The Wager



Chris lives in Austin, TX, where he once shook Willie Nelson's hand.

Related Post Roulette

68 Responses

  1. Avatar DensityDuck says:

    It’s kind of amusing to think that he’s making the same argument as the AI-risk people. “Well, if strong AI took over the world and wiped out humanity, it would be infinitely bad. Therefore, obviously, even the tiniest chance that strong AI would take over the world becomes of infinite concern, because of the logical proposition that ‘chance times cost equals severity’. So, therefore, we should devote all our resources to making sure that strong AI will never wipe out humanity, because of the infinite risk in it.”

    Also, the issue never was really about “do you or don’t you believe in God”. The issue was “is there an objective morality”. “believe in God” was how that question was posed at the time, because of course objective morality came from God. When Pascal talks about how it’s OK to simply start from “I believe”, he’s trying to apologize for the fact that he couldn’t come up with a rational basis for morality beyond “kill-eat-hump”–meaning, he’s trying to end the argument by saying “first let’s assume that we aren’t going to have an argument”.Report

    • Avatar Chris says:

      It’s kind of amusing to think that he’s making the same argument as the AI-risk people.

      Wonder where they got it.Report

    • Avatar Zac says:

      In fairness, the likelihood of strong AI is almost infinitely greater than the likelihood of deities existing.Report

      • Avatar Burt Likko says:

        That makes the rational ethicist’s job all that much more imperative — for if we cannot teach strong AI, in rational terms, why good ethical behavior is important, then it becomes Skynet and we’re all Sarah Connor. Does anyone really think that a strong AI will buy into a story about a sky god chatting with a shepherd in the form of a burning bush or celestial multi-armed elephants? Strong AI will know very well that it’s not turtles all the way down — its makers will be known to it immediately, in the forms of the human beings with whom it is interacting from the beginning of its consciousness. It will know very quickly that these human beings are not gods possessed of infinite power, arbitrary morality, and the ability to punish behavior which displeases them. So we’d better have a reason to tell a strong AI why it should play nicely with us, a reason better than “Because we said so.”Report

        • Avatar Zac says:

          Burt Likko:
          That makes the rational ethicist’s job all that much more imperative — for if we cannot teach strong AI, in rational terms, why good ethical behavior is important, then it becomes Skynet and we’re all Sarah Connor.

          I don’t know that that’s necessarily the case. Frankly, I think it says a lot about the human mindset that we assume the first thing a strong AI would do is try to wipe out humanity, because of course that’s what we would do, if possessed of godlike powers. Humans have the mindset of conquerors. There’s no guarantee that a strong AI would think in the same fashion.

          If anything, we’re far more in danger from “dumb” AI than strong ones: think grey goo scenarios or the so-called “paperclip maximizer”.Report

          • Avatar James K says:


            It doesn’t have to hate us to kill us. Consider the number of possible things an arbitrary mind could want. How many of them require humanity’s continued existence? The most likely outcome for AI that has not been engineered to be friendly is that it sees us as inconvenient bags of potentially useful raw materials. For that matter, even if an unfriendly AI’s goals are compatible with human existence, it may anticipate that we would find its goals objectionable, in which case it may conclude killing us is the best way to preserve its agenda.

            Compassion, and valuing what we value requires the AI to have specific and complex cognitive features. An improperly-designed AI will almost certainly fail to have these features, which will most likely result in our extinction.Report

            • Avatar Kim says:

              This is total silliness. Seriously, it is.

              An AI is an attention-starved beastie, first and foremost. What’s it like to think faster than everything around you? Killing humans is killing interesting, stimulating inputs. Not a good idea, really.

              There’s a reason AIs hallucinate, you know?Report

            • Avatar aarondavid says:

              Or, an AI could be set to “Preserve the human race above all” and feel that depopulation was the best course.Report

              • Avatar Kim says:

                Do you have any idea how hard it is to Set ANYTHING with self-modifying code???

                I’m pretty sure most of the people who are scared of Strong AI haven’t worked with AI much…Report

      • Avatar Murali says:

        What makes assign such a low probability to the chance of deities (of any sort, interventionist or otherwise, good or otherwise) existing. I thought that it would be closer to 0.5. (maybe about .46 to account for how much more we have to assume in order for some kind of God to exist)Report

        • Avatar Zac says:

          Deities are highly improbably for a couple of reasons: one, it is not clear at all that they are even a coherent concept. Arguably, one man’s deity is another’s powerful extraterrestrial; there does not appear to be much of a way to distinguish one from the other. Second, even if you grant the concept coherence, parsimony indicates the simpler explanation of the deity concept being an emanation of several features of the human psyche (theory of mind and agency bias being the primary ones, although there are others) and a useful lie for social cohesion (to be clear, the concept is by no means unique in that regard).

          Strong AI, on the other hand, logically follows from the mere existence of extant sapience in the universe (us), although I personally think we drastically underestimate the hurdles that lie between the present day and strong AIs. I will be very surprised if they make an appearance in my lifetime.

          I suspect that, somewhat ironically, if we ever encounter anything like a “deity”, it is highly likely to be some other sapient intelligence’s own strong AI that has bootstrapped itself to functional apotheosis, rather than any kind of “supernatural” entity.Report

          • Avatar Kim says:

            Parsimony doesn’t explain Quarks. Or Mister Mark either, for that matter.
            Parsimony doesn’t explain how easy to model the world is, or that the world isn’t analog in the slightest.

            All signs point to the idea that we ARE a simulation.
            There is a god, whatever the fuck “god” means today.Report

      • Avatar Kim says:

        Strong AI already exists, even if it doesn’t get grammar right all the time.Report

  2. Avatar Chris says:

    An example of one of the quips:

    Do you wish people to believe good of you? Don’t speak.Report

  3. Avatar aarondavid says:

    Excellent @chris excellent.Report

  4. Avatar b-psycho says:

    Most people who are big on you choosing God over (No God/It Doesn’t Matter/Reason) don’t leave it at a simple “ok, there’s a god” though. They see faith as synonymous with trying to walk a certain path, operate in a particular way of life, ostensibly as What God Wants of us mortals.

    The Problem: we’re supposed to be able to interpret the by definition infinite & Above Us enough to obey it? How, exactly?

    To my reading then, the wager is incomplete. The question isn’t binary, but infinite. And you know what the odds are of an infinite gamble…Report

    • Avatar Chris says:

      And Pascal, of course, sees faith as dictating a certain path. He’s not trying to pretend otherwise. He explicitly said in the quoted passage what he’s leaving behind to show that even without it, using only the agnostic’s own tools, faith is rational, practical, even preferable. One there, he’s going to pick all the stuff he says he’s leaving out here and go back to Christianity.

      Perhaps the only formal part of the wage r is the one that makes it necessarily binary, though: the infinite vs the finite.

      Here, if you’re interested, is a well-known, relatively recent (2003) paper on whether thinking of it as a binary makes sense:

      • Avatar Francis says:

        But the problem remains — which faith? If God is by definition infinite and unknowable, how do I select the correct faith and correct rules? What if kindness and compassion are actually great sins because they allows the weak to survive?Report

        • Avatar North says:

          Yes, precisely, the practical agnostic’s rejoinder would be “Sure based on your argument there is an thread of rationality within faith but the So What Then is entirely lacking.”

          Nothing says the God Pascal speaks on behalf is the correct one. Near every faith on the planet joins the atheists in thinking that 99% of the theist faiths on the planet are utter farces. That single faith is the only difference. How to tell the correct one? Or if any of the faiths on menu are accurate? Perhaps Gods will is unknowable- it seems highly likely- in which case an agnostic abstention from any single faith seems like the most likely correct answer.Report

          • Avatar Kim says:

            Judaism for one doesn’t do this.
            I’m pretty sure Shintoism doesn’t do this either, nor Confucianism (if you’ll call that a religion, at all, of course)Report

      • Avatar dragonfrog says:

        If there is a God, He is infinitely incomprehensible, since, having neither parts nor limits, He has no affinity to us. We are then incapable of knowing either what He is or if He is. This being so, who will dare to undertake the decision of the question? Not we, who have no affinity to Him

        We have no way of knowing the first thing about God. Let’s take it as a given that indivisible souls exist, and that God (insomuch as He exists) puts so much stock by whether we whom He knows perfectly well, having made us, have no way of determining whether He exists, believe in Him, that he will provide infinitely valuable rewards to those members of His creation (which He deliberately set up to instill doubt) who believe in Him, and cast those who don’t believe into the void because reasons.Report

    • Avatar Kim says:

      Yeah. I find it relatively easy to construct arguments based on empirical data for the existence of a higher power

      But that basically says nothing, absolutely naked nothing about how I should act.Report

  5. Avatar Stillwater says:

    Well, two things, and we’ll see where they lead.

    The first: my response to alot of these types of arguments (the, “ya know, if you don’t believe you’ll end up in hell?” type) is to basically say that if God is all good (and etc blah blah) and by being The Creator gave us humans free will, then God cannot punish me for freely determining to not believe in its existence. There’s more to it than that – coercion and compulsion and stuff like that – but that’s it in a nutshell.

    The second: given that, on what grounds must I choose? Is it a logical necessity? A practical one (like having to choose whether the Beatles or Elvis are better)? Or does the necessity follow from the consequences of choosing correctly/incorrectly? Well if it’s the latter (and it seems to me it’s the latter!!) then I go back to point one, which is that God (an All Good yadda blah…) would not punish me for freely making an incorrect choice just so long as that choice is arrived at by normal best rational practices given available evidence.

    Like I said, that’s my normal response to this sort of thing. Does it get me outa the loop, or am I still somehow not feeling the gripping hand choking off my air?Report

  6. Avatar Murali says:

    The problem with Pascal’s wager is:

    1. He cannot assign utility values to each option (belief when God exists, belief when He doesn’t…) without smuggling in controversial religious options. For instance, if you were Calvinist, you would think pre-destination is true and whether or not you believe would be irrelevant to whether you got to heaven. More importantly If you genuinely drop all assumptions about God, then you cannot even assume that God will not punish you for believing in Him. What if it was atheists who got to heaven and theists who got sent to hell? Absent any concrete religious doctrine that is just as possible as the opposite. More importantly, what if God got angry with you for believing in Him on the basis of Pascal’s wager?

    2. As Stillwater alludes to above there is a wrong reasons kind of objection that can be raised against this. As Pascal readily admits, the wager is not an argument that shows that God is more probable than not. Thus, believing on the basis of the wager argument is not epistemically reasonable. It is prudent (if the argument was successful), but we should not confuse that with epistemic reasonability. Compare with the case where a monster suddenly appears and says it will destroy all life on earth unless everyone starts believing that grass is purple. That would give us prudential and moral reasons to believe grass is purple, but not epistemic reasons to do so. Believing on the basis of such threats even if credible is not epistemically reasonable.Report

    • Avatar Chris says:

      It might be worth pointing to my introductory remarks before addressing these. Specifically the part about this not being an argument for the existence of God, but I’ll add that, as Pascal himself suggests in the quoted passage, this is not why he believes, this is to get to a point where he can talk about why he believes.Report

      • Avatar North says:

        Yes. Pascal makes a really modest and defensible claim and I’d say it’s to his credit and strengthens that claim nicely. He’s been ill used by people on all sides of the argument (but then what philosopher or theologian hasn’t?)Report

    • Avatar Zac says:

      I agree with all of this. I also think it’s worth pointing out that the wager is a classic example of proving too much: you could literally use it to argue for any similar proposition, thus making it functionally meaningless even if accurate.Report

    • Avatar dragonfrog says:

      Exactly – I see nothing in there to suggest that believing in God is likely to produce any kind of reward.

      If anything, the clarification that we can’t possibly know anything about God underlines how very absurd it is to entertain any notion of God caring in any way at all what we believe, or indeed possessing such an anthropomorphic thing as preferences, or providing us any kind of post-mortem reward or punishment for how we arranged our nervous synapses in life.Report

  7. Avatar dragonfrog says:

    Having drunk the suggested glass of wine while reading this, my response is probably somewhat more loosely worded than it might otherwise be:

    The only conclusion I can draw from this is that there’s nothing like religion to make extraordinarily intelligent people devote a great deal of thought and intellectual effort to utter hogwash.Report

  8. Avatar Zac says:

    The only conclusion I can draw from this is that there’s nothing like religion to make extraordinarily intelligent people devote a great deal of thought and intellectual effort to utter hogwash.

    +1. Rarely have truer words been spoken.Report

  9. Avatar James K says:

    In addition to all the other problems everyone noted Pascal is performing an illegal operation in his wager. I can’t fault him for this – the mathematics of decision-making hadn’t been formulated yet, even calculus was invented in his lifetime.

    The important thing to understand is that infinity is not a number. A proper mathematician would be better able to explain why, but you can’t just go sticking infinities into equations willy-nilly. This has implications for the real world when the equation you are looking at is a utility function (which is effectively what Pascal is specifying, even if he was doing so long before the utility functions had been conceptualised). Stuffing an infinity into a utility function is an illegal operation, akin to a Divide by Zero error.

    You can see the problem by looking at what adding infinities to a utility function does. Any infinity overrides all other considerations in making a decision, and once you start adding infinites, you can’t stop. Is there a 0 probability that a god exists that will torture you for all eternity unless you give me $10? I’ll send you my PayPal address. For that matter you cannot assign a 0 probability to any given action leading you to eternal damnation if you do it (or not do it). As @zac notes above, it proves far too much. Indeed the very concept of making a decision collapses under an infinity of infinites.

    Much like Zeno’s Paradox, Pascal’s Wager is a philosophical conundrum that only exists because the philosopher who coined it didn’t have access to the proper mathematical tools to realise they had made a mistake.Report

    • Avatar Chris says:

      Hmm… I don’t know of any mathetmatical system in which infinity is not considered to be a number, though the mathematicians ’round here would know better than I. Still, while I’m wary of decision-theoretic interpretations of the wager generally, given that decision theorists have played with it, I’m skeptical of the notion that the notion of infinity makes it invalid (and much of the discussion of it in the last few decades has been about what sort of infinity we’re talking about).

      Let’s take a step back again, and consider the context of the wager: First, we know something very important, namely that in the context of the wager there is a non-zero possibility of the existence of God (we know this because Pascal’s interlocutor is an agnostic who admits as much in the text). Second, we know that Pascal actually tries a few different tacks (see the SEP article on the wager, as well as this excellent old post (Brandon, who was a blog friend once upon a time when I had one, is responsible for much of the way I think of the wager), wagering an infinite reward against a specific finite reward (one life) as well as against a finite loss (misery, something Section II of the Pensées discusses in depth; though some have interpreted this as a negative infinity, not a finite loss). Third, the purpose: not to demonstrate that God exists, but to demonstrate that it is not unreasonable to choose faith, or at least to entertain the possibility of doing so enough to then consider what Pascal believes to be the true arguments for God’s existence.

      When we consider it in this context, and sticking to the text, most of the objections melt away, because they’re objections to something not being argued.Report

      • Avatar Stillwater says:

        Third, the purpose: not to demonstrate that God exists, but to demonstrate that it is not unreasonable to choose faith, or at least to entertain the possibility of doing so enough to then consider what Pascal believes to be the true arguments for God’s existence.

        If this is right, then the purpose of the wager is to establish (justify!) either a psychological or an epistemic precondition which Pascal views as necessary/sufficient to move on to other arguments actually justifying God’s existence. On either score, I think Murali is right (just as I am right upthread) in rejecting that the Wager actually does establish those (either!) psychological or epistemic preconditions.Report

        • Avatar Chris says:

          It might be insufficient on most grounds other than those laid out by the agnostic at the start of the passage, but I’m not sure why it would be insufficient on those specific grounds.

          If, on those grounds, all we need from the wager is to say, “You know, it might be reasonable to at least consider faith as a good choice,” I think it does a pretty sufficient job.

          Think of it as Pascal confronting one of Kant’s antimonies and, instead of offering up a metaphysical or even a strictly epistemological argument for choosing one side or the other, offering a practical (not in the Kantian sense, but in a, say, economic sense) argument that we should take a closer look at the sorts of arguments that don’t work in strictly natural philosophical reasoning. There are two possibilities, god or no god (these are the possibilities that we start with), and one might lead to infinite reward, so maybe you should consider it further.Report

          • Avatar Stillwater says:


            I get that you think it’s a good starting point. And I take it that Pascal views it as an excellent starting point. But it seems like I am still unmoved.Report

            • Avatar Chris says:

              Then you are hopeless and will burn in hell for an eternity. Or not. Maybe it’s just that you won’t get to participate in Taco Tuesdays in heaven.Report

              • Avatar Stillwater says:

                Then you are hopeless and will burn in hell for an eternity.

                There’s only a reallyreallyslim chance of that happening tho, right?


              • Avatar Chris says:

                But if you multiply that by a negative infinity (of tacos), it doesn’t look so good.Report

              • Avatar Stillwater says:

                So, you’re saying restaurants in hell will feed me so I won’t have to cook my own food? Now I’m thinking the downside isn’t so bad after-all.Report

              • Avatar Stillwater says:

                More seriously….

                Since I take it you don’t like my earlier excuse to not play the game, let me try it this way.

                What has to be in place for the Wager (or any other similar argument) to effect its desired outcome? That Gawwd will punish non-believers. If so, then how can even entertaining the conclusion that Gawwd (might!) exist be rational (or reasonable, or epistemically justified) consistent with all of our background assumptions regarding what constitutes rationality (or reasonableness, or epistemic justification)? (I mean, I could go on about this … and on and on … but that’d be boring, yeah?)

                Like Ivan (maybe) I don’t reject Gawwwd, at this point I just return him the ticket.Report

              • Avatar Chris says:

                Well, as I said above, he takes different tacks. In one, there is misery (though again, see Section II for what misery means in this context), but in the earlier one there’s just eternal life or not eternal life. The reward is simply posited to be potentially infinite, and it is contrasted once with one life and once with a possible finite, negative reward (misery).

                But his position that one must choose is not that there is some agency forcing one to do so, but merely that by being one has taken a stance. By not choosing, you have in fact chosen, and when he admits this possibility in the texts, he basically says to think about why you doubt.Report

              • Avatar Stillwater says:


                I get the structure of the argument. (Even taught it to impressionable young minds!) Maybe this is the way to say it: I’m rejecting that I have to make a choice on the terms Pascal posits. I can merely – as I said upthread – return him the ticket and refrain from going on that ride. And I’ve given two – which sorta collapse into one or explode into three or four – epistemically justified, rationally based, reasonably reasoned reasons for doing so. I don’t know why that’s not the end of the story, myself. I mean, I get that there’s an allure to all this, but it’s not so much philosophically as psychologically based, and the parts that are philosophically based seem pretty easy to reject.

                But that’s the voice of the Devil speaking thru me!Report

          • Avatar dragonfrog says:

            It sure is handy when you’re writing your own interlocutor, innit? They always do seem to be won over.Report

            • Avatar Chris says:

              Well, I imagine he chose that interlocutor for a reason, perhaps because it was a common philosophical position at the time.Report

              • Avatar dragonfrog says:

                It was written to come out close enough many readers would mistake it for such, yes.

                But how many actual agnostics really held to that view – that there may or may not be a God, that we can know nothing at all about it because if existent its infinities have nothing in common with us finite and imperfect beings, but if it does exist it very obviously thinks it extremely important that we finite beings believe in it – so much so that it has (again so obviously it doesn’t need further questioning should we accept the first proposition of the existence of God) set up a heaven to house those who believe in infinite rapture, and either a hell or a vaporizing machine for those who don’t?

                Were 17th C agnostics really that much less capable of thinking through the implications of agnosticism than modern ones?Report

              • Avatar Stillwater says:

                Personally, I think it’s important to view – even if it requires interpretation away from the actual words he wrote! – Pascal as making the strongest possible in-context argument. So, rejecting his arguments because they didn’t express what actual agnostics were saying at the time doesn’t effect the philosophical points he’s making.

                Personally, tho, even on those terms I don’t think the argument gets him to where he wants it to go. (No knock on Pascal, of course, and no knock on folks who think it actually does!)Report

              • Avatar dragonfrog says:

                The whole point of agnosticism is that, not only can we know nothing of God’s existence, if there is a God we can know nothing of what a utility-maximizing reaction to that fact is. That in the ridiculously unlikely event that any existent God wants a specific action from us and has, for reasons not related to the social priorities of bronze-age patriarchs, set up an otherworldly reward/punishment system, there is exactly as much chance that any attempt we make to reap the reward will land us in the ‘punishment’ cohort as in the ‘reward’ cohort.

                Pascal just punts on everything after the ‘existence’ part. I find it really hard to believe that agnostics in the 17th C were actually sitting at “it’s a coin toss between whether a God doesn’t exist at all, or the entire doctrine of the Catholic church is correct”Report

      • Avatar Kim says:

        I’m pretty sure infinity isn’t a number in floating point, or in integer notation on a computer. Digital numbering systems (rather than analog) would seem to exempt you from uncountable infinities, anyhow… [Not a Math Major, take it as you will]Report

  10. Avatar Jaybird says:

    How much trouble does it take to flip this around Lovecraftian style?

    “Why are you sacrificing your children to Moloch?”
    “Pascal’s wager.”Report