risk compensation if people are net doing better. I don't think that "if even one person is hurt by this, that's too much" is a meaningful basis for decisions, especially when there's a risk that even one person will be hurt by not doing the thing. So at the risk of reducing people to numbers, if, say, 100 teenagers send sexts when they otherwise wouldn't have and get screenshotted, but 1,000 teenagers send sexts when they otherwise would have sent them to a non-disappearing-by-default client, and now their photos don't get copied because of social pressure / high-but-not-impossible technical barriers, that still seems like a clear win.
That's the sort of data that I think would be very interesting to inform good engineering decisions, and also pretty impossible to get.
So, I'm okay with risk compensation if people are net doing better. I don't think that "if even one person is hurt by this, that's too much" is a meaningful basis for decisions, especially when there's a risk that even one person will be hurt by not doing the thing. So at the risk of reducing people to numbers, if, say, 100 teenagers send sexts when they otherwise wouldn't have and get screenshotted, but 1,000 teenagers send sexts when they otherwise would have sent them to a non-disappearing-by-default client, and now their photos don't get copied because of social pressure / high-but-not-impossible technical barriers, that still seems like a clear win.
That's the sort of data that I think would be very interesting to inform good engineering decisions, and also pretty impossible to get.