Pascal's Wager is S-Tier


At the risk of just repeating this entire post, Pascal’s Wager is actually a pretty strong, tricky-to-escape argument. For those unfamiliar, the simplified version of Pascal’s Wager is: if hell is real, going there would be overwhelmingly bad, so even if you think it almost certainly isn’t, you should do what it takes, such as being religious, to avoid the possibility of winding up there. I am an atheist, insofar as that means that I only have a minority credence in the class of worldviews that might plausibly be labeled “religions” 1, but when I see things like Drew McCoy ranking Pascal’s Wager in “F-Tier” among religious arguments, I cringe.

Here’s the thing: most religious arguments have a serious problem. Things like the ontological argument, the teleological argument, and the Kalam cosmological argument appeal to high-level metaphysical properties of the universe. Not only is each of them found in vastly more and less compelling forms, and sort of hard to intuitively grasp in their strongest forms, they don’t really get you that far. Your average defender of Kalam would probably be able to defend the argument’s indication of something as specific as monotheism (even this seems too far to me, but I’ll allow that it is related to the terms of the argument at least), but from there going even as far as Christianity or Judaism or Islam or anything else isn’t plausible, much less more specific sects.

In order to get to any specific religious view, even a devotee who has some reason they think is compelling for each step taken towards their view will have to acknowledge the number of steps they must take. This comic put it well, and for us atheists it is rather noticeable that devotees of things like the “Southern Methodist Episcopalian Wesleyan Church” do not say “having considered many steps and branches in the relevant religious arguments, I have come to have a 0.001% plurality credence in the teachings of SMEWC, versus competing religious views”. 2

Pascal’s Wager is different in three very important ways. For one thing, as McCoy admits, it is catchy. Most people have some intuitive grasp of expected value theory or at least risk/reward reasoning, even if its full implications are more tricky, and this is all you really need to understand the basic logic of this argument. For another thing, Pascal’s Wager, unlike these other arguments, hammers home that you have a reason to care a whole lot about which specific view is right. Saying you’re some sort of agnostic with certain metaphysical views that a number of religions share and you have some minor preferences about the specifics is not enough to satisfy Pascal’s Wager. You need to be something like an actual religious devotee, to pick a side. Finally, and most importantly to this post, Pascal’s Wager is an argument about instrumental rationality, so it implicitly swallows all arguments about epistemic rationality that are relevant to it. That is, in order to follow the argument’s dictates properly, you need to get a range of credences in different religious views somehow, as well as a sense of the stakes involved in each, and then pick a side. 3

McCoy treats the wide variety of different possible religions that Pascal’s Wager has to consider as a definitive argument against Pascal’s Wager. As Amanda Askell points out in argument 1, this is not a great objection. I would like to go further and say that it is Pascal’s Wager’s secret superpower. On Pascal’s Wager, it is not enough to endorse some vague, deist conception of a divine being with certain properties relative to the fabric of reality, you have to contend with all of the specific religions that religious apologists themselves follow. Furthermore, it may be possible to try making some progress here.

A note in advance: I am NOT a religious scholar. What follows is mostly based on a brief stint through Wikipedia to get enough material for my example reasoning.

The two things that will matter to Pascal’s Wager most are how likely a religious view is to be true, and how high the stakes in accepting/rejecting it are. High stakes seems relatively easy. If a religion requires your endorsement of it in order to give you a chance to not go to hell (and additionally offers you the chance to go to heaven if you do believe in it, though honestly I find the hell situation more intuitively desperate), hell being defined as infinite torture, then it is going to be higher stakes than religions with nothing this dramatic. Furthermore there are forms of hell that are intuitively worse than others. In particular, most conceptions of hell get to the “infinite” part of the torture by lasting forever. Since this factor is the same for many of them, it at least very much seems like the severity of torture in each moment should be the tie-breaker between them.

As an example, some of the people who went to Hades under the ancient Greek religious teachings were punished forever. It is not a conventional, modern version of hell, but if we accept that Sisyphus for instance is suffering on net each moment of the afterlife 4, and this lasts forever, then this adds up to infinite net suffering. On the other hand, other hells seem much more unpleasant. I have never attended a sermon describing the Catholic hell myself, but if the chapter of “A Portrait of the Artist as a Young Man” I referenced in my earlier post is at all accurate (and Joyce was raised a Catholic, and continued interacting with Catholic thought throughout his writings, so it seems credible this draws on experience), it seems utterly horrible in each moment, designed to maximize the torture of each sense, and enhance the nature of torture itself by ensuring you can never get more used to the suffering.

Because of the severity of the different possible hells one could end up in, likelihood might wind up being swamped as a consideration in Pascal’s Wager, but I want to look at the relative likelihood of different religions anyway. Pretty much all major religions (at least those that believe in hell) seem so out there and weirdly specific to me, that I have no intuitive way to begin to approach comparing their likelihoods of being true. In a case like this, arguably the best way to start comparing likelihood is by comparing number of adherents. There is little chance that number of adherents is anti-correlated with the likelihood of a view being correct, whereas there is reason to think the likelihood of a view being correct might be positively correlated with its number of adherents (for instance other people might see reasons that you don’t why the religion is plausible – either arguments or real religious enlightenment).

I cannot find precise statistics in perfect agreement on either, but it seems like the two biggest religions in the world are probably Sunni Islam and Catholicism. Sunni Islam is probably the bigger but it seems like pretending to believe in Islam without actually believing classically lands you in the deepest reaches of hell, so trying to believe in Sunni Islam might be counter-productive. I cannot find any such rule in the Catholic conception of hell, though I doubt you get totally let off the hook for not sincerely believing. Still, it would be super interesting if it turned out that both the risk and likelihood considerations disproportionately favor Catholicism among religious views, because Blaise Pascal was a Catholic. This would make Pascal’s Wager that rarest of religious arguments, one that might uniquely vindicate following the specific religion of its advocate.

Now, to be clear, I think this is ultimately a coincidence. I see no reason this argument couldn’t have been first proposed by a Shia Muslim or an Eastern Orthodox Christian or anyone else who believed in a very high stakes afterlife tied to endorsement of the religion. In fact, there are almost certainly religious views preferred over traditional Catholicism under Pascal’s Wager, views that are rarer. I mentioned for instance that Islam has the disadvantage relative to Christianity that it seems to be especially punitive of pretending to believe. On other views it at least seems likely that there is not much you can do to make yourself believe in the religion, and the religion is less likely to reward you for just following the motions. As Amanda Askell notes in argument 7 of her post, a religious view that only required you to take certain actions and endorse the view, not necessarily to actually believe in the religion, would be preferable.

I also mentioned that the Catholic version of hell seems much worse moment to moment than the ancient Greek one. Maximizing on this dimension would involve causing uncountably infinite suffering moment to moment for an infinite span of time. It is not clear to me what this would even mean, but it’s not the strangest thing Pascal’s Wager asks us to imagine the possibility of. The same would go for Heaven, presumably the best version of it wouldn’t just be everlasting bliss, but everlasting bliss that is uncountably infinite in each moment. Any view that has these two attributes will be docked some points for likelihood on the number of followers heuristic, but probably not docked enough to overcome their strengths. I think a religious view of this sort can also regain points by accepting most of the tenets of nearby religious views. As an example, perhaps the endorsement of a belief one does not think is epistemically rational could be explained as a variation on the classic idea of “faith”. You could believe it as part of a version of something like Catholicism, on the possibility that most Catholics are just mistaken on the subtleties of that one point.

I think there are two more serious concerns with Pascal’s Wager that remain, both of which Askell has discussed to some extent. The first is that Pascal’s Wager is just a reductio against pure expected value decision theories. I don’t really get Askell’s reply to this one, she just says,

“I don’t think the reductio worry helps people who don’t want to buy Pascal’s wager though, since it doesn’t warrant acting as if some fanaticism-avoiding decision theory were true.” (Askell)

I don’t really understand this reply, and would need it elaborated. It really does seem like if Pascal’s Wager holds up, for many this would be seen as evidence that expected value theory doesn’t work, rather than a sufficient reason to endorse belief in some religion-like view. Or at the very least evidence that we are not in fact motivated by expected value at some extremes, whether it would be prudent or not.

A related objection to this one is that the argument seems bad faith, because it appeals to a type of reasoning most people would not, or at least in practice do not, endorse in a consistent way. If Pascal’s Wager was mostly invoked by galaxy-brain philosophers like Nick Bostrom or, well, Amanda Askell, who write difficult arguments on the margins of rationality about infinite ethics and the simulation hypothesis, then it would be more believable. In fact your average Christian apologist does not look like this, but rather like a fairly normal, conservative thinker, who endorses the religion they grew up with and no other bizarre expected value edge cases. It seems ad hoc.

The other objection is that it is the type of logic that has particularly sinister effects. One of the most interesting points to come up in Askell’s interview with Julia Galef which did not come up in her original post was a specific objection to Pascal’s Mugging, a variation on Pascal’s Wager proposed by Nick Bostrom. In this version, a mugger has forgotten his gun, but approaches you with offers of a crazy amount of utility if you hand over your wallet. This claim is of course incredibly far-fetched, but the way the mugger sets the argument up, he is able to manipulate its terms almost arbitrarily until the deal is sweet enough to make the expected value of giving him the wallet carry through. Galef’s concern, or the strongest version of it, is that if you are the sort of person who responds “rationally” to Pascal’s Muggings, you are going to wind up getting mugged an awful lot.

This is something like an expected value version of Newcomb’s Problem, following the procedure gives you the right answer on a case by case basis, but reasoning in this way can allow manipulations that leave you predictably worse off. Likewise there is something toxic in the conclusions that fall out of Pascal’s Wager. I mentioned that stakes matter, the higher the stakes of a religion, the more reason to follow it. This gives an edge to religions that endorse belief in hell, and in particular promotes ones with worse and worse visions of hell. I know of some atheists, like Colin McGinn, who apparently like the idea of hell and wish it existed, but I cannot even begin to relate to that. Hell seems to me the cruelest idea ever conceived of by humans, and I am barely reassured at all by the idea that it only happens to the “right people”. I am far more sympathetic to, for instance, Galef’s sentiment in this same interview,

“I grew up in a non-religious family, but I remember playing on the playground with another 7 year old who told me about hell, and I couldn’t sleep for like a week. I couldn’t understand how other people went on with their lives, even believing this was a thing. Not even if they thought they were going to hell, but just knowing that anyone could go to hell, like, how could you just live normally?” (Galef)

Another thing I mentioned which Pascal’s Wager promotes is popular religions. The more people accept a religion, the more likely it is to be correct. This would be self-feeding in practice, growth would encourage yet more growth. Knowledge of this pressure should cause some discounting of the popularity consideration, but it still has at least some possible force (you can never be sure its number are increasing just because of people following similar Pascalian reasoning, maybe it is still because of more people being enlightened). And, since it is really hard to tell which specific religions are more or less likely otherwise, it still might wind up being one of the dominating pressures.

I am even somewhat concerned that something like this has played a real role in the proliferation of hell-supporting religions. Not in that people go through the explicit logic of Pascal’s Wager and take all of its implications seriously, but more as a sort of implicit logic. Seeing so many people believe in hell makes other people more and more scared of hell, and more and more ready to join, and endorse hell’s existence to escape its possibility themselves.

Regardless of the precise reasoning here, I have no immediate plans to convert to some optimally hell-avoiding religion myself. I find both of the reasons I have mentioned compelling, I am both not the sort of person who in fact takes reasoning like this seriously at these extremes, and I’m worried about what run-away dynamics it would commit me to if I became the sort of person who did. Still, you cannot object that Pascal’s Wager does not prove god exists, because it isn’t trying to. You cannot accuse it of presenting a false dichotomy between the religion of the arguer and atheism, because all major religious arguments are limited in how far they can immediately take you, and even if the arguer is wrong in applying Pascal’s Wager the way they do, it does not defeat the basic reasoning. You cannot accuse Pascal’s Wager of applying equally to all options, because it is vanishingly unlikely that it does. Finally, you cannot accuse Pascal’s Wager of being impossible to make progress thinking through, because there are ways to try that are better than nothing. I may not intend to follow its dictates, but I do know that putting Pascal’s Wager, possibly the most formidable religious argument ever devised, in “F-Tier”, is not on the table for me.


  1. There is of course some serious ambiguity here about what counts as a religion. As an example, I tend to side with reductionism concerning personal identity, and lean towards panpsychist theories of consciousness. I have also argued before that there is some moral reason in favor of and no factual reason against thinking of one’s identity as encompassing all consciousness that exists. This sounds trippy and even kind of religious, but there isn’t exactly a church associated with it, and there is no special factual reason in favor of it either. Its reasonability just falls out pretty straight-forwardly from reductionist views of personal identity in general. ↩︎

  2. Ed. Note: For more explanation of why this sort of thing is a problem, see “Burdensome Details”. ↩︎

  3. This is actually not enough in practice. There are some religious views on which hell is real, and you can avoid it by following certain rules, but you don’t need to believe in any specific doctrine, and others by which you also need to believe in the specific religion. Both types are worth acting on according to Pascal’s Wager, but since it may be possible to act in accordance with the rules of more than one religion, ultimately there will be a reason for formally endorsing a religion that will punish you for not endorsing it. ↩︎

  4. And no referencing Camus, smart-asses. ↩︎



Tagged: