The Moral Knowledge Argument Sucks

No atheist should be convinced by it

Dustin Crummett and Philip Swenson—philosophers I often find myself agreeing with—have a book chapter where they argue that there is a strong argument from moral knowledge to the existence of God. Fellow substackers Amos Wollen and Bentham’s Bulldog—who I also agree with on many issues—find this argument compelling too. So it’s sort of in the cards that I should think it’s a strong argument. But I don’t. Actually, I think it’s a decidedly terrible argument.

But What Is the Argument?

I hear you asking. Very briefly the idea is this: We have moral knowledge because [insert some arguments for moral realism]. Now, given atheism it’s very surprising that we would have this knowledge. Firstly, the fact that we make the moral judgements that we do isn’t very surprising on atheism/naturalism, as they’re basically what we’d expect given evolution, regardless of whether moral realism were true or not. At the same time, plausibly nothing about the way we formed these judgements will have been affected in any way by what the moral facts are, since moral facts don’t move around atoms, and generally don’t have causal powers. That means that if the moral facts were different, we would have exactly the same moral beliefs, meaning our beliefs don’t track the moral facts. This is pretty bad news, and basically completely undercuts our moral beliefs.

But if theism is true, it’s very probable that we would have accurate moral beliefs because God would want us to, etc., etc. This means that the fact that we have moral knowledge is very improbable given atheism, and very probable given theism, meaning we get strong evidence for theism. Pretty straightforward!

I Wonder What Time it Is

Now a short irrelevant intermission, where I tell you an unrelated story about my watch. I know that the time is *checks watch* 12:58, and I’m very confident about this. But a clock-skeptic now comes up to me and tells me that my watch actually doesn’t have any batteries and is stopped. He knows this because the store I bought it from sells watches without batteries already in them (for some reason), and I never put one in myself. I know this fact about the store, and I also remember pretty distinctly never having put batteries in.

But luckily I see the skeptic’s challenge for what it is: a great argument for watch-God. You see, it’s super obvious that the time is… well, by now it’s probably 12:59. Anyways, it’s super obvious that this is the time—in fact I know this is the time. And you see, if the clock-skeptic is right, that would mean that my temporal beliefs were not properly tracking the time, and I would not be justified in thinking it’s 12:59. If watch-theism is true, on the other hand, watch-God would want me to have correct temporal beliefs, meaning he would have ordained it that my watch somehow did have a battery, or that I had correct temporal intuitions, or whatever. This means that my temporal knowledge gives me strong evidence in favor of watch-theism.

Hmm, what a strange unrelated story, I wonder whether it has any significance for the rest of the post…

What’s Wrong With the Argument

HAHA, I tricked you! This story isn’t completely unrelated—in fact it’s VERY related! It might not have been clear to any but the most extraordinarily sharp of you, but I was actually making a parody argument of the argument from moral knowledge. It looks to me like the inference I make in the watch case is relevantly analogous to the one I make in the moral knowledge argument: I think I know some fact, but am presented with some undercutting defeater for my knowledge. In response to this, I retain my belief in the fact, and Moorean shift to some theory where the defeater isn’t a defeater.

If you’ve read my post on Moorean shifts, you might know what I think the problem with this is. Basically, I don’t think you can Moorean shift away from an undercutting defeater—and the move made in the argument is essentially a Moorean shift. The reason for this is that an undercutting defeater explains why the reason you believe P isn’t actually reason to believe P, given some background assumptions. But when Moorean shifting to retain belief in P, you do this on the basis of the strength of your reason to believe P, and since it’s exactly that which has been removed, this move is no longer valid.

Consider for illustration the position of the moral-realist-atheist who hears this argument, and is convinced that we can only have moral knowledge on theism. This means that the reasons they previously believed they had for their moral beliefs are now shown to have been faulty all along, given their other commitments. That completely removes their reason for believing moral realism. It is of course true that, if they accept theism, they will no longer have this defeater, but they antecedently judge theism not to be plausible, and they have no positive reason to revise their beliefs.

This all means that the intellectual cost of rejecting moral realism is nil, since they no longer have any reason to believe it. On the other hand, the cost of retaining moral realism and accepting theism is not nothing, since they presumably have some reasons for being atheists, which they will have to overcome. So it just looks completely obvious to me that it’d be wholly irrational for an atheist to become a theist on the basis of the argument from moral knowledge. At best they should become moral anti-realists, if they accept the premise that moral knowledge is impossible under atheism.

I think the fact that this argument is considered convincing speaks to a sort of naive way some people have of reasoning about moral realism—and other “Moorean” type beliefs in general, that are super intuitive on the face of it. In his his paper on evolutionary debunking and axiarchism, Brian Cutter (a philosopher I otherwise respect) remarks “moral realism is immensely plausible—so plausible that (for me, anyway) rejecting it isn’t really a live option.” While I myself lean towards some sort of moral realism, and think phenomenal conservatism is probably correct, these sorts of remarks just seem to ignore the fact that you can have defeaters for your beliefs. Something can seem obvious, but it can turn out that you thought this for some reason completely unrelated to its truth—and this is exactly what debunking arguments are supposed to show for moral realism. This kind of sloppy thinking just substantiates claims that phenomenal conservatism is used as a crutch to hold any belief you find convenient. I think people should be better at distinguishing when you have really good reason to hold a belief in the face of skepticism, and when you just really don’t want to let go of it.

Might There Still Be Hope?

What I have said here is certainly not very optimistic about the prospects of this type of argument. But I do think there might be one domain where it may have some strength: beliefs that cannot rationally be rejected. What do I mean by this? Basically beliefs that, if rejected, undermine the reason to reject them. One example could be a principle of ontological parsimony. Consider a parallel debunking argument against such a principle:

We should expect that we would believe more parsimonious theories to be more likely, given evolution, since it takes more brainpower to process less parsimonious theories. But the explanation for our belief has no relation to whether such theories are actually more likely, and so we have a defeater for this belief.

The problem is that if we don’t accept that theories that are more parsimonious are more likely, all else being equal, then the argument given no longer has any force. After all, we might accept that we have no reason to think our belief tracks the truth, but then just stipulate that there is a mysterious force that makes us track the truth of it anyways, but which is otherwise undetectable. If we don’t think that more parsimonious theories are less likely, then we have no reason to reject this theory. In that case, though, the argument against the principle of parsimony is no longer successful. So the argument defeats itself.

This of course requires that we actually can give a sufficiently strong debunking argument given atheism.¹ On top of this, it requires that the belief debunked actually can’t be rationally rejected. What I have said here is clearly not sufficient to motivate this, but this is the general structure a successful “argument from XX knowledge” would need to look like, and I think something along these lines might work.

Bentham’s Bulldog likewise points out that the argument isn’t just about moral knowledge (for example here). While I think some of the domains he mentions could potentially be the basis of a successful argument, we have to be strict and keep in mind that only those beliefs that cannot be rationally rejected ought to count. For example, while it’s intuitively plausible that we have modal knowledge, I don’t think there’s anything obviously self-defeating about rejecting it (at least rejecting the modal knowledge you won’t have on atheism). You can’t assume that knowledge in some domain is a theory-neutral datapoint that you can hold fixed; chances are this style of argument will at best show that an atheist should be a skeptic in that domain, not that they should be a theist to retain their belief.

Also, I don’t think it’s sufficient that a fact cannot move around atoms. For example, the fact that 2+2=4, rather than 3, presumably can’t move around atoms. Yet, if Grok (circa 40000 BC) saw two lions enter a cave, followed by two more lions, and then saw three leave, he had better hope he had true arithmetic beliefs. Likewise, if he thought that all lions are dangerous, and that Leo the Lion in front of him is a lion, then it would be pretty good for him to be able to figure out logical entailments. So I think it’s much harder to make a successful debunking argument than simply considering which facts have causal powers and which don’t.²

And at the very least, the argument should be renamed something other than the argument from moral knowledge, as moral knowledge is perhaps the weakest possible “datum” to use as a starting point for the argument.

1. Obviously this cannot be done for all possible atheist theories. So what it should really say is “atheism of a sort that isn’t much less antecedently plausible than theism” or something.

2. Generally I suspect that even the most hardcore naturalist theories will get knowledge of these types of deductive domains. One reason to think this is that the basic inference-rules will be ensured by examples like the above, and any large deduction will simply consist in repeated application of the basic inference rules. It also seems like having this capacity to realize entailments will transfer when you consider systems with axioms that don’t correspond to everyday experience (like non-Euclidean geometry).