In an AI-assisted (according to Pangram), self-declared “powerful takedown of the anthropic argument,” with the cover image an unflattering AI-generated photo of and me, argues against the anthropic argument for theism. Unsurprisingly, I don’t find any of these objections at all convincing. My summary of the argument will be pretty brief—read more here and here.
The anthropic argument for theism starts by arguing for the self-indication assumption. This is the view that your existence favors theories on which there are more people. For example, if a coin is flipped that creates one person if heads and ten people if tails, and you are created as a result of the coinflip, SIA says you ought to think at 10:1 odds that the coin came up tails. If ten times more people exist, it’s ten times likelier that *you *would exist.
From the SIA, it follows that you should think the number of people that exists is some very large infinity. If your existence is 10x likelier if there are 10 times more people and 100 times likelier if there are 100x more people, then your existence is infinitely likelier if there are infinitely more people. The same logic applies to larger infinities too—so from your existence, you should be certain that the number of people is some very large infinity. It must be equal in number to the total number of possible people.
Then, once you think this, theism gets a massive boost in probability. If there is a God, then it’s very likely that some ridiculously large number of people would be created. God has no limit to how much he can create. Creating people is good. So if theism is true, then it’s pretty likely that God would make a ridiculously large multiverse.
In contrast, given that the number of possible people is (conservatively) more than the number of numbers, and a number much too large for an atheistic multiverse to have any non-trivial probability of producing it, given atheism, it is ridiculously unlikely that the number of people that would exist would be the very most there could be.
So three stages:
SIA is true.
If SIA is true, then the number of people that exists is the very most it could be.
The number of people existing being the very most people it could be is vastly likelier given theism than atheism.
A fact is evidence for a hypothesis if it’s a lot likelier given the hypothesis than its negation. Thus, your existence is strong evidence for theism. Tettu has seven objections. Let’s see if they hold up.
This section has a lot of different somewhat disjoint thoughts.
Let us begin by granting SIA, although SIA itself is controversial.
It’s controversial for bad reasons.
Applying SIA across cosmological or metaphysical theories presupposes some sufficiently well-defined hypothesis space, probability distribution, reference class, and account of the physical or metaphysical constraints governing observer production.
It doesn’t require any of these. SIA is notable for, unlike SSA, not needing a reference class (this is an advantage of the theory, widely recognized). What matters, given SIA, is the number of candidates for being your present self. If you’re confused what that means, read the Carlsmith piece’s discussion of the minimal reference class.
No, you don’t need a “well-defined hypothesis space” and “probability distribution.” In no other context do we think you need this to update in the face of ridiculously strong evidence. Nobody thinks that before declaring that there’s dark matter based on powerful evidence from physics, we need some well-defined and exhaustive probability space. Nor would it be an adequate defense, after getting 27 consecutive royal flushes in poker, that the person accusing you of cheating doesn’t have a “well-defined hypothesis space.”
The argument just depends on considering two hypotheses: theism and atheism. Theism makes it likelier that there’d be the most people there could than atheism. So that fact obtaining is evidence for theism.
Now, maybe the idea is supposed to be that you only update in favor of hypotheses where there are more people if there is some precise and well-defined hypothesis space. But there’s no reason to think that. SIA is a theory about how a theory becomes more probable in light of your existence.
There’s no reason that this norm would stop applying when you don’t have super precise prior probabilities. You don’t need some precise and determinate prior in evolution to be pretty sure it happened on the basis of overwhelming evidence. In this case, the evidential update, given SIA, in favor of an infinite universe is infinitely massive! In addition, all the arguments in favor of SIA in general can also be given in favor of SIA for hypotheses where you don’t need a determinate prior.1
I also have no idea why one is supposed to need to know the “account of the physical or metaphysical constraints governing observer production.” I mean, I’m pretty sure I know the standard way people like us are produced: childbirth. If some fact is expected given one hypothesis, and the other hypothesis has to hold out hope for some ad hoc unknown way of explaining the data, it’s evidence for the first hypothesis. Again, if you didn’t grant this, then any theory could explain any evidence at no cost by declaring there to be some unknown explanation.
SIA does not itself settle these background questions. If one theory appears to allow 10^1000 observers and another only 10^50, it would be epistemically reckless simply to declare the former overwhelmingly better supported before evaluating our independent reasons for believing the laws and metaphysics that generate those populations. Otherwise SIA begins to look suspiciously like a machine for awarding probability to whichever theory writes the largest number next to “observers.”
Or, more precisely: if one theory says there are N times more observers than the other, and its prior is less than N times less than the other, then the first is more likely. So if one theory predicts 10^1000 observers and the other 10^50, the first is overwhelmingly confirmed by your existence. Of course, this result is counterintuitive, but it’s backed by about a billion decisive arguments. And on its face, it’s not that weird that if one theory predicts there being 10^950 times more people than another, it makes it 10^950 times likelier that you’d exist.
Tettu’s next point is that SIA favors theories on which there are a lot of people, not where there are a lot of people with great lives. That’s true. The inference is based on the fact that theism makes it much likelier than atheism that there’d be the right number of people.
Tettu’s argument is that theism predicts fewer people than the biggest atheistic multiverses. After all, you get more people if you create people in both good, medium, and bad states than just in good states. Thus, e.g. modal realism is supposed to be majorly favored by your existence over theism.
I don’t agree with this.
I mean, first of all, I don’t agree with the claim about the metaphysics of people. I think people are essentially souls. God can make every possible soul and then give them all great lives.
Second, theism can get just as many people as modal realism: imagine that God makes the giant modal multiverse and then after one millisecond transfers every single person to the world that’s best for them. Because every person in the modal multiverse still exists, there are just as many people.
Third, comparing infinities of the same cardinality (or both of sizes exceeding any cardinality) is very tricky. God has no limits on how much creating He can do. He can first make absolutely infinity people, then do it again, then do it again, then do it again. It’s not obvious how many people you get this way—nor is it obvious how it compares to modal realism. Both exceed any cardinality and there’s no obvious way to compare them.
Fourth, at least on standard atheistic views, there’s no limit to the number of possible people. After creating the modal realist multiverse, you can always clone it once, then again, then another time. Thus, modal realism can’t secure that every possible person gets created—it just secures that every distinct possible world gets created, perhaps once, perhaps infinite times. So there’s no meaningful sense in which modal realism creates a strictly larger set of people than theism. God can create absolute infinity worlds and clone them any number of times.
Fifth, worst case scenario, this is only a modest update in favor of modal realism. Assume there are equal good, bad, and neutral worlds with observers like us. Then the update in favor of modal realism over theism will be by a factor of 3. That’s not much. And this is more than offset by the fact that basically all the modally realist worlds with you have random counter-inductive nonsense that is incompatible with your present experience.
Sixth, if this is right, modal realism might be likelier than theism. But still theism will get a relative boost in probability from the evidence. One argument rarely settles the debate: it is successful if it boosts the likelihood of the hypothesis that it favors.
Perfect being theism doesn’t, by itself, predict the number of people existing being the most that there could be. It needs an extra assumption that:
Creating a happy person is good.
If 1, God would create a bunch of happy people.
But note: for the argument to give a giant boost to theism, these assumptions just have to be plausible. They don’t have to be certain.
These assumptions are also very likely to be true. I’ve defended the first here. The second is about as plausible as the notion that a maximally evil being would create a bunch of horribly badly off beings. If God is perfect, and can bring about infinite good lives at no cost, and doing so is good, He very likely would.
In addition, none of these affect the degree of the update. If God wouldn’t create, then theism was already false, given that we exist. The claim made by the anthropic argument is that SIA raises the likelihood of theism. “God wouldn’t create” is a fully general objection to theism, which if correct would render theism false. It thus doesn’t affect how much likelier you should think theism is in light of SIA.
The next objection is that lives as they actually exist don’t look like how you’d expect them to on theism. There are a lot of bad lives. Now, the theist shouldn’t grant that there are actually bad lives—if theism is true, then God is likely to give us an infinite afterlife rather than capriciously snuff us out after a few decades. But still, it’s surprising given theism that some lives are bad during our time on Earth.
Indeed! This is called the problem of evil (see here for my preferred response). It’s not specific to the anthropic argument. Maybe some things about our properties favor atheism and others favor theism (say, the fact that there’s psychophysical harmony, fine-tuning, psychonomic harmony, and more). If we don’t want to make every argument a free-for-all where every other argument slips in, we must hold other evidentially relevant data points in the background. In this case, we hold facts about what our lives are like in the background and just update on SIA raising our assessment of the number of people.
So you can’t rebut every theistic argument by repeating the problem of evil over and over again. This is also, by the way, why it’s silly to rebut the problem of evil by suggesting that evil requires consciousness and consciousness is evidence for theism. When considering evidence, you should consider the pieces one by one, rather than using every contrary argument as an excuse to relitigate your whole case.
The next objection is that if you think that theism only requires that God make a good life overall, He can do that no matter what life people have during their time on Earth. Thus, theism makes no predictions. Fair enough, but I don’t think this is the only thing that theism predicts. It predicts God make a best life if there is one, and if not, a very good one.
The next objection is the problem of evil. See my comments before. You don’t just get to repeat every argument you have for atheism as an objection to any argument that mentions any feature of human life. Rather, you only get to mention things relevant to that *specific *feature.
The next argument is that if you have some mechanism for generating a bunch of observers, it will predict the right number of people while also explaining evil. A few notes:
Tettu never says what this theory is supposed to be! I’ve critiqued the existing proposals
here. So far, this is just a promissory note.As an objection to the claim that SIA raises the odds of theism, this objection is impotent. Even if some atheistic hypotheses can explain the relevant data, the probability of one of those being true is super low given atheism. If there is no God, it is very unlikely that you’ll get the right number of people. If all you knew was that there was no God, you’d think it was super unlikely that
absolute infinitypeople existed. And all the leading proposals on the market for getting the right number of people accidentallyexplode induction, because they don’t secure that most observers are in inductive worlds.Now, maybe Tettu grants that SIA gives evidence for theism, but he just thinks theism is nevertheless improbable even after SIA. Fair enough! Of course, repeatedly taking large independent probabilistic hits each time new evidence is rolled out isn’t a sustainable long-term strategy—but it’s fine to do once or twice. The reason I don’t buy this is that all the proposals for getting the right number of people strike me as
ludicrously improbableand don’t adequately explain any of theother evidence in favor of theism.
Tettu then says that the argument depends on too many assumptions. You need an adequate theodicy, the idea that creating (infinitely) good lives is good, the idea that God would create a bunch of lives, and the idea that there’d be an infinite afterlife.
All these assumptions, other than that there’s a successful theodicy, strike me as very plausible. I’m more than 90% sure that they’re all true. And a theodicy is needed by any theist. It’s independent of the anthropic argument. If the atheist gets to import the problem of evil here, then I get to import other theistic argument and note that the atheist also needs an explanation of religious experience, psychophysical harmony, fine-tuning, us not being in a skeptical scenario, consciousness, etc.
Tettu has seven objections. The first one consists of errors…
