Going by only the data Yvain made public, and defining “experienced rationalists” as those people who have 1000 karma or more (this might be slightly different from Yvain’s sample, but it looked as if most who had that much karma were in the community for at least 2 years), and looking only at those experienced rationalists who both recorded a cryonics probability and their cryonics status, we get the following data (note that all data is given in terms of percentages—so 50 means 50% confidence (1 in 2), while 0.5 means 0.5% confidence (1 in 200)):
For those who said “No—and do not want to sign up for cryonics”, we have for the cryonics success probability estimate (and this is conditioning on no global catastrophe) (0.03,1,1) (this is (Q1,median,Q3)), with mean 0.849 and standard deviation 0.728. This group was size N = 32.
For those who said “No—still considering it”, we have (5,5,10), with mean 7.023 and standard deviation 2.633. This group was size N = 44.
For those who wanted to but for some reason hadn’t signed up yet (either not available in the area (maybe worth moving for?) or otherwise procrastinating), we have (15,25,37), with mean 32.069 and standard deviation 23.471. This group was size N = 29.
Finally, for the people who have signed up, we have (7,21.5,33), with mean 26.556 and standard deviation 22.389. This group was size N = 18.
If we put all of the “no” people together (those procrastinating, those still thinking, and those who just don’t want to), we get (2,5,15), with mean 12.059 and standard deviation 17.741. This group is size N = 105.
I’ll leave the interpretation of this data to Mitchell_Porter, since he’s the one who made the original comment. I presume he had some point to make.
(I used Excel’s population standard deviation computation to get the standard deviations. Sorry if I should have used a different computation. The sample standard deviation yielded very similar numbers.)
Thanks for the calculations… and for causing me to learn about quartiles.
Part of Yvain’s argument is that “proto-rationalists” have an average confidence in cryonics of 21%, but “experienced rationalists”, only 15%. The latter group is thereby described as “less credulous”, because the average confidence is lower, but “better at taking ideas seriously”, because more of them are actually signed up for cryonics.
Meanwhile, your analysis – if I am parsing the figures correctly! – suggests that “experienced rationalists” who don’t sign up for cryonics have an average confidence in cryonics of 12%, and “experienced rationalists” who do sign up for cryonics, an average confidence of 26%.
This breaks apart the combination of contrary traits that forms the headline of this article. We don’t see a single group of people who are simultaneously more cryo-skeptical than the LW newbies, and yet more willing to sign up for cryonics. Instead, we see two groups: one that is more cryo-skeptical and which doesn’t sign up for cryonics; and another which is less cryo-skeptical, and which does sign up for cryonics.
This breaks apart the combination of contrary traits that forms the headline of this article. We don’t see a single group of people who are simultaneously more cryo-skeptical than the LW newbies, and yet more willing to sign up for cryonics. Instead, we see two groups: one that is more cryo-skeptical and which doesn’t sign up for cryonics; and another which is less cryo-skeptical, and which does sign up for cryonics.
It seems like you should do the same quartile breakdown for the newbies, because I read Yvain’s core point as the existence of high-probability newbies who aren’t signed up as a failure to act on their beliefs.
I haven’t separated out the newbie cryocrastinators from the newbie considerers, though, and it seems that among the experienced the cryocrastinators give higher numbers than those who have signed up, which also seems relevant to a comparison.
Maybe procrastinators are trying to over-estimate it to get themselves to do it...
The probabilities are nuts though. For the whole thing to be of use,
1: you must die in a right way to get frozen soon enough and well enough. (Rather unlikely for a young person, by the way).
2: cryonics must preserve enough data.
3: no event that causes you to lose cooling
4: the revival technology must arise and become cheap enough (before you are unfrozen)
5: someone should dispose of the frozen head by revival rather than by garbage disposal or something even nastier (someone uses frozen heads as expired-copyright data).
Note that it’s the whole combined probability that matters for the decision to sign up. edit: and not just that, but compared to the alternatives—i.e. you can improve your chances by trying harder not to die, and you can use money/time for that instead of cryonics.
edit2: also, just 3 independent-ish components (freezing works, company doesn’t bust, revival available) with high ignorance get you down to 12.5%
You might be interested in reading some other breakdowns of the conditions required for cryonics to work (and their estimates of the relevant probabilities):
If we distinguish between
“experienced rationalists” who are signed up for cryonics
and
“experienced rationalists” who are not signed up for cryonics
… what is the average value of P(Cryonics) for each of these subpopulations?
Going by only the data Yvain made public, and defining “experienced rationalists” as those people who have 1000 karma or more (this might be slightly different from Yvain’s sample, but it looked as if most who had that much karma were in the community for at least 2 years), and looking only at those experienced rationalists who both recorded a cryonics probability and their cryonics status, we get the following data (note that all data is given in terms of percentages—so 50 means 50% confidence (1 in 2), while 0.5 means 0.5% confidence (1 in 200)):
For those who said “No—and do not want to sign up for cryonics”, we have for the cryonics success probability estimate (and this is conditioning on no global catastrophe) (0.03,1,1) (this is (Q1,median,Q3)), with mean 0.849 and standard deviation 0.728. This group was size N = 32.
For those who said “No—still considering it”, we have (5,5,10), with mean 7.023 and standard deviation 2.633. This group was size N = 44.
For those who wanted to but for some reason hadn’t signed up yet (either not available in the area (maybe worth moving for?) or otherwise procrastinating), we have (15,25,37), with mean 32.069 and standard deviation 23.471. This group was size N = 29.
Finally, for the people who have signed up, we have (7,21.5,33), with mean 26.556 and standard deviation 22.389. This group was size N = 18.
If we put all of the “no” people together (those procrastinating, those still thinking, and those who just don’t want to), we get (2,5,15), with mean 12.059 and standard deviation 17.741. This group is size N = 105.
I’ll leave the interpretation of this data to Mitchell_Porter, since he’s the one who made the original comment. I presume he had some point to make.
(I used Excel’s population standard deviation computation to get the standard deviations. Sorry if I should have used a different computation. The sample standard deviation yielded very similar numbers.)
Thanks for the calculations… and for causing me to learn about quartiles.
Part of Yvain’s argument is that “proto-rationalists” have an average confidence in cryonics of 21%, but “experienced rationalists”, only 15%. The latter group is thereby described as “less credulous”, because the average confidence is lower, but “better at taking ideas seriously”, because more of them are actually signed up for cryonics.
Meanwhile, your analysis – if I am parsing the figures correctly! – suggests that “experienced rationalists” who don’t sign up for cryonics have an average confidence in cryonics of 12%, and “experienced rationalists” who do sign up for cryonics, an average confidence of 26%.
This breaks apart the combination of contrary traits that forms the headline of this article. We don’t see a single group of people who are simultaneously more cryo-skeptical than the LW newbies, and yet more willing to sign up for cryonics. Instead, we see two groups: one that is more cryo-skeptical and which doesn’t sign up for cryonics; and another which is less cryo-skeptical, and which does sign up for cryonics.
It seems like you should do the same quartile breakdown for the newbies, because I read Yvain’s core point as the existence of high-probability newbies who aren’t signed up as a failure to act on their beliefs.
I haven’t separated out the newbie cryocrastinators from the newbie considerers, though, and it seems that among the experienced the cryocrastinators give higher numbers than those who have signed up, which also seems relevant to a comparison.
Maybe procrastinators are trying to over-estimate it to get themselves to do it...
The probabilities are nuts though. For the whole thing to be of use,
1: you must die in a right way to get frozen soon enough and well enough. (Rather unlikely for a young person, by the way).
2: cryonics must preserve enough data.
3: no event that causes you to lose cooling
4: the revival technology must arise and become cheap enough (before you are unfrozen)
5: someone should dispose of the frozen head by revival rather than by garbage disposal or something even nastier (someone uses frozen heads as expired-copyright data).
Note that it’s the whole combined probability that matters for the decision to sign up. edit: and not just that, but compared to the alternatives—i.e. you can improve your chances by trying harder not to die, and you can use money/time for that instead of cryonics.
edit2: also, just 3 independent-ish components (freezing works, company doesn’t bust, revival available) with high ignorance get you down to 12.5%
You might be interested in reading some other breakdowns of the conditions required for cryonics to work (and their estimates of the relevant probabilities):
Break Cryonics Down (March 2009, at Overcoming Bias)
How Likely Is Cryonics To Work? (September 2011, here at LW)
More Cryonics Probability Estimates (December 2012, also at LW)
That’s what I thought.
Having trouble reading this data. Are the numbers percentages? (i.e. is the mean for No—don’t want 0.85%?)
Yes. I should have made that clearer. I’ll edit my comment.