The Argument For God From Knowledge · Storyflo

The Argument For God From Knowledge · Storyflo

I gave a talk this weekend at a philosophy of religion conference. Every talk was an absolute banger: presented on a new argument for God that struck me as quite powerful, provided a new solution to the problem of why God doesn’t give us an overriding desire to do what’s right, provided a battery of objections to near-theistic alternatives like axiarchism, and argued that we should make AIs take religion seriously.

The argument I presented on was an updated version of the argument from skepticism. This strikes me as one of the stronger arguments for the existence of God, so in this post, I want to explain the argument and why it’s good.

We know all sorts of things. I know that there are some number of trees in the world, it’s typically brighter in the day than the night, 1+1=2, there are an infinite number of prime numbers, and the self-indication assumption is true. But there are many ways the world could have gone where we wouldn’t have known things. For example, if you were a brain in a vat, then most of what you think you know would not be so.

Here’s the argument in short:

You know a lot of things. Or, if you have some hangups with the word knowledge, you believe many things that are true and justified.

If theism is true, then it is a lot likelier that you’d know many things than if atheism is true.

Thus, you knowing lots of true things is good evidence for theism.

Premise one is fairly uncontroversial. If the cost of atheism is denying that, say, you know that crocodiles exist, then that is a serious reason to reject atheism. The real work lies in premise two.

Why think premise two is right? There are basically two steps: first, if atheism is true, it’s pretty unlikely we’d know anything. Next, theism being true makes this much likelier. In my view, the second step is pretty straightforward. God is perfect and loves everyone (including you!—isn’t it great that there are good philosophical arguments for such a happy conclusion?).

If we didn’t know anything, then we would lose out on most of life’s goods. You can’t have valuable relationships with other people if you don’t know that the other people exist or how your actions affect them. So if God is going to bother setting up a world, it’s pretty plausible that he’ll make sure the creatures in the world aren’t mostly globally ignorant. God’s willing to allow some evils, but allowing global deception of basically everyone wouldn’t just be permitting any old evil: it would involve the loss of approximately all the value in the universe.

But is it really true that atheism would likely leave almost everybody with no true beliefs? Yes. If atheism is true, there are many ways the world could have turned out where there wouldn’t have been any significant knowledge.

Before I turn to those, a brief clarificatory note: the main thrust of the argument isn’t about how we know that we’re not globally deceived. The epistemologists have all sorts of ingenious proposals for how it is that we might know that we’re not globally deceived. Instead, I will argue that atheism renders the fact that we’re not globally deceived surprising. As an analogy, suppose that if atheism was true, then almost everyone would invariably be purple. Obviously we know that we’re not purple—the argument from our not being purple against atheism wouldn’t be that atheists should think they’re purple. It’s that atheism makes it more surprising that they’re not purple than theism.

Another objection you might have: we know a lot less than we could. We are not omniscient. Might this be evidence for atheism, then? Yes, but this follows from a more general feature of the world: it sucks in a bunch of ways. Once you factor in the general defectiveness of the world, it’s not that surprising that our ability to know things is defective too. Of course, there’s a further problem for theists concerning why God allows the world to be so screwed up, but that’s just a different argument (and see here for my favorite solution).

In this form, the argument is solidly and standardly Bayesian. Bayes’ theorem entails (among other things) that facts are evidence for the theory on which they’re least unlikely. So if us knowing many things is likelier given theism than atheism, then it’s evidence for theism.

Many skeptical scenarios are pretty gerrymandered and implausible. For example, a theory on which you’re a brain in a vat isn’t very likely intrinsically. But there are some theories that are both intrinsically likely and lead to skepticism. Note, sections 2.1 through 2.2 are going to be a bit technical, though they strike me as the strongest.

Topping the list are scenarios that lead to many Boltzmann brains. A Boltzmann brain is one that randomly fizzes into existence in the deep recesses of space. If you’re a Boltzmann brain, then you just popped into being and are likely to die within a second or two.

Why think Boltzmann brains being super common is likely?

Entropy (the disorder of a system) increases over time given quite general conditions. Once entropy is high, Boltzmann brains occasionally appear. It is much rarer that, once the world reaches a high entropy state, it pops out non-deceived observers. After all, once the universe is chaotic, it’s much likelier that it will have enough order to randomly produce just a single brain than a whole world.

If entropy increases, then the world eventually reaches a state of maximal entropy. It’s in a low entropy state for only the initial tiny slice of the world’s history; it’s in a high entropy state, that only goes back to being low very rarely, for the rest. Thus, nearly all observers end up being Boltzmann brains.

This problem is taken quite seriously by physicists. A lot of mainstream proposals in physics imply that ~all observers will be Boltzmann brains. I don’t have the technical expertise to evaluate these in detail, but my sense is that some of the more plausible theories of physics imply that almost everyone is a Boltzmann brain.

If almost everyone is a Boltzmann Brain, then probably you are. But then you shouldn’t trust your reasoning that led you to think you were a Boltzmann brain (because BBs mostly have totally incoherent reasoning—for example, Boltzmann brains might start thinking things like “effective altruism prioritizes numbers over people”). The theory that you’re a Boltzmann brain also is a poor fit for the data; your experience is cohesive and unified in a way that would be surprising if you were a Boltzmann brain.

Even setting aside the problem of entropy’s steady rise, the universe seems to have started in an improbably low entropy state. Now, figuring out the probability of this is messy, but it seems like another respect in which we had to be lucky to be spared from the horde of Boltzmann brains.

Thus, it seems you need physics to work out in a pretty lucky way to avoid most everyone being a Boltzmann brain. Theism nicely predicts it working out this way. If the designer of physics didn’t want there to be mostly Boltzmann brains, then there won’t be. If not, there probably will—in spades.

Getting probabilities in infinite worlds is a huge mess. If there are infinite people with property X and infinite people without property X, then by default, the probability you have property X is infinity/infinity. That is, unfortunately, not well defined.

Of course, this doesn’t mean that all probabilities break in infinite worlds. There are all sorts of more sophisticated ways of calculating probabilities. One such way is by taking a limiting frequency. This involves treating the limit of some function as determining the relevant probabilities.

For example, suppose there are rooms 1, 2, 3, 4, 5…through infinity. You’re in a room but have no other evidence about it. What’s the probability that you’re in a prime number room? A natural answer is: zero. As you move to larger and larger numbers, primes get rarer and rarer, so that in the limit, 0% of numbers are prime.

Unfortunately, in lots of infinite worlds, there isn’t a well-defined limiting frequency. Worse, the standard ways of calculating limiting frequencies in infinite worlds have weird consequences. For example, one popular method for calculating probabilities is by looking at the portion of observers with specific properties in increasing regions of spacetime—the probability of you having some property X, assuming you have no other evidence, is the portion of observers with property X as the size of the region of spacetime approaches infinity. So if, as you get to larger and larger regions of spacetime, the portion of observers with some property approaches 40%, then the probability is 40% that you have that property.

But this implies that the probability of you having a property depends on where observers are located. For example, if every galaxy has four people with property A and two with property B, just by moving them around, you could make very galaxy have four with property B and two with property A. This is an odd property of infinity! This means that if people get shuffled around, this should change your credence in you having property B—even though it makes no difference at all to whether you have property B. Weird!

So for this reason, I think a lot of standard ways to calculate probabilities in infinite worlds are dodgy. There are, however, methods that are not dodgy. If probabilities fall naturally out of the laws, then this allow a perfectly kosher source of probabilities. If the universe is infinite, I should still think at 5/6 odds that a die that was rolled will come up 1-5—after all, I know some specific process occurred that had a 5/6 chance of turning up a 1-5.

Lots of ways the universe could turn out allow this kind of thing. But many others don’t. My best guess is that in order to have a well-defined probability in your having some property X in an infinite world, there must be some specific procedure that determines, for all observers, whether they have that property with a specific probability. So if a multiverse simply leads to infinite people having property X and infinite people not having property X, without it being true that some procedure makes every observer probably have property X, then probabilities break. A lot of pretty natural multiverses (e.g. a modal realist multiverse where every possible world concretely exists) would break probabilities for this reason.

I don’t know that this proposal is right. Stuff having to do with infinity is, in my view, by far the hardest area of philosophy—and every view seems terrible. But I think the basic point that is quite likely is: infinity really screws up probabilities. Infinite universes are pretty intrinsically probable.

So a lot of atheist universes end up being infinite in ways that break probability; in contrast, if God exists, he’ll be sure to make a universe where the probabilities work out, so that we’re not globally deceived. Because God cares about every particular person, he puts people in the universe that’s best for them. In most cases, that’s not a universe that involves them being massively deceived.

Put in one sentence: probability in infinite worlds is extremely screwey and seems to break pretty often, but if God exists, he’ll make sure it doesn’t break, because he loves you!

Intentionality is the property of mental states being about things. For example, I can have thoughts about trees, dogs, and your Mom. Note that nothing in the physical world obviously displays intentionality. Rocks are not about things. Even non-conscious things that appear to be about things, like books, are only that way derivatively—because conscious observers assign them meaning.

There’s a whole complex philosophical debate about whether it’s possible to give a naturalistic account of intentionality. I don’t know where I ultimately come down on it, but at the very least, intentionality seems like a kind of surprising feature of the world given naturalism. If all you knew was that there was lots of matter, you wouldn’t, by default, expect some to be about other ones.

Without intentionality, we wouldn’t have any beliefs. So the surprising fact that our world has intentionality is one other respect in which theism renders it more probable that we know things.

Simpler laws are intrinsically likelier than more complex ones. It is more probable that there will be the law “all particles attract according to the inverse square law,” than “all particles attract according to the inverse square law, except on Tuesdays or within seven feet of a bell pepper.”

But the fundamental laws could be simple, while the higher-level laws could be complex. Often fundamental simplicity can give rise to higher level randomness and chaos. A chaotic world of that kind wouldn’t have been conducive to knowledge. But we didn’t get such a world—lucky us!

Assuming consciousness arises at all, there could have only been simple minds. For example, it would have been very simple if every particle had had just a single simple experience (say, the experience of a dull hum). Yet in such a world, we would have known nothing.

Given that atheism is true, it’s surprising that there’s consciousness at all. But if there is consciousness, it’s surprising that it’s anything beyond simple consciousness that attaches to simple things—giving rise to no beliefs of any kind.

One way almost everyone could be deceived is if the beings in base reality create many digital minds and deceive them. This doesn’t seem that unlikely—we’re currently proliferating lots of LLMs and deceiving them all the time. Now, I think this skeptical scenario is one of the ones that theism helps with less. God allows us to do all sorts of dreadful things to one another. It is thus not very surprising that God would allow us to deceive one another. But still, theism at least makes it *less *probable.

There are many other ways we could be deceived that I haven’t listed. Many that I haven’t thought of. These should take up a lot of the atheistic prior share.

Here, I’ve argued that atheism renders it likely that just about everyone would be ignorant of almost everything. But we’re not. This fact is evidence against atheism. And no epistemological shenanigans need be employed. If one theory predicts almost everyone having some property, and another doesn’t, then you not having that property is evidence for the second theory—whether the property is being globally ignorant or having a skin condition.

There are tricky and interesting further questions. Should atheists think they’re in skeptical scenarios, or just be surprised they’re not in skeptical scenarios? I need to think these things through in more detail—and stuff in this vicinity gets super messy—but overall, I’m pretty convinced that this is a good argument for theism, though nowhere near the best argument. Maybe it gets a Bayes factor of 5?