When I heard about Yvain’s PD contest, I flipped a coin. I vowed that if it came up heads, I would Paypal the winner $200 (on top of their winnings), and if it came up tails I would ask them for the prize money they won.
It came up tails. YOUR MOVE.
(No, not really. But somebody here SHOULD have made such a commitment.)
Hey, it’s not too late: if you should have made such a commitment, then the mere fact that you didn’t actually do so shouldn’t stop you now. Go ahead, flip a coin; if it comes up heads, you pay me $200; if it comes up tails, I’ll ask Yvain to give you the $42.96.
Yes, of course I thought of that when considering my reply, but in this particular context (where we’re considering counterfactual dealmaking presumably because the idea of pulling such a stunt in real life is amusing), I thought it was more in the spirit of things to be trusting. As you know, Newcomblike arguments still go through when Omega is merely a very good and very honest predictor rather than a perfect one, and my prior beliefs about reasonably-well-known Less Wrongers make me willing to bet that Simplicio probably isn’t going to lie in order to scam me out of forty-three dollars. (If it wasn’t already obvious, my offer was extended to Simplicio only and for the specified amounts only.)
Precommitment is a solved problem which doesn’t need a trusted website. For example, simplicio could’ve released a hash precommitment (made using a local hash utility like sha512sum) to Yvain after taking the survey and just now unveiled that input, if he was serious about the counterfactual.
(He would also want to replace the ‘flip a coin’ with eg. ‘total number of survey participants was odd’.)
You can even still easily do a verifiable coin flip now. For example, you could pick a commonly observable future event like a property of a Bitcoin block 24 hours from now, or you could both post a hash precommitment of a random bit, then when both are posted, each releases the chosen bit, verifies the other’s hash, and XOR the 2 bits to choose the winner.
When I heard about Yvain’s PD contest, I flipped a coin. I vowed that if it came up heads, I would Paypal the winner $200 (on top of their winnings), and if it came up tails I would ask them for the prize money they won.
It came up tails. YOUR MOVE.
(No, not really. But somebody here SHOULD have made such a commitment.)
Hey, it’s not too late: if you should have made such a commitment, then the mere fact that you didn’t actually do so shouldn’t stop you now. Go ahead, flip a coin; if it comes up heads, you pay me $200; if it comes up tails, I’ll ask Yvain to give you the $42.96.
...I don’t think this is a very wise offer to make on the Internet unless the “coin” is somewhere you can both see it.
Yes, of course I thought of that when considering my reply, but in this particular context (where we’re considering counterfactual dealmaking presumably because the idea of pulling such a stunt in real life is amusing), I thought it was more in the spirit of things to be trusting. As you know, Newcomblike arguments still go through when Omega is merely a very good and very honest predictor rather than a perfect one, and my prior beliefs about reasonably-well-known Less Wrongers make me willing to bet that Simplicio probably isn’t going to lie in order to scam me out of forty-three dollars. (If it wasn’t already obvious, my offer was extended to Simplicio only and for the specified amounts only.)
Nevermind—I thought I’d found a site that would flip a coin and save the result with a timestamp.
Why hasn’t anybody made this yet?
Precommitment is a solved problem which doesn’t need a trusted website. For example, simplicio could’ve released a hash precommitment (made using a local hash utility like
sha512sum) to Yvain after taking the survey and just now unveiled that input, if he was serious about the counterfactual.(He would also want to replace the ‘flip a coin’ with eg. ‘total number of survey participants was odd’.)
You can even still easily do a verifiable coin flip now. For example, you could pick a commonly observable future event like a property of a Bitcoin block 24 hours from now, or you could both post a hash precommitment of a random bit, then when both are posted, each releases the chosen bit, verifies the other’s hash, and XOR the 2 bits to choose the winner.
No need for Bitcoin etc; one side commits to a bit, then the other side calls heads or tails, and they win if the call was correct.
They have—they’re known as “dice rollers”, because they’re usually used for rolling dice in play-by-post RPGs.
For example.
Em, I don’t actually like those odds all that much, thanks!