Would a (hypothetically) pure altruist have children (in our current situation)?

I'm curious about your personal experiences with physical pain. What is the most painful thing you've experienced and what was the duration?

I'm sympathetic to your preference in the abstract, I just think you might be surprised at how little pain you're actually willing to endure once it's happening (not a slight against you, I think people in general overestimate what degree of physical pain they can handle as a function of the stakes involved, based largely on anecdotal and second hand experience from my time in the military).

At the risk of being overly morbid, I have high confidence (>95%) that I could have you begging for death inside of an hour if that were my goal (don't worry, it's certainly not). An unfriendly AI capable of keeping you alive for eternity just to torture you would be capable of making you experience worse pain than anyone ever has in the history of our species so far. I believe you that you might sign a piece of paper to pre-commit to an eternity of torture vice simple death. I just think you'd be very very upset about that decision. Probably less than 5 minutes into it.

So D "wins" the bid, and B pays him $15 to go get the kids from their grandma's.

Shouldn't it be more something like 15+(100-15)/2$? So both win (about) the same amount of utility? Otherwise, the one who was ready to pay 100$ saved ("won") 85$ and the other won nothing (s/he was indifferent to pay or do it for 15$).

Nice post by the way. Such techniques seem useful if you trust the other will make a bid that really represents the amount s/he's ready to pay.

If doubting is/was accepted in our current society, and we wanted Archimede to doubt about his beliefs, would we have to doubt about the value doubting, or be certain about the value of doubting?

It's a joke. As Eliezer said "to get nonobvious output, you need nonobvious input", so obviously, we'd just have to find something nonobvious. :-)

I wonder if we will ever come up with something that is as nonobvious to us right now as bayesian thinking was to Archimede.

Eliezer sometime ask something that I would now like to ask him: how would the world looks like if mathematics didn't precede mathematicians? And if it did?

But why would your distribution be two deltas at 10^-4 and 10^-16 and not more continuous?

Because it's a toy example and it's easier to work out the math this way. You can get similar results with more continuous distributions, the math is simply more complicated.

I don't think I'm rationalizing an answer; I'm not even presenting an answer. I meant only to present a (very simplified) example of how such a conclusion might arise.

I'm totally willing to chalk the survey results up to scale insensitivity, but such results aren't necessarily nonsensical. It could just mean somebody started with "what credence do I assign that aliens exist and the Fermi Paradox is/isn't an illusion" and worked backwards from there, rather than pulling a number out of thin air for "chance of life developing in a single galaxy" and then exponentiating.

Since the latter method gives sharply differing results depending on whether you make up a probability a few orders of magnitude above or below 10^-11, I'm not sure working backwards is even a worse idea. At least working backwards won't give one 99.99999% credence in something merely because their brain is bad at intuitively telling apart 10^-8 and 10^-14.

Edit: I think some degree of dichotomy is plausible here. A lot of intermediate estimates are ruled out by us not seeing aliens everywhere.

Would some people be interested in answering 10 such questions and give their confidence about their answer every month? That would provide better statistics and a way to see if we're improving.

There's both PredictionBook and the Good Judgment Project as venues for this sort of thing.