In almost all cases like this, there's going to be good, telling evidence besides your own mere existence. It's only when you build a subtly unrealistically large amount of information into the statement you're conditionalizing on that weirdness arises.
In this case, you have (for example) evidence of the form "I came out of a process of sexual reproduction that contraception can prevent." Even assuming with 100% certainty that I exist as part of my prior, this massively favors the no-contraception scenario. If it turned out that the coin landed heads and everyone used contraception starting in the 70's, my self-certain prior would make it extraordinarily likely that I am some sort of sui-generis swampman-type being, or something otherwise wild. Since in fact I have evidence that there is no such being, this serves as evidence that the coin landed tails.
Now of course the thought experiment can be made to explicitly stipulate that everyone definitively is born through normal sexual reproduction. So you might say I can't appeal to the epistemic possibility that I'm a swampman (or whatever) anymore, because you're asking me to conditionalize it away and thus assign probability zero to it. And then you can laboriously try to stipulate similar things for all the countless other aberrant epistemic possibilities besides swampman that would get me out of the dilemma in the above fashion.
But this is in fact extraordinarily unrealistic! I'm never going to be truly certain about empirical statements like that, and as long as I'm a little bit uncertain, the above reasoning goes through. So I don't mind biting the bullet, if you call it that, because the bullet only fires in a case (perfect knowledge of something I can never have perfect knowledge of) that I expect to be highly pathological anyway.
I think it's fair to treat SIA as a broken clock, which is correct twice a day. The case you've described in the post is where it produces the correct answer but it doesn't mean that this kind of reasoning is generally correct. And it definitely doesn't mean that there is no other way to produce the correct answer in case described in the post, except SIA.
Consider an alternative scenario. In 1970 a coin was flipped. On Tails everyone except your biological parents got a very effective contraception, while your parents can't use any contraception whatsoever, on Heads, you parents become infertile and the rest of the world extremely fertile and can't used any contraception anymore. Heads world has more people, but your in particular existence as a child of your parents isn't possible there. So the fact of your existence makes you very confident in Tails, in spite of what SIA claims.
The real rule has to be about investigation of the causal process that created you and seeing which outcomes made this process more likely. And in different settings your existence can depend on very different processes. Sometimes it's fair to treat yourself as a random person from a group of people, but there is no rule of the universe, that we know of, claiming that it always has to be the case.
In this particular case I agree with you. Here SIA indeed produces correct result.
But, this doesn't contradict 5. at all. As 5. explicitly claims that there are cases where SIA is correct among cases where SIA is wrong. Are you claiming that there no cases where SIA produces wrong answers?
Would you agree that if your 1. is wrong then your 2. is also wrong because we would be able to construct a theory not equal to SIA which uses SIA in every case where SIA produces correct result and doesn't use SIA in cases where SIA produces wrong results?
No because you can show that the things touted as wrong results are implied by clearly right results and there's no plausible broad theory of how to reason about one's existence that gets the right results other than SIA.
Just thinking things through here - in the original case, if the reference class is extended in the way you describe, then the two options would be:
Heads: 50 billion people born before 1970 and ~1000 born after 1970
Tails: 50 billion people born before 1970 and ~500 billion people born after 1970
So there, it does seem like the SSA gives the right answer, right? If you find yourself born after 1970, that's much more likely with tails. It's not that the extra people make it more likely that you exist, period - just that the extra people make the situation you find yourself in more likely in terms of the percentage of your reference class. It would basically be like saying God creates 50 billion red jackets and 1000 blue jackets on heads, or 50 billion red jackets and 500 billion blue jackets on tails. Wouldn't SSA say a blue jacket is evidence for tails?
If you modify the case like you described, then the classes would be:
Heads: 5 billion people created in 1970 and ~1000 born after 1970
Tails: 5 billion people created in 1970 and ~500 billion people born after 1970
Isn't this still a situation where the SSA gives the right conclusion? It's still 5 billion red jackets and 1000 blue jackets on heads, 5 billion red jackets and 5000 billion blue jackets on tails. It almost seems like that makes it *better* for SSA, not worse? Am I missing something?
Another pretty decent argument for SIA is that implies that Thirders are right in Sleeping Beauty (whereas SSA suggests that Halfers are right). Which is not only the plurality position among relevant experts, but also happens to be the option that most laypeople intuitively tend to pick when first confronted with the problem - so many people have strong independent reasons to accept SIA.
> Suppose that in 1970 a coin was flipped. If it came up heads then everyone in the world would have had to use extremely effective contraception. If it came up tails, they wouldn’t. Given that you exist, I submit that you should think the coin probably came up tails. But why?
Because if no one was born after 1970 in the heads scenario, then I being born in 2005 is impossible. So heads could not have been flipped. This has nothing to do with reference classes or the sum total number of people.
> But what’s the alternative explanation if you reject SIA?
That under one scenario it would be impossible for me to exist, and under the other it would not be.
> First, the case can be modified so that there was no one birthed before 1970. Perhaps the world was made in 1970 and there were no preceding generations that were born in any way.
This makes no sense. Whether or not the world was made in the same way, contraception functions identically. If you made a world in 1970 that could have no babies, and then someone was born after 1970, a contradiction has occurred. It’s logically impossible.
> For instance, imagine that in 1970 a time machine was invented and then a coin was flipped…
Truthfully I am not sure what you are trying to say here. But the same argument basically applies. If everyone had babies before 1970 then It would be 100% impossible for me to exist. So this could not have happened because I do exist. The reason this is “clearly wrong” has nothing whatsoever to do with SIA.
Overall your argument appears to be saying “X is obvious”, which is true, and then linking it to “so my hypothesis must be obvious” even though one need not accept your hypothesis to accept the predicate.
In almost all cases like this, there's going to be good, telling evidence besides your own mere existence. It's only when you build a subtly unrealistically large amount of information into the statement you're conditionalizing on that weirdness arises.
In this case, you have (for example) evidence of the form "I came out of a process of sexual reproduction that contraception can prevent." Even assuming with 100% certainty that I exist as part of my prior, this massively favors the no-contraception scenario. If it turned out that the coin landed heads and everyone used contraception starting in the 70's, my self-certain prior would make it extraordinarily likely that I am some sort of sui-generis swampman-type being, or something otherwise wild. Since in fact I have evidence that there is no such being, this serves as evidence that the coin landed tails.
Now of course the thought experiment can be made to explicitly stipulate that everyone definitively is born through normal sexual reproduction. So you might say I can't appeal to the epistemic possibility that I'm a swampman (or whatever) anymore, because you're asking me to conditionalize it away and thus assign probability zero to it. And then you can laboriously try to stipulate similar things for all the countless other aberrant epistemic possibilities besides swampman that would get me out of the dilemma in the above fashion.
But this is in fact extraordinarily unrealistic! I'm never going to be truly certain about empirical statements like that, and as long as I'm a little bit uncertain, the above reasoning goes through. So I don't mind biting the bullet, if you call it that, because the bullet only fires in a case (perfect knowledge of something I can never have perfect knowledge of) that I expect to be highly pathological anyway.
I think it's fair to treat SIA as a broken clock, which is correct twice a day. The case you've described in the post is where it produces the correct answer but it doesn't mean that this kind of reasoning is generally correct. And it definitely doesn't mean that there is no other way to produce the correct answer in case described in the post, except SIA.
Consider an alternative scenario. In 1970 a coin was flipped. On Tails everyone except your biological parents got a very effective contraception, while your parents can't use any contraception whatsoever, on Heads, you parents become infertile and the rest of the world extremely fertile and can't used any contraception anymore. Heads world has more people, but your in particular existence as a child of your parents isn't possible there. So the fact of your existence makes you very confident in Tails, in spite of what SIA claims.
The real rule has to be about investigation of the causal process that created you and seeing which outcomes made this process more likely. And in different settings your existence can depend on very different processes. Sometimes it's fair to treat yourself as a random person from a group of people, but there is no rule of the universe, that we know of, claiming that it always has to be the case.
Salutations to a fellow anthropics enthusiast! Could you address these several points?
1. The fact that SSA sometimes outputs correct answers doesn't mean that it's true in general.
2. The fact that SIA sometimes outputs correct answers doesn't mean that it is true in a general.
3. The fact that SSA sometimes outputs wrong answers doesn't mean that SIA is true in general.
4. The fact that SIA sometimes outputs wrong answers doesn't mean that SSA is true in general.
5. Both SIA and SSA have cases where they produce correct answers and wrong answers.
6. Therefore both SIA and SSA are obviously wrong in a general case.
I reject 5. My claim in this article is that there is no view other than SIA that makes sense of the verdicts in this case.
In this particular case I agree with you. Here SIA indeed produces correct result.
But, this doesn't contradict 5. at all. As 5. explicitly claims that there are cases where SIA is correct among cases where SIA is wrong. Are you claiming that there no cases where SIA produces wrong answers?
My claim is
1. SIA doesn't produce any wrong results
2. There are no alternative views that explain some right results of SIA which shows that SIA is the right view
Would you agree that if your 1. is wrong then your 2. is also wrong because we would be able to construct a theory not equal to SIA which uses SIA in every case where SIA produces correct result and doesn't use SIA in cases where SIA produces wrong results?
No because you can show that the things touted as wrong results are implied by clearly right results and there's no plausible broad theory of how to reason about one's existence that gets the right results other than SIA.
Hmm. Suppose I show you that thirdism in Sleeping Beauty is wrong. Which obviously right result implies it?
> there's no plausible broad theory of how to reason about one's existence that gets the right results other than SIA.
Well, maybe because there is no one general way to reason about one's existence in all circumstances? Why do you just discard such possibility?
Just thinking things through here - in the original case, if the reference class is extended in the way you describe, then the two options would be:
Heads: 50 billion people born before 1970 and ~1000 born after 1970
Tails: 50 billion people born before 1970 and ~500 billion people born after 1970
So there, it does seem like the SSA gives the right answer, right? If you find yourself born after 1970, that's much more likely with tails. It's not that the extra people make it more likely that you exist, period - just that the extra people make the situation you find yourself in more likely in terms of the percentage of your reference class. It would basically be like saying God creates 50 billion red jackets and 1000 blue jackets on heads, or 50 billion red jackets and 500 billion blue jackets on tails. Wouldn't SSA say a blue jacket is evidence for tails?
If you modify the case like you described, then the classes would be:
Heads: 5 billion people created in 1970 and ~1000 born after 1970
Tails: 5 billion people created in 1970 and ~500 billion people born after 1970
Isn't this still a situation where the SSA gives the right conclusion? It's still 5 billion red jackets and 1000 blue jackets on heads, 5 billion red jackets and 5000 billion blue jackets on tails. It almost seems like that makes it *better* for SSA, not worse? Am I missing something?
Another pretty decent argument for SIA is that implies that Thirders are right in Sleeping Beauty (whereas SSA suggests that Halfers are right). Which is not only the plurality position among relevant experts, but also happens to be the option that most laypeople intuitively tend to pick when first confronted with the problem - so many people have strong independent reasons to accept SIA.
And yet 1/2 is the correct answer in Sleeping Beauty. Though not for the reasons SSA claims.
https://www.lesswrong.com/posts/SjoPCwmNKtFvQ3f2J/lessons-from-failed-attempts-to-model-sleeping-beauty
https://www.lesswrong.com/posts/gwfgFwrrYnDpcF4JP/the-solution-to-sleeping-beauty
> Suppose that in 1970 a coin was flipped. If it came up heads then everyone in the world would have had to use extremely effective contraception. If it came up tails, they wouldn’t. Given that you exist, I submit that you should think the coin probably came up tails. But why?
Because if no one was born after 1970 in the heads scenario, then I being born in 2005 is impossible. So heads could not have been flipped. This has nothing to do with reference classes or the sum total number of people.
> But what’s the alternative explanation if you reject SIA?
That under one scenario it would be impossible for me to exist, and under the other it would not be.
> First, the case can be modified so that there was no one birthed before 1970. Perhaps the world was made in 1970 and there were no preceding generations that were born in any way.
This makes no sense. Whether or not the world was made in the same way, contraception functions identically. If you made a world in 1970 that could have no babies, and then someone was born after 1970, a contradiction has occurred. It’s logically impossible.
> For instance, imagine that in 1970 a time machine was invented and then a coin was flipped…
Truthfully I am not sure what you are trying to say here. But the same argument basically applies. If everyone had babies before 1970 then It would be 100% impossible for me to exist. So this could not have happened because I do exist. The reason this is “clearly wrong” has nothing whatsoever to do with SIA.
Overall your argument appears to be saying “X is obvious”, which is true, and then linking it to “so my hypothesis must be obvious” even though one need not accept your hypothesis to accept the predicate.