"Now, this is hard. The numbers are a bit made up at times. But reasoning with made up numbers is often better than reasoning with no numbers at all. Human intuition isn’t good at figuring out probability, so it can often be improved by Bayesian analysis. Even though it often won’t be clear whether the odds ratio is 5:1 or 10:1, you can usually have a rough order of magnitude estimate."
I know you qualified this a lot, but I still think it dramatically understates the difficulties involved. Rough order of magnitude estimates cannot be taken for granted. There are plenty of questions where it would be a major scientific achievement to come up with well grounded rough order of magnitude estimates. E.g., what's the probability of life emerging on a habitable planet? If you could nail that down to within a few orders of magnitude, origins of life researchers would be thrilled. And that's given lots and lots of rich background information about what the universe is like. If I'm told to make analogous estimates (e.g., that there should exist matter) totally a priori--given no factual background assumptions at all--I think the honest answer is nothing I come up with should be taken seriously at all.
What's the alternative? I think we're probably a lot better at making empirically informed posterior probability judgments, and then working backwards to think what priors we'd have to have for those to be sensible posteriors. But if that's the approach you take, it's going to be a lot harder to take totally a priori constraints on prior probabilities as premises, and use them in a dialectically effective manner to impose constraints on posterior probabilities.
I don't disagree that this is a fraught enterprise and it's easy to go wrong. I guess I think that there's some evidential significance if you notice that one view wins even when you use conservative estimates for its Bayes factors, but I agree it's nowhere near decisive.
I guess what I want to dispute is that we have any good grip on what amount to "conservative" estimates when we're trying to come up with probabilities totally a priori.
Hmm, that seems a little extreme. Suppose, for instance, that someone gave fine-tuning a Bayes factor of 1.00001 or evil a Bayes factor of 1.2. Surely that could be aptly described as conservative?
I'm not sure! For what it's worth I'm much much less impressed by the problem of evil than most atheists. I have a really hard time describing what sort of world I think a good God should have created. Maybe this seems like a tangent but I think it's no accident that there's no compelling fiction set in heaven; it's really hard to tell an interesting story where everything is always awesome. So I'm not much inclined to criticize a theist who says the existence of evil barely moves them at all.
As mentioned in a comment below, I think the real issue is to pick point estimates given how uncertain these factors are--if you can think of a reasonable range, you should do a calculation incorporating the whole range of plausible values.
I have some pretty strong methodological worries here:
For one, you seem to be just throwing in a new ratios without updating the extent to which the data already included the prior credence function confirm the hypothesis. But that’s clearly not the right way to think about things. The problem of evil does directly disconfirm theism, but it dually undercuts the evidential value of the various arguments in favor of theism — they aren’t considerations we can just completely bifurcate and reason through independent of one another. We expect a benevolent god to fine-tune a *maximally good* world, not just any world at all. Likewise, the benevolent god theory predicts that god would create tons of *maximally good* lives, not a mixed bag with some pretty good lives and lots of horrible lives. When we learn evidence that suggests the world isn’t maximally good and that he hasn’t created an absurd amount of maximally good lives, we also need to significantly lower the extent to which fine-tuning considerations confirm theism because we realize we’ve observed a world that theism doesn’t actually predict. The value of stuff like fine-tuned laws, lots of life, consciousness, etc. is just instrumental to value creation, but if the latter doesn’t obtain, we shouldn’t expect think the former alone confirms theism. In other words, the predicted datum is <fine-tuning + maximal goodness>, not two separate data of <fine-tuning> and <maximal goodness> such that the former by itself would be strongly confirming. Two quick points of clarification:
A] Of course, we might think that our world is close enough to maximal goodness (perhaps because your credence in the problem of evil isn’t super close to 1) that our observations do give us at least some reason to believe that <fine-tuning + maximal goodness> has obtained in our world. But that’s a VERY different way of reasoning than the way you’ve reasoned in this post, since you are treating them as independent data.
B] Maybe you don’t share my view that value requires phenomenal consciousness - e.g. maybe you think a cool waterfall is valuable even if there’s literally zero minds to perceive it or appreciate it, but this is something you’d need to substantively justify to get your arguments off the ground. I also think this is probably wrong - Uriah Kriegel has some good arguments about why we should reject that view.
Second, I think trying to reason about really abstract, complicated matters like the probability of theism with formal Bayesian mathematics probably hurts more than it helps. Assigning numbers is always pretty arbitrary, and the complexity of mathematical modeling means we are more likely to make mistakes when updating our prior credences than we otherwise would be. As they say, a Bayesian is someone who reasons normally but moans Bayes name as they do it!
Finally, why think there’s an objective correct prior that we should all have? My impression is that most epistemic internalists tend to be subjective Bayesians, so I’m curious why you aren’t.
(As a side note, I think 4:1 is seriously underestimating the strength of arguments from evil, but that is a separate matter that I think would be really stupid to try to convince you of in a comment section here — I know you have a lot to say on this that you’ve written about elsewhere.)
Wait sorry I gave the problem of evil a Bayes factor of 10,000, not 4.
I'm not a subjective Bayesian because some priors just seem crazy. If you started out with a prior of 1 in flat earth, that would seem clearly irrational.
I was imagining that at every point where we counted a fact we included the other facts in the background. I don't think the POE undercuts the evidence that much--even in a world where God allows evil, it seems like in a decent portion he'd want agents who can make significant decisions, rather than defective disharmonious blobs. Even if you think the odds are only .1 of creating such an agent given theism, you'd still end up with a Bayes factor of 3200:1.
I agree we'd expect God to probably make the best world! The fact he doesn't is surprising. Now, he could make maximally good finely-tuned worlds, of course. But this was why I gave not a maximally good world a high Bayes factor.
I agree that consciousness is needed for value. But it's not super obvious that the universe that God would create would, at every moments, appear maximal in regards to all value facts. It might be that there are subtle connection so that a world like ours is part of some ideal total existence. Seems like this has odds of over 1/10000!
Thanks for the reply Matthew - I think it’s great that you are so willing to engage with criticism. That takes a lot of epistemic virtue; you are setting a great example for this platform. Anyways, here’s my thoughts on the substance:
1. Whoops — sorry I got the numbers mixed up, 4 was for absence of miracles. Apologies for not reading carefully enough.
2. As for the flat earth claim, I don’t think this is an argument against subjective Bayesianism. The subjective Bayesian would just say that someone who doesn’t find smth crazy would have a higher prior. If you think intuitions provide defensible evidence, then I’m not sure why someone who thinks about a strange proposition but finds it intuitive wouldn’t have justification for it before evaluating any evidence. In other words, it seems like phenomenal conservatives should like subjective Bayesianism. It doesn’t seem very conservative to be committed to the view that someone’s intuition doesn’t provide justification if we consider it crazy. (Also I think most people actually have a high prior in flat earth - that’s why most humans believed it for most of history. What is crazy would be if I didn’t believe in flat earth after considering all the evidence because soooo much disconfirms it. After updating, I have a low credence, but before I know anything about science I probably don’t)
3. Regarding the problem of evil undercutter: this seems like it just gets into a debate about the merits of various theodicies. But if we think the problem of evil isn’t solved by various theodicies (e.g. if we think it’s false that “he’d want agents who can make significant [sometimes evil] decisions”), then the undercutter remains extremely strong. I personally don’t find the free will theodicy compelling - I don’t think leeway freedom is required for free will, nor do I see why a perfect world would include it. I’d imagine that maximally good world would involve free agents who naturally have great dispositions such that they freely act morally at all times. Theres also lots of natural suffering. The point about the possibility that non-maximal goodness in this world being necessary for maximal goodness across all worlds along belongs here — it’d offer reason to think god would actually allow evil in this world. I’m not sure how much weight to give this, partially because it’s unclear why evil in this world could be causally linked to the amount of goodness in other words such that the former effects the latter - and partially because some of the arguments for a multiverse depend on claims about the existence of a good god which becomes less likely given evil, so updating here gets tricky because there’s a recursive element.
I don’t think it’s right to calculate the odds as 3200:1, because it’s not just undercutting the total odds, but rather undercutting each individual argument for theism - so fine-tuning of constants is multiplied by 0.1, anthropic arg by 0.1, etc. etc. (there are certainly some arguments that seemingly escape unscathed like the moral argument, but those tend to be pretty weak anyway - the good arguments tend to rely on a premise that god would do X to actualize value, and we observe X). If you just count it as 0.1 at the end, you are basically updating on only one datum being undercut which isn’t the claim being made here. You’d have to lower the degree of confirmation of every piece of evidence that relies on a value-actualization premise, which is most of them. If 5 arguments are undercut (not sure the exact number - I haven’t looked back at all the exact args you make, but let’s just say it’s 5, the math would look like this: 32,000*.00001=0.32
The final ratio would then be 1:3.125 (aka ~3x more likely that theism is false)
Again, maybe our credence in arguments form evil is low enough that the undercutting aspect isn’t enough to tip the balance of the scale, but I hope it’s at least clear how it might. I think the most evidentially salient question in the debate about theism is what to make of evil, because it really does have drastic implications for literally every other important argument.
I don't think it would undercut each bit of evidence because they're not independent. What I gave at least odds of .1 to--but fine, let's have it be .01 to be generous--is that God, in a world where he allows evil, would want to make agents who can understand the world. That simply entails doing the other things he did, and so it won't affect the cumulative Bayes factor by more than a factor of 100.
For instance, conditional on God having reason to allow evil and create a physical universe, the odds of him creating laws are near 1, as are the odds of him having nomologically harmonious laws.
If the arguments for theism are not independent, then why are we treating them as independent when calculating the 32,000:1 odds? It seems like we’d be treating them as discrete confirming data when it aids theism but then suddenly assimilating them when asking about undercutters.
If you want to maintain that they are *not* independent, we don’t get to calculate the odds for theism in the way you wanted because we only multiply them (without considering joint and conditional probabilities) if we’re dealing with probabilistically independent data.
There’s basically a double-bind: either a) they are dependent, in which case you’ve improperly calculated the probability of theism wrong since you just multiplied a bunch of numbers together, or b) they’re independent, so you undercut all of them.
I’m pretty sure you can change the prior and conditional probabilities, with support, to get any answer here. And the justifications provided don’t suggest anything about the leap from believing in God to believing in a personal God.
Wow, this is uncharacteristically bad analysis. Here's some of the mistakes you're making:
1) Misdirection. You're starting with the low prior of 1/1000 chance of Theism and then listing a huge number of arguments in favor and 1 against. To the untrained eye, it looks impressive that you can be "conservative" with such a low prior and still end up with 32,000:1, but the prior is meaningless when the evidence selection is skewed.
2) Even with your early caveat that all the odds are subjective and the result can't be taken too literally, you're still giving dramatically too much weight to the final output. Using Bayesian analysis for purely subjective arguments is less than worthless, because it gives us a false precision. You're casually throwing around odds of 200:1 in favor of Theism for the fact that we have consciousness with essentially no justification. Why not 20:1? Why not 2000:1?
3) You're giving the "absence of miracles" a 4:1 ratio against Theism? In other words, the fact that every unexplained phenomenon in all of history originally attributed to divinity has since been explained away is worth a 4:1 against Theism while the fact that "When you put your hand in hot water and feel pain, you pull your hand away" is worth 20:1 in favor? lol?
4) There are so many more arguments against "an all-knowing, all-powerful, all-good being" than the problem of evil. For example, divine hiddenness, inconsistent revelations. Give them each 100:1 as you casually gave many of the arguments in favor of Theism and suddenly the odds are 3.2:1 not 32,000 to 1!
5) On your Point #3: "Absent a necessary being explaining the existence of contingency, contingent stuff existing at all is surprising. It would be much simpler for nothing to exist at all." Umm, okay. What explains the existence of the being? This is like Theism 101 stuff. To assume the existence of the being is to make at least the same logical leap as to assume the existence of "contingency."
6) On your Point #3: "Taking into account the existence of a physical universe, the odds of such a thing are very near one. After all, God has no reason to create particles that don’t do anything." So why did God create a Universe that is trillions of times bigger than our planet?
7) On Point #4 regarding fine-tuning- I know you have written in depth about this argument. But I and many others find that anthropic reasoning completely negates the appearance of fine-tuning.
8) On Point #5 regarding Consciousness: This is simply a misapplication of Bayesian reasoning. The fact that we cannot predict consciousness with certainty from the outset doesn’t mean it was an improbable outcome, it just means that it wasn’t expected based on the models we had prior to discovery. Consciousness is today often considered an emergent property of complex systems.
9) On Point #6 regarding psychophysical harmony: have you heard of evolution? We already have a theory that predicts a relationship between mental and physical states being reinforced over time. And the fact that when we say, "I see a brown desk," it aligns with what’s going on in our mind simply reflects that our language evolved to describe our perceptions. This demonstrates our ability to interface with the world around us and communicate our thoughts effectively, not evidence of divine coordination.
10) On Point #7 regarding Misc: "You exist! Out of the uncountably infinite possible people who could exist, you happen to. On atheism, this is very surprising—even if someone exists, it’s very unlikely it would be you." Wow. Are you also surprised that your shoes happen to be the exact size to fit you? Do you think it's amazing that our atmosphere happens to have the right % of oxygen which we happen to breathe? You are taking our individual identity as a given and then concluding it is wildly improbable for that identity to exist. But the fact that you are you is simply a result of your surroundings. Just as we evolved to fit our environment, so too has our identity emerged. No matter "who" you are today, it was never predestined. The fact that you give a 100:1 ratio for this in favor of Theism about wraps up my criticism here.
1) I mean, any specific hypothesis will have a low prior. So .001 isn’t some insanely low prior for rhetorical effect. My aim was to be accurate and conservative.
2) Well, it’s hard to be precise about these things. I could imagine a person who was only inclined to give it a Bayes factor of 20 (though 1 in 20 odds of consciousness randomly appearing given naturalism seems a bit nutty to me). But one could do that with all the numbers. As I said, feel free to plug in your own numbers.
3) I disagree that every miracle has been better explained by naturalism. And I included elsewhere a broadly indifferent universe operant according to natural laws, so this is just about the extra fact that there aren’t random suspensions.
4) I list the all! But I don’t find them very persuasive, because they’re just examples of bad things, and I think whatever explains why all the evil in the world exists will probably explain those too!
5) Maybe it would be helpful to read the argument on the contingency argument. The difference is that God has a plausible claim to be necessary, unlike the universe. Necessary things don’t need some outside cause.
6) Again, I counted the argument from scale separately, and gave it a generous Bayes factor even though I think it’s actually an own goal benthams.substack.com/p…
The fact I listed there was merely: a physical universe exists!
7) The anthropic reply is rejected by roughly 100% of philosophers for good reason. Exactly the same logic would say that your existence doesn’t give you good evidence that your parents didn’t use highly effective contraception because if they had you wouldn’t have existed.
8) If a fact is more expected given hypothesis A than B, even if B post hoc comes up with an explanation, then it’s evidence for A. That’s exactly how Bayes theorem works. See also the examples I gave where you retroactively declare something necessary.
9) No, I’ve never heard of it! What’s that!
You should probably, if you are going to say that psychophysical harmony is undermined by evolution, read the things that are included in every single piece ever written about psychophysical harmony explaining why it doesn’t. The one sentence version is that even though evolution explains why people behave in various ways, for that’s what natural selection selects for, it doesn’t explain why those adaptive behaviors are associated with harmonious mental states. For more detail, see philarchive.org/archive….
10) No because there’s a selection effect that makes it likely that your shoes would fit you. Most people have shoes that fit them, so you shouldn’t be surprised about that fact. Most possible people don’t exist (at least given naturalism) and for reasons I have explained at length including in published philosophy papers, your existence favors hypotheses on which there are more people. (A theory on which there is a mutiverse with 100,000 earths makes your existence 100,000 times likelier than a theory on which there’s just one earth). If you want to know more, read the linked pieces on the argument.
If you are going to be this snarky, I would advise learning something about the arguments you are making fun of.
> Exactly the same logic would say that your existence doesn’t give you good evidence that your parents didn’t use highly effective contraception because if they had you wouldn’t have existed.
“Your existence” alone does not imply anything about your parents or contraception. You need additional facts about human reproduction and the effects of contraception to make that conclusion. With that data, you know that the chances of you being the biological child to people who used contraception is low.
In contrast, your argument isn’t about which specific person you are. It’s generalized to the level of “you exist”. But you had to exist in some form somehow to do probability calculations, and naturalism predicts that some beings will exist. So pure existence (agnostic about the specific circumstances of your existence) isn’t evidence for anything in the way that.
Since the Anthropic argument was always about existence in the abstract being necessary, I don’t think this is the kind of ad hoc epicycle that reduces the persuasiveness of an argument.
Two mathematical points, leaving aside all the many other issues that others have raised.
1. When multiplying point estimates and ignoring uncertainty, you can get misleading results, see e.g. Sandberg, Drexler, and Ord's reevaluation of the Fermi paradox--even if point estimates for the parameters imply a strong conclusion that aliens should exist, doing some simulations just by incorporating a range in which your parameters might fall can give you an estimate that there is a 1/3 chance we are alone.
If I understand it right, what you get (and I think this will be generic, for simple assumptions about the uncertainty range of your parameters) is a power law distribution, with small numbers of universes with huge numbers of aliens, and huge numbers of universes with no aliens.
I think it would be interesting to redo your analysis along these lines, and see that the overall distribution looks like.
2. A point one of the judges, Eric Stansifer, made in the Rootclaim debate over COVID origins: if you're picking pieces of evidence at random, even if the overall evidence is evenly balanced and you pick without bias, there is a high probability that you end up with an overwhelming Bayes factor decisively in favour of one hypothesis over the other... And if your selection of evidence is a little bit biased, this is even more pronounced.
To counteract this it's best to either have a canonical factorization of your problem into component pieces so that there's no choice in what pieces of evidence to look at, or to agree in advance what pieces of evidence to look at.
I think to be convincing you should make it clear either why you think, a priori, these are exactly the things we should agree to look at to decide this question, or to present some model where you factor the question of "does God exist" into a calculation where these are the only parameters.
"The numbers are a bit made up at times. But reasoning with made up numbers is often better than reasoning with no numbers at all"
Sure I guess, if you also acknowledge the limitations and potential for error when the underlying data is unreliable. Pulling numbers out of a hat is just a recipe for false precision.
The most likely 'probability' in these estimates is that the universe just exists. Nothing else is needed to explain it. This would be more likely than theism by your logic.
> God might be the very simplest entity: a mind or agent totally without limit.
That's not simple. To take the example of the sequences, "a witch did it" sounds simple because we've cached the incredibly complicated concept of a witch (all the complexities of a human plus all the complexities of "magic" (something so complicated it doesn't even exist) plus all the complexities of a human-magic interface) into one word. Same with god.
Minds and agents are incredibly complicated. Mind bogglingly complicated. And they don't get simpler as they approach infinite mindfulness/agency. The steel man I can give of a limitless mind is if there's something like panpsychism and the whole universe creates one mind (in the same way that a whole brain creates one mind), but then I'd say, call that something other than "god" or you're gonna create a whole lot of confusion.
It's not enough that the English sentence describing a thing is simple: it has to be ontologically simple. You say that minds and agents are complex, but if dualism is true, then that isn't so--on dualism minds are fundamental and irreducible. And I think there are good arguments for dualism https://benthams.substack.com/p/marys-room-refutes-physicalism. At the very least one should have credence in dualism well above 1%.
Dualism, and in particular arguments for it like Mary's Room, may suggest that certain features of mental life like qualia might be fundamental, but they don't automatically imply that beliefs, attitudes and other psychological concepts that you want to build into theism are fundamental.
I suspect you'll reply that even if Mary's Room or the like don't do the heavy lifting of showing that these concepts are fundamental, they do indicate the truth of dualism generally, and that you don't think conditional on dualism that those specific things being fundamental is all that crazy (regardless of whether you actually have a strong argument for their fundamentaliy or not). But I think it's hard to make progress against moves like this. You have to make sure that you're being similarly generous to all of naturalism's possible "not that crazy" purely speculative sub-hypotheses, of which there are innumerably many.
I think that either the prior on theism should be something like one in a gazillion, or else it should be as high as you say (because of under-explored epistemic unknowns) but the likelihood of naturalism on fine-tuning/psychophysical harmony/whatever should also be relatively high (because of under-explored epistemic unknowns). Of course, I would also add on to this that I think there are separate objections that largely defuse these arguments so I don't have to rely on the unknowns, but still.
Even under that, just because electromagnetism is fundamental doesn't mean the universe is a supercomputer, or that a supercomputer is simple. I think that even with consciousness as fundamental, minds as we know them are a unique, complicated arrangement of the fundamentals, just like computers are unique arrangements of the electromagnetic fundamentals. So maybe with those fundamentals there is an unbounded supercomputer and an unbounded mind, but I don't think it would look like anything we know of as a computer or mind, and I don't think the mind version would look like what most people would call god
If electromagnetism is simple then if there was something that was just pure limitless electromagnetism, then it would be pretty simple! And in any case, minds are more joint-carving than electromagnetism, which is just a random physical force.
Even if minds we know are complicated, that doesn't mean all possible minds are. As an analogy, life we know is very complicated--much too complex to form by abiogenesis--but that doesn't mean abiogenesis is impossible!
How could we tell whether qualia/consciousness is simpler than electromagnetism? If they are equally complex, should we have a 1:1000 prior on a pure limitless electromagnetic "thingy" (for lack of a better word) existing?
If a mind was limitless, wouldn't all minds be part of it? Anything that would separate our minds from the limitless one would be limits, which it can't have
On the separate class of reality point, why couldn't we also say that electromagnetism is a class of reality with parts that interact with physical world in predictable ways? Such as how the qualia class of reality interacts in predictable ways with the physical world (although we have a better understanding of electromagnetism/we're better at predicting it). For example heroin interacts with our brain which interfaces with the qualia realm to produce pleasure, and on the other direction, the mind class interfaces with our brain to produce signals to control our musculature, and so on for the various inputs and outputs of minds
Mary's Room is a bad argument for dualism because it's question begging. But it doesn't stop there. What would zombie Mary say when she sees red for the first time? If she's surprised by the redness of red, then zombies aren't possible. If she isn't, then the original Mary does not experience any nonphysical red. Believing the Mary's Room argument destroys the other conceptual tools that are meant to be pro dualism without the argument giving independent justification for being a dualist.
>A physical world is contingent—it could have been otherwise! Contingent stuff existing at all is somewhat puzzling on atheism. Absent a necessary being explaining the existence of contingency, contingent stuff existing at all is surprising.
Calling something "necessary" does no explanatory work in explaining its existence. Or if it does, the atheist can just claim the physical world has that property, or that God lacks that property. And by the nature of the methodological assumptions behind the contingent/necessary dichotomy, there will be no resolution either way. I feel like this is a problem that keeps reappearing in these theistic considerations so the ultra giga bayes factor you end up with at the end should be heavily in favor of there being no fact of the matter/this being a meaningless discourse with too many faulty assumptions to evaluate probabilities.
Plus, even given an uncaused causer, I would think the original unmoved mover is more likely to be as a grain of sand is to a pearl, than as any normal concept of god is to the universe
Can anyone tell me how this is any different from saying 'There is a lot of stuff in the world that I can't make sense of, so God must have done it.'? It is essentially the same as creationists marvelling at all the wondrous features of bananas.
Well it is just replacing the banana with the cosmos, natural laws, consciousness, psychophysical harmony and evil i.e. just a long chain of arguments from ignorance.
1: an intelligent creator of the universe exists, and he is all-powerful, all-knowing, and all-good.
2: an intelligent creator of the universe exists, but he lacks one or more of those properties.
3: the universe was not intelligently created.
You focus on 1 and 3. But you ignore 2. I consider 2 a lot more likely than 1.
So even if I spot you that there is a cosmic appearance of design (consciousness, psychophysical harmony, nomological harmony, etc.) it seems like all it gets us is an intelligent creator, not the "triple omni" version of theism you seem to want.
Moreover, we can distinguish eight different conceptions of a creator, depending on which of these three properties he has or lacks (1: all-powerful, all-knowing, and all-good... 2: all-powerful, all-knowing, but not all-good... etc.).
Arguably, the Christian one of those—the one where he has all three omni properties—is the least plausible of the eight. At best, your arguments tell in favor of the disjunction of these eight options.
The conclusion of these types of analyses seems to hinge a lot on how independent you think the various arguments / lines of evidence are. You say that atheism has only one significant piece of evidence in its favor, whereas theism has several. Even though for atheism you mention the problem of suffering, divine hiddenness, religious confusion, and absence of miracles, you say that really these are not independent data points, because we only to need to have an explanation for one thing, a blind and indifferent world, and once we have that, then all of these observations are explained in one fell swoop.
Ok, fair enough, but I don't see why the theistic arguments aren't similarly collapsible in this way. Scott Alexander had a post earlier this year about how Tegmark's Mathematical Universe Hypothesis defeats many of the arguments for God's existence. Of the ones you mentioned here, I think MUH could explain fine-tuning, nomological harmony, and the anthropic argument. If MUH is true, then it changes your conclusion by several orders of magnitude because it explains multiple of your theistic pieces of evidence.
Basically, conditional on knowing that there is a good atheistic explanation for one of the theistic arguments, that ought to at least somewhat impact your credence in the others as well, because it could be the same underlying theory that explains multiple of them. For this reason, I think it is a mistake to simply multiply the odds together as you have done here.
> So in deciding upon the prior you should ask: if you were in a dark room with no access to the world, and hadn’t considered your own existence (perhaps you were tripping on a powerful drug that eliminates your sense of self), how likely would you think it is that an omnipotent, omniscient, and omnibenevolent God exists.
If I were in a dark room with no access to the world, I’m not sure I could reasonably come to the conclusion that omnibenevolence was part of the equation that got me there.
If I then were let out into the world to find out I’d certainly die some day and very likely in agony, as would all other living things that were for whatever reason made capable of agony, I think omnibenevolence would be even further from my mind.
Omnibenevolence has never been a good explanation for suffering, let alone for astronomical amounts of it.
> This isn’t surprising given theism—consciousness is the source of most value
What is surprising is that physical brain defects can nullify the consciousness and value of the soul. It's probably because souls don't comport well with a broader picture of psychology, neuroscience, etc.
"Now, this is hard. The numbers are a bit made up at times. But reasoning with made up numbers is often better than reasoning with no numbers at all. Human intuition isn’t good at figuring out probability, so it can often be improved by Bayesian analysis. Even though it often won’t be clear whether the odds ratio is 5:1 or 10:1, you can usually have a rough order of magnitude estimate."
I know you qualified this a lot, but I still think it dramatically understates the difficulties involved. Rough order of magnitude estimates cannot be taken for granted. There are plenty of questions where it would be a major scientific achievement to come up with well grounded rough order of magnitude estimates. E.g., what's the probability of life emerging on a habitable planet? If you could nail that down to within a few orders of magnitude, origins of life researchers would be thrilled. And that's given lots and lots of rich background information about what the universe is like. If I'm told to make analogous estimates (e.g., that there should exist matter) totally a priori--given no factual background assumptions at all--I think the honest answer is nothing I come up with should be taken seriously at all.
What's the alternative? I think we're probably a lot better at making empirically informed posterior probability judgments, and then working backwards to think what priors we'd have to have for those to be sensible posteriors. But if that's the approach you take, it's going to be a lot harder to take totally a priori constraints on prior probabilities as premises, and use them in a dialectically effective manner to impose constraints on posterior probabilities.
I don't disagree that this is a fraught enterprise and it's easy to go wrong. I guess I think that there's some evidential significance if you notice that one view wins even when you use conservative estimates for its Bayes factors, but I agree it's nowhere near decisive.
I guess what I want to dispute is that we have any good grip on what amount to "conservative" estimates when we're trying to come up with probabilities totally a priori.
Hmm, that seems a little extreme. Suppose, for instance, that someone gave fine-tuning a Bayes factor of 1.00001 or evil a Bayes factor of 1.2. Surely that could be aptly described as conservative?
I'm not sure! For what it's worth I'm much much less impressed by the problem of evil than most atheists. I have a really hard time describing what sort of world I think a good God should have created. Maybe this seems like a tangent but I think it's no accident that there's no compelling fiction set in heaven; it's really hard to tell an interesting story where everything is always awesome. So I'm not much inclined to criticize a theist who says the existence of evil barely moves them at all.
As mentioned in a comment below, I think the real issue is to pick point estimates given how uncertain these factors are--if you can think of a reasonable range, you should do a calculation incorporating the whole range of plausible values.
I have some pretty strong methodological worries here:
For one, you seem to be just throwing in a new ratios without updating the extent to which the data already included the prior credence function confirm the hypothesis. But that’s clearly not the right way to think about things. The problem of evil does directly disconfirm theism, but it dually undercuts the evidential value of the various arguments in favor of theism — they aren’t considerations we can just completely bifurcate and reason through independent of one another. We expect a benevolent god to fine-tune a *maximally good* world, not just any world at all. Likewise, the benevolent god theory predicts that god would create tons of *maximally good* lives, not a mixed bag with some pretty good lives and lots of horrible lives. When we learn evidence that suggests the world isn’t maximally good and that he hasn’t created an absurd amount of maximally good lives, we also need to significantly lower the extent to which fine-tuning considerations confirm theism because we realize we’ve observed a world that theism doesn’t actually predict. The value of stuff like fine-tuned laws, lots of life, consciousness, etc. is just instrumental to value creation, but if the latter doesn’t obtain, we shouldn’t expect think the former alone confirms theism. In other words, the predicted datum is <fine-tuning + maximal goodness>, not two separate data of <fine-tuning> and <maximal goodness> such that the former by itself would be strongly confirming. Two quick points of clarification:
A] Of course, we might think that our world is close enough to maximal goodness (perhaps because your credence in the problem of evil isn’t super close to 1) that our observations do give us at least some reason to believe that <fine-tuning + maximal goodness> has obtained in our world. But that’s a VERY different way of reasoning than the way you’ve reasoned in this post, since you are treating them as independent data.
B] Maybe you don’t share my view that value requires phenomenal consciousness - e.g. maybe you think a cool waterfall is valuable even if there’s literally zero minds to perceive it or appreciate it, but this is something you’d need to substantively justify to get your arguments off the ground. I also think this is probably wrong - Uriah Kriegel has some good arguments about why we should reject that view.
Second, I think trying to reason about really abstract, complicated matters like the probability of theism with formal Bayesian mathematics probably hurts more than it helps. Assigning numbers is always pretty arbitrary, and the complexity of mathematical modeling means we are more likely to make mistakes when updating our prior credences than we otherwise would be. As they say, a Bayesian is someone who reasons normally but moans Bayes name as they do it!
Finally, why think there’s an objective correct prior that we should all have? My impression is that most epistemic internalists tend to be subjective Bayesians, so I’m curious why you aren’t.
(As a side note, I think 4:1 is seriously underestimating the strength of arguments from evil, but that is a separate matter that I think would be really stupid to try to convince you of in a comment section here — I know you have a lot to say on this that you’ve written about elsewhere.)
Wait sorry I gave the problem of evil a Bayes factor of 10,000, not 4.
I'm not a subjective Bayesian because some priors just seem crazy. If you started out with a prior of 1 in flat earth, that would seem clearly irrational.
I was imagining that at every point where we counted a fact we included the other facts in the background. I don't think the POE undercuts the evidence that much--even in a world where God allows evil, it seems like in a decent portion he'd want agents who can make significant decisions, rather than defective disharmonious blobs. Even if you think the odds are only .1 of creating such an agent given theism, you'd still end up with a Bayes factor of 3200:1.
I agree we'd expect God to probably make the best world! The fact he doesn't is surprising. Now, he could make maximally good finely-tuned worlds, of course. But this was why I gave not a maximally good world a high Bayes factor.
I agree that consciousness is needed for value. But it's not super obvious that the universe that God would create would, at every moments, appear maximal in regards to all value facts. It might be that there are subtle connection so that a world like ours is part of some ideal total existence. Seems like this has odds of over 1/10000!
Thanks for the reply Matthew - I think it’s great that you are so willing to engage with criticism. That takes a lot of epistemic virtue; you are setting a great example for this platform. Anyways, here’s my thoughts on the substance:
1. Whoops — sorry I got the numbers mixed up, 4 was for absence of miracles. Apologies for not reading carefully enough.
2. As for the flat earth claim, I don’t think this is an argument against subjective Bayesianism. The subjective Bayesian would just say that someone who doesn’t find smth crazy would have a higher prior. If you think intuitions provide defensible evidence, then I’m not sure why someone who thinks about a strange proposition but finds it intuitive wouldn’t have justification for it before evaluating any evidence. In other words, it seems like phenomenal conservatives should like subjective Bayesianism. It doesn’t seem very conservative to be committed to the view that someone’s intuition doesn’t provide justification if we consider it crazy. (Also I think most people actually have a high prior in flat earth - that’s why most humans believed it for most of history. What is crazy would be if I didn’t believe in flat earth after considering all the evidence because soooo much disconfirms it. After updating, I have a low credence, but before I know anything about science I probably don’t)
3. Regarding the problem of evil undercutter: this seems like it just gets into a debate about the merits of various theodicies. But if we think the problem of evil isn’t solved by various theodicies (e.g. if we think it’s false that “he’d want agents who can make significant [sometimes evil] decisions”), then the undercutter remains extremely strong. I personally don’t find the free will theodicy compelling - I don’t think leeway freedom is required for free will, nor do I see why a perfect world would include it. I’d imagine that maximally good world would involve free agents who naturally have great dispositions such that they freely act morally at all times. Theres also lots of natural suffering. The point about the possibility that non-maximal goodness in this world being necessary for maximal goodness across all worlds along belongs here — it’d offer reason to think god would actually allow evil in this world. I’m not sure how much weight to give this, partially because it’s unclear why evil in this world could be causally linked to the amount of goodness in other words such that the former effects the latter - and partially because some of the arguments for a multiverse depend on claims about the existence of a good god which becomes less likely given evil, so updating here gets tricky because there’s a recursive element.
I don’t think it’s right to calculate the odds as 3200:1, because it’s not just undercutting the total odds, but rather undercutting each individual argument for theism - so fine-tuning of constants is multiplied by 0.1, anthropic arg by 0.1, etc. etc. (there are certainly some arguments that seemingly escape unscathed like the moral argument, but those tend to be pretty weak anyway - the good arguments tend to rely on a premise that god would do X to actualize value, and we observe X). If you just count it as 0.1 at the end, you are basically updating on only one datum being undercut which isn’t the claim being made here. You’d have to lower the degree of confirmation of every piece of evidence that relies on a value-actualization premise, which is most of them. If 5 arguments are undercut (not sure the exact number - I haven’t looked back at all the exact args you make, but let’s just say it’s 5, the math would look like this: 32,000*.00001=0.32
The final ratio would then be 1:3.125 (aka ~3x more likely that theism is false)
Again, maybe our credence in arguments form evil is low enough that the undercutting aspect isn’t enough to tip the balance of the scale, but I hope it’s at least clear how it might. I think the most evidentially salient question in the debate about theism is what to make of evil, because it really does have drastic implications for literally every other important argument.
I don't think it would undercut each bit of evidence because they're not independent. What I gave at least odds of .1 to--but fine, let's have it be .01 to be generous--is that God, in a world where he allows evil, would want to make agents who can understand the world. That simply entails doing the other things he did, and so it won't affect the cumulative Bayes factor by more than a factor of 100.
For instance, conditional on God having reason to allow evil and create a physical universe, the odds of him creating laws are near 1, as are the odds of him having nomologically harmonious laws.
If the arguments for theism are not independent, then why are we treating them as independent when calculating the 32,000:1 odds? It seems like we’d be treating them as discrete confirming data when it aids theism but then suddenly assimilating them when asking about undercutters.
If you want to maintain that they are *not* independent, we don’t get to calculate the odds for theism in the way you wanted because we only multiply them (without considering joint and conditional probabilities) if we’re dealing with probabilistically independent data.
There’s basically a double-bind: either a) they are dependent, in which case you’ve improperly calculated the probability of theism wrong since you just multiplied a bunch of numbers together, or b) they’re independent, so you undercut all of them.
I wasn't treating them as independent! I was assuming at each juncture that we had the other facts in the background.
I’m pretty sure you can change the prior and conditional probabilities, with support, to get any answer here. And the justifications provided don’t suggest anything about the leap from believing in God to believing in a personal God.
Yeah the Jack!
Wow, this is uncharacteristically bad analysis. Here's some of the mistakes you're making:
1) Misdirection. You're starting with the low prior of 1/1000 chance of Theism and then listing a huge number of arguments in favor and 1 against. To the untrained eye, it looks impressive that you can be "conservative" with such a low prior and still end up with 32,000:1, but the prior is meaningless when the evidence selection is skewed.
2) Even with your early caveat that all the odds are subjective and the result can't be taken too literally, you're still giving dramatically too much weight to the final output. Using Bayesian analysis for purely subjective arguments is less than worthless, because it gives us a false precision. You're casually throwing around odds of 200:1 in favor of Theism for the fact that we have consciousness with essentially no justification. Why not 20:1? Why not 2000:1?
3) You're giving the "absence of miracles" a 4:1 ratio against Theism? In other words, the fact that every unexplained phenomenon in all of history originally attributed to divinity has since been explained away is worth a 4:1 against Theism while the fact that "When you put your hand in hot water and feel pain, you pull your hand away" is worth 20:1 in favor? lol?
4) There are so many more arguments against "an all-knowing, all-powerful, all-good being" than the problem of evil. For example, divine hiddenness, inconsistent revelations. Give them each 100:1 as you casually gave many of the arguments in favor of Theism and suddenly the odds are 3.2:1 not 32,000 to 1!
5) On your Point #3: "Absent a necessary being explaining the existence of contingency, contingent stuff existing at all is surprising. It would be much simpler for nothing to exist at all." Umm, okay. What explains the existence of the being? This is like Theism 101 stuff. To assume the existence of the being is to make at least the same logical leap as to assume the existence of "contingency."
6) On your Point #3: "Taking into account the existence of a physical universe, the odds of such a thing are very near one. After all, God has no reason to create particles that don’t do anything." So why did God create a Universe that is trillions of times bigger than our planet?
7) On Point #4 regarding fine-tuning- I know you have written in depth about this argument. But I and many others find that anthropic reasoning completely negates the appearance of fine-tuning.
8) On Point #5 regarding Consciousness: This is simply a misapplication of Bayesian reasoning. The fact that we cannot predict consciousness with certainty from the outset doesn’t mean it was an improbable outcome, it just means that it wasn’t expected based on the models we had prior to discovery. Consciousness is today often considered an emergent property of complex systems.
9) On Point #6 regarding psychophysical harmony: have you heard of evolution? We already have a theory that predicts a relationship between mental and physical states being reinforced over time. And the fact that when we say, "I see a brown desk," it aligns with what’s going on in our mind simply reflects that our language evolved to describe our perceptions. This demonstrates our ability to interface with the world around us and communicate our thoughts effectively, not evidence of divine coordination.
10) On Point #7 regarding Misc: "You exist! Out of the uncountably infinite possible people who could exist, you happen to. On atheism, this is very surprising—even if someone exists, it’s very unlikely it would be you." Wow. Are you also surprised that your shoes happen to be the exact size to fit you? Do you think it's amazing that our atmosphere happens to have the right % of oxygen which we happen to breathe? You are taking our individual identity as a given and then concluding it is wildly improbable for that identity to exist. But the fact that you are you is simply a result of your surroundings. Just as we evolved to fit our environment, so too has our identity emerged. No matter "who" you are today, it was never predestined. The fact that you give a 100:1 ratio for this in favor of Theism about wraps up my criticism here.
Posting my reply:
Wow, this is uncharacteristically bad analysis :)
1) I mean, any specific hypothesis will have a low prior. So .001 isn’t some insanely low prior for rhetorical effect. My aim was to be accurate and conservative.
2) Well, it’s hard to be precise about these things. I could imagine a person who was only inclined to give it a Bayes factor of 20 (though 1 in 20 odds of consciousness randomly appearing given naturalism seems a bit nutty to me). But one could do that with all the numbers. As I said, feel free to plug in your own numbers.
3) I disagree that every miracle has been better explained by naturalism. And I included elsewhere a broadly indifferent universe operant according to natural laws, so this is just about the extra fact that there aren’t random suspensions.
4) I list the all! But I don’t find them very persuasive, because they’re just examples of bad things, and I think whatever explains why all the evil in the world exists will probably explain those too!
5) Maybe it would be helpful to read the argument on the contingency argument. The difference is that God has a plausible claim to be necessary, unlike the universe. Necessary things don’t need some outside cause.
6) Again, I counted the argument from scale separately, and gave it a generous Bayes factor even though I think it’s actually an own goal benthams.substack.com/p…
The fact I listed there was merely: a physical universe exists!
7) The anthropic reply is rejected by roughly 100% of philosophers for good reason. Exactly the same logic would say that your existence doesn’t give you good evidence that your parents didn’t use highly effective contraception because if they had you wouldn’t have existed.
8) If a fact is more expected given hypothesis A than B, even if B post hoc comes up with an explanation, then it’s evidence for A. That’s exactly how Bayes theorem works. See also the examples I gave where you retroactively declare something necessary.
9) No, I’ve never heard of it! What’s that!
You should probably, if you are going to say that psychophysical harmony is undermined by evolution, read the things that are included in every single piece ever written about psychophysical harmony explaining why it doesn’t. The one sentence version is that even though evolution explains why people behave in various ways, for that’s what natural selection selects for, it doesn’t explain why those adaptive behaviors are associated with harmonious mental states. For more detail, see philarchive.org/archive….
10) No because there’s a selection effect that makes it likely that your shoes would fit you. Most people have shoes that fit them, so you shouldn’t be surprised about that fact. Most possible people don’t exist (at least given naturalism) and for reasons I have explained at length including in published philosophy papers, your existence favors hypotheses on which there are more people. (A theory on which there is a mutiverse with 100,000 earths makes your existence 100,000 times likelier than a theory on which there’s just one earth). If you want to know more, read the linked pieces on the argument.
If you are going to be this snarky, I would advise learning something about the arguments you are making fun of.
> Exactly the same logic would say that your existence doesn’t give you good evidence that your parents didn’t use highly effective contraception because if they had you wouldn’t have existed.
“Your existence” alone does not imply anything about your parents or contraception. You need additional facts about human reproduction and the effects of contraception to make that conclusion. With that data, you know that the chances of you being the biological child to people who used contraception is low.
In contrast, your argument isn’t about which specific person you are. It’s generalized to the level of “you exist”. But you had to exist in some form somehow to do probability calculations, and naturalism predicts that some beings will exist. So pure existence (agnostic about the specific circumstances of your existence) isn’t evidence for anything in the way that.
Since the Anthropic argument was always about existence in the abstract being necessary, I don’t think this is the kind of ad hoc epicycle that reduces the persuasiveness of an argument.
Thank you for writing this it calmed me down after reading the article.
You’re very welcome 😂
1. But didn’t he count both theistic and atheist evidence?
I think 3 is just divine hiddeness and in the article he explained why he didn’t put it as highly as the problem of evil.
On contingency a necessary being wouldn’t need to be explained in the same way a contingent thing would.
This is a great analysis!
I love you but I just can't. I can't. I'm too slow, I guess.
Two mathematical points, leaving aside all the many other issues that others have raised.
1. When multiplying point estimates and ignoring uncertainty, you can get misleading results, see e.g. Sandberg, Drexler, and Ord's reevaluation of the Fermi paradox--even if point estimates for the parameters imply a strong conclusion that aliens should exist, doing some simulations just by incorporating a range in which your parameters might fall can give you an estimate that there is a 1/3 chance we are alone.
If I understand it right, what you get (and I think this will be generic, for simple assumptions about the uncertainty range of your parameters) is a power law distribution, with small numbers of universes with huge numbers of aliens, and huge numbers of universes with no aliens.
I think it would be interesting to redo your analysis along these lines, and see that the overall distribution looks like.
2. A point one of the judges, Eric Stansifer, made in the Rootclaim debate over COVID origins: if you're picking pieces of evidence at random, even if the overall evidence is evenly balanced and you pick without bias, there is a high probability that you end up with an overwhelming Bayes factor decisively in favour of one hypothesis over the other... And if your selection of evidence is a little bit biased, this is even more pronounced.
To counteract this it's best to either have a canonical factorization of your problem into component pieces so that there's no choice in what pieces of evidence to look at, or to agree in advance what pieces of evidence to look at.
I think to be convincing you should make it clear either why you think, a priori, these are exactly the things we should agree to look at to decide this question, or to present some model where you factor the question of "does God exist" into a calculation where these are the only parameters.
100% agreed, both very relevant concerns regardless of whether you agree with the content of the post
"The numbers are a bit made up at times. But reasoning with made up numbers is often better than reasoning with no numbers at all"
Sure I guess, if you also acknowledge the limitations and potential for error when the underlying data is unreliable. Pulling numbers out of a hat is just a recipe for false precision.
I don't think my essay could be described as false precision when I say you shouldn't take the numbers very seriously.
Right, that's the problem! What's the point of this fancy pants analysis when the inputs are shit.
The most likely 'probability' in these estimates is that the universe just exists. Nothing else is needed to explain it. This would be more likely than theism by your logic.
> God might be the very simplest entity: a mind or agent totally without limit.
That's not simple. To take the example of the sequences, "a witch did it" sounds simple because we've cached the incredibly complicated concept of a witch (all the complexities of a human plus all the complexities of "magic" (something so complicated it doesn't even exist) plus all the complexities of a human-magic interface) into one word. Same with god.
Minds and agents are incredibly complicated. Mind bogglingly complicated. And they don't get simpler as they approach infinite mindfulness/agency. The steel man I can give of a limitless mind is if there's something like panpsychism and the whole universe creates one mind (in the same way that a whole brain creates one mind), but then I'd say, call that something other than "god" or you're gonna create a whole lot of confusion.
I too have read that Eliezer essay :).
It's not enough that the English sentence describing a thing is simple: it has to be ontologically simple. You say that minds and agents are complex, but if dualism is true, then that isn't so--on dualism minds are fundamental and irreducible. And I think there are good arguments for dualism https://benthams.substack.com/p/marys-room-refutes-physicalism. At the very least one should have credence in dualism well above 1%.
If minds are simple, why is the brain so complex? What do we need 86 billion neurons and trillions of synapses for?
Dualism, and in particular arguments for it like Mary's Room, may suggest that certain features of mental life like qualia might be fundamental, but they don't automatically imply that beliefs, attitudes and other psychological concepts that you want to build into theism are fundamental.
I suspect you'll reply that even if Mary's Room or the like don't do the heavy lifting of showing that these concepts are fundamental, they do indicate the truth of dualism generally, and that you don't think conditional on dualism that those specific things being fundamental is all that crazy (regardless of whether you actually have a strong argument for their fundamentaliy or not). But I think it's hard to make progress against moves like this. You have to make sure that you're being similarly generous to all of naturalism's possible "not that crazy" purely speculative sub-hypotheses, of which there are innumerably many.
I agree there are some further questions, but even with these I think its prior should be at least .001.
I think that either the prior on theism should be something like one in a gazillion, or else it should be as high as you say (because of under-explored epistemic unknowns) but the likelihood of naturalism on fine-tuning/psychophysical harmony/whatever should also be relatively high (because of under-explored epistemic unknowns). Of course, I would also add on to this that I think there are separate objections that largely defuse these arguments so I don't have to rely on the unknowns, but still.
Even under that, just because electromagnetism is fundamental doesn't mean the universe is a supercomputer, or that a supercomputer is simple. I think that even with consciousness as fundamental, minds as we know them are a unique, complicated arrangement of the fundamentals, just like computers are unique arrangements of the electromagnetic fundamentals. So maybe with those fundamentals there is an unbounded supercomputer and an unbounded mind, but I don't think it would look like anything we know of as a computer or mind, and I don't think the mind version would look like what most people would call god
If electromagnetism is simple then if there was something that was just pure limitless electromagnetism, then it would be pretty simple! And in any case, minds are more joint-carving than electromagnetism, which is just a random physical force.
Even if minds we know are complicated, that doesn't mean all possible minds are. As an analogy, life we know is very complicated--much too complex to form by abiogenesis--but that doesn't mean abiogenesis is impossible!
How could we tell whether qualia/consciousness is simpler than electromagnetism? If they are equally complex, should we have a 1:1000 prior on a pure limitless electromagnetic "thingy" (for lack of a better word) existing?
If a mind was limitless, wouldn't all minds be part of it? Anything that would separate our minds from the limitless one would be limits, which it can't have
It's just a single thing that's limitless qua mind, so nothing binds or constrains it. But that won't subsume other minds because it's singular.
Electromagnetism is only one part of the physical world rather than a whole class of reality.
I'm not following 😅
On the separate class of reality point, why couldn't we also say that electromagnetism is a class of reality with parts that interact with physical world in predictable ways? Such as how the qualia class of reality interacts in predictable ways with the physical world (although we have a better understanding of electromagnetism/we're better at predicting it). For example heroin interacts with our brain which interfaces with the qualia realm to produce pleasure, and on the other direction, the mind class interfaces with our brain to produce signals to control our musculature, and so on for the various inputs and outputs of minds
Mary's Room is a bad argument for dualism because it's question begging. But it doesn't stop there. What would zombie Mary say when she sees red for the first time? If she's surprised by the redness of red, then zombies aren't possible. If she isn't, then the original Mary does not experience any nonphysical red. Believing the Mary's Room argument destroys the other conceptual tools that are meant to be pro dualism without the argument giving independent justification for being a dualist.
>A physical world is contingent—it could have been otherwise! Contingent stuff existing at all is somewhat puzzling on atheism. Absent a necessary being explaining the existence of contingency, contingent stuff existing at all is surprising.
Calling something "necessary" does no explanatory work in explaining its existence. Or if it does, the atheist can just claim the physical world has that property, or that God lacks that property. And by the nature of the methodological assumptions behind the contingent/necessary dichotomy, there will be no resolution either way. I feel like this is a problem that keeps reappearing in these theistic considerations so the ultra giga bayes factor you end up with at the end should be heavily in favor of there being no fact of the matter/this being a meaningless discourse with too many faulty assumptions to evaluate probabilities.
Plus, even given an uncaused causer, I would think the original unmoved mover is more likely to be as a grain of sand is to a pearl, than as any normal concept of god is to the universe
Why?
Occams razor
Can anyone tell me how this is any different from saying 'There is a lot of stuff in the world that I can't make sense of, so God must have done it.'? It is essentially the same as creationists marvelling at all the wondrous features of bananas.
The banana bit is an argument from biological design, which according to Bentham "holds no force" (stated near end).
He's granting you evolution, natural selection, etc, but arguing that even with such things being true, God still exists.
Well it is just replacing the banana with the cosmos, natural laws, consciousness, psychophysical harmony and evil i.e. just a long chain of arguments from ignorance.
Distinguish three hypotheses:
1: an intelligent creator of the universe exists, and he is all-powerful, all-knowing, and all-good.
2: an intelligent creator of the universe exists, but he lacks one or more of those properties.
3: the universe was not intelligently created.
You focus on 1 and 3. But you ignore 2. I consider 2 a lot more likely than 1.
So even if I spot you that there is a cosmic appearance of design (consciousness, psychophysical harmony, nomological harmony, etc.) it seems like all it gets us is an intelligent creator, not the "triple omni" version of theism you seem to want.
Moreover, we can distinguish eight different conceptions of a creator, depending on which of these three properties he has or lacks (1: all-powerful, all-knowing, and all-good... 2: all-powerful, all-knowing, but not all-good... etc.).
Arguably, the Christian one of those—the one where he has all three omni properties—is the least plausible of the eight. At best, your arguments tell in favor of the disjunction of these eight options.
The conclusion of these types of analyses seems to hinge a lot on how independent you think the various arguments / lines of evidence are. You say that atheism has only one significant piece of evidence in its favor, whereas theism has several. Even though for atheism you mention the problem of suffering, divine hiddenness, religious confusion, and absence of miracles, you say that really these are not independent data points, because we only to need to have an explanation for one thing, a blind and indifferent world, and once we have that, then all of these observations are explained in one fell swoop.
Ok, fair enough, but I don't see why the theistic arguments aren't similarly collapsible in this way. Scott Alexander had a post earlier this year about how Tegmark's Mathematical Universe Hypothesis defeats many of the arguments for God's existence. Of the ones you mentioned here, I think MUH could explain fine-tuning, nomological harmony, and the anthropic argument. If MUH is true, then it changes your conclusion by several orders of magnitude because it explains multiple of your theistic pieces of evidence.
Basically, conditional on knowing that there is a good atheistic explanation for one of the theistic arguments, that ought to at least somewhat impact your credence in the others as well, because it could be the same underlying theory that explains multiple of them. For this reason, I think it is a mistake to simply multiply the odds together as you have done here.
> So in deciding upon the prior you should ask: if you were in a dark room with no access to the world, and hadn’t considered your own existence (perhaps you were tripping on a powerful drug that eliminates your sense of self), how likely would you think it is that an omnipotent, omniscient, and omnibenevolent God exists.
If I were in a dark room with no access to the world, I’m not sure I could reasonably come to the conclusion that omnibenevolence was part of the equation that got me there.
If I then were let out into the world to find out I’d certainly die some day and very likely in agony, as would all other living things that were for whatever reason made capable of agony, I think omnibenevolence would be even further from my mind.
Omnibenevolence has never been a good explanation for suffering, let alone for astronomical amounts of it.
You're taking into account the updates but you're supposed to focus just on the prior.
> This isn’t surprising given theism—consciousness is the source of most value
What is surprising is that physical brain defects can nullify the consciousness and value of the soul. It's probably because souls don't comport well with a broader picture of psychology, neuroscience, etc.
>One of the most surprising facts about our world is: experience exists.
It's no more surprising than computers or engines or nukes existing. Which is weird, but not in a difference in kind way.