I think there are two serious problems with this argument.
Arguments like this don't actually compare theism with atheism given some datum they compare atheism or naturalism simpliciter with a very specific version of theism: one on which God is a perfectly good being with unconstrained power, knowledge, and a particular set of intentions toward creation. So the comparison isn't A vs T, it's really A vs (T & tri-omni & a-set-of-intentions), where all the heavy lifting is being done by the add-ons rather than by bare theism itself. But once you see that, the move loses much of its force because you can attach analogous add-ons to atheism or naturalism just as easily. Nagel's teleological naturalism is exactly this: atheism plus the thesis that mind and value are fundamental features of reality with an intrinsic tendency toward realization. Aristotelian naturalism is another: naturalism plus a robust account of natural teleology grounded in what kind of thing we are. Both of these predict moral knowledge and agency without God. So if the datum is evidence for anything, it's not evidence for theism over atheism as such for that you'd need to show that bare theism, once updated on the datum, yields a higher posterior than bare atheism does. What the argument actually shows, at best, is that the datum is evidence for versions of theism that already have built into them the intentions and values that make the datum expected.
The second problem concerns the first premise of the argument itself. If we grant that there are objective moral facts, then there are also objective facts about evil: that suffering is bad, that cruelty is wrong, that vast amounts of undeserved harm occur, that a vast amount of evil appears to be gratuitous. These facts are in significant tension with the specific version of theism being invoked here, namely one featuring an omnipotent, omniscient, and perfectly good creator. So even if we agree that P(moral knowledge & agency & facts | atheism) is very low, it's not obvious why P(moral knowledge & agency & facts | theism-with-omni-properties) wouldn't be even lower (it appears to be).
Classical theism just is the hypothesis of a perfectly good, rational creator. Those aren’t ad hoc add-ons.
By contrast, your enriched forms of naturalisms build in teleology and value to do the same explanatory work, just less unified, and in a very blatantly ad hoc manner.
On evil: of course it’s part of the data. But then the question is comparative. You’d need to show it outweighs the explanatory advantage theism has re: moral knowledge and agency.
I think the right move for you is to grant that moral knowledge and agency are evidence for theism but hold that theism is still unlikely overall.
An ad hoc add-on is one invoked specifically to rescue a hypothesis from disconfirming data, with no independent motivation. But Nagel and Aristotelian naturalists aren't responding to the moral knowledge and agenecy datum at all they developed their views on independent grounds. The enriched naturalisms weren't constructed to save atheism from your argument. So whatever their other flaws, ad hoc they are not.
You might want to say they are less unified than classical theism, and that this affects their prior. That may be a legitimate point but it is a completely different argument from the one being made in your original post. The original argument is that moral knowledge and agency constitute evidence for theism over atheism. If the reply is now that classical theism has a better prior than enriched naturalism due to greater unity or parsimony (and thus greater posteriori) we have shifted the dialectic entirely. My critique was directed at the original argument, and that stands on its own terms independent of this.
You say I need to show that evil outweighs the explanatory advantage theism has regarding moral knowledge and agency. But I don't think that's right. You are the one asserting premise four that on theism, moral knowledge and agency are to be expected. But what theism must predict is not just moral knowledge and agency in isolation. It includes objective moral facts and that landscape includes evil. So the burden falls on you.
You say "I think the right move for you is to grant that moral knowledge and agency are evidence for theism but hold that theism is still unlikely overall." but I do not think this is right. It is not evidence for theism but for specific conception of theism and that is not particular interesting. For the moral data is evidence for classical theism over bare atheism in roughly the same way that the existence of cheese is evidence for Cheese Theism over bare atheism both are data in which the enriched part not thei theist part is predicting. The question of whether God exists is supporter is not settled by noting that some elaborated version of theism (whether it is a classical tradition or not) predicts.
You’re still mislocating where the explanatory work is being done.
The issue isn’t whether Nagel or Aristotelian naturalists introduced their views to answer this argument. It’s that once you enrich naturalism with fundamental value and teleology, you’ve moved to a view on which the datum is no longer surprising. That’s fine — but then the comparison is no longer “theism vs atheism,” it’s between competing enriched worldviews. And at that point, unity and theoretical economy are directly relevant to the evidential question, not a shift in dialectic.
On evil: of course the total data matters. But that cuts both ways. You don’t get to treat evil as disconfirming theism without also asking how well atheism accommodates objective moral evil as such. The comparison remains holistic.
Finally, the “cheese theism” analogy just misses the structure of the argument. The claim is not that some arbitrarily enriched theism predicts the data, but that a well-motivated, independently supported conception — classical theism — renders it more expected. That’s exactly how evidential arguments are supposed to work.
At this point the positions are clear, so I’ll leave it there.
On the first point: I want to be precise about what the cheese analogy is actually showing, because I think you're not getting its force correctly. The point is about what the data is evidence for. Suppose God exists. The datum of moral knowledge and agency is then evidence about what kind of God exists one with the goodness and revelatory intentions that classical theism attributes to him. It confirms those attributes given theism. But that is different from being evidence that theism is true in the first place. Analogously, the datum of cheese confirms that this God has cheese-desiring intentions. That is evidence about God's nature given his existence, not evidence that God exists. The parallel holds regardless of how independently or well motivated the enriched hypothesis is, the evidence is for the enriched aspect not the bare-bone parts.
For D (moral data) to be evidence for theism, the proportion of theism-space that predicts D well would need to be significantly larger than the proportion of atheism-space that predicts D. That is whats relevant, not whether some specific corner of a vast theistic possibility space is well and independently motivated, that is orthogonal.
If the argument shifts to comparing classical theism against enriched naturalisms on grounds of theoretical unity and parsimony, that may well be a fruitful discussion but not relevant, its conclusion would be that classical theism is a better worldview than Nagelian/Aristotle naturalism, not that moral knowledge and agency are evidence for theism over atheism, the former can be true while the latter false.
Point about evil I agree that the comparison is holisticbut that does not discharge the burden I identified you have by asserting the forth premise.
First objection: The concept of God here just is the classical one — a perfect being. So, it’s not contrived, unusual, or unmotivated. Additionally, there are multiple independent lines of argument for the existence of God so conceived. I’ve discussed a number of them in other posts.
Second objection: This is just the problem of evil — and yes, any perfect being theist needs some kind of response to it. I’ve outlined what I take to be the right way of thinking about that problem here: https://reflectionsonwhatmatters.substack.com/p/pondering-the-problem-of-evil?utm_campaign=post-expanded-share&utm_medium=web. Setting aside whether you think I’m right about that, theists generally don’t take the problem of evil to be decisive (hence their theism), so shouldn’t be expected to reject this argument because of it.
The question isn't whether classical theism is an old tradition or well-motivated independently but whether the moral data itself gives us reason to prefer theism over atheism, and that requires showing that bare theism predicts the datum better than bare atheism. That case hasn't been made. The moral evidence would be better regarded as evidence of God's nature and goals (that the classical thesis associates with him) given that he exists, than as evidence of a God simpliciter.
For the second point the argument. The point is not whether the theist takes EPOE to be decisive we agree that they dont what is at issue is whether evil does not disconfirm this classical theistic thesis to such a degree that it balances out the support that moral fact & agency & knowledge gives. The argument needs an additional premise: that whatever disconfirmation evil imposes (important to remember evil is part of the data invoked) on T* is outweighed by the confirmation provided by moral knowledge and agency to such an extent that the argument remain powerful.
I worry that your second paragraph might be an instance of lumping together things that need not be lumped. Sure, it might the case that the evidential problem of evil disconfirms theism to such a degree that it cancels out the boost it got from the existence of moral facts. But like, those are two separate arguments. We can coherently evaluate whether theism gets confirmed by the existence of moral knowledge and separately evaluate whether it gets disconfirmed by the existence of evil. In other words, we can start by assessing P(mknowledge|theism) vs P(mknowledge|atheism) and then put the existence of moral knowledge in the background (after updating on it) to evaluate P(evil|mknowledge&theism) vs P(evil|mknowledge&atheism). Or we could do it in reverse and start by evaluating P(evil|theism) vs P(evil|atheism) and then move on to compare P(mknowledge|theism&evil) vs P(mknowledge|atheism&evil). Or we could even do a c-inductive argument where our modest goal is to evaluate just P(mknowledge|theism) vs P(mknowledge|atheism) and leave the problem of evil to people who specialize on that area. In other words, I think there are many ways to go about it. It could be that once moral knowledge is accounted for, the existence of evil provides evidence that cancels it out, but that would be a separate argument that doesn't need to trouble this one. In general, the ability to break big chunks of data into smaller ones is what gives us the ability to discuss separate arguments for or against God without having to evaluate literally everything all at once.
The argument doesn't just happen to invoke moral knowledge it invokes a range of moral data that includes objective evil facts as part of the same package. So evil isn't a separate independent datum that comes from a different argument it is entailed by the first premise the argument.
You can evaluate P(d1|T) and P(d2|T) separately. But that works when d1 and d2 are independent data points. Here d1 (moral knowledge) and d2 (evil facts) are not independent. You cannot accept premise one, use it to licence moral knowledge as your datum, and then say evil facts belong to a separate argument that need not trouble this one. The premise that generates the datum also generates the problem.
Perhaps you want to run a seperate argument rather than the one Andrew gave. One just from moral knowledge. But that does not work either. Knowledge is factive to know that P is the case entails that P is the case. So when the new argument invokes moral knowledge as its datum, it isn't just invoking the doxastic fact that humans form moral beliefs. It's invoking the stronger claim that those beliefs are true. But now look at what we actually know morally. The package of moral knowledge we possess isn't limited to pleasant normative facts it includes knowledge about the existence of evil. The datum itself, in virtue of being knowledge entails the existence of the very evil. When two propositions stand in an entailment relation like this, you cannot treat them as probabilistically independent.
Would it be possible to use the formulation where the datum is the doxastic fact that humans have lots of moral beliefs that happen to be true? This would help avoid your worry that there is a strict entailment relationship involved here. If the argument is just that it is surprising that we have lots of moral knowledge (and not that every moral belief we have classifies as knowledge), then the fact that we have a moral belief that P, doesn't need to logically entail that P is true.
We can imagine this by imagining a possibility space for all the worlds under consideration. Some of them will be worlds where lots of their inhabitants' moral beliefs are true and some of them will be worlds where that isn't the case. We update our credence in theism vs atheism relative to our proportions here and then put this in the background. This is basically the full extent to which the imagined moral knowledge argument would go.
Once we do that, we can further enter into the traditional problem of evil territory. Lots of us have a strong belief that there is evil in this world. Our first step here would be deciding whether this belief is true or false. If we decide that it is true, we would then decide the extent to which we think it is evidence against God's existence and update accordingly. But this entire last part seems to be separable from the first part.
The doxastic reformulation is interesting but I think it violates the requirement of total evidence principle, you must conditionalize on your total evidence, not a strategically selected subset of it. The total doxastic datum available to us isn't merely that humans have lots of moral beliefs that happen to be true. It's that humans have lots of moral beliefs that happen to be true and warranted, and a significant portion of those true warranted moral beliefs are specifically about the existence of evil. Both of these are part of our evidence simultaneously.
I think there are two serious problems with this argument.
Arguments like this don't actually compare theism with atheism given some datum they compare atheism or naturalism simpliciter with a very specific version of theism: one on which God is a perfectly good being with unconstrained power, knowledge, and a particular set of intentions toward creation. So the comparison isn't A vs T, it's really A vs (T & tri-omni & a-set-of-intentions), where all the heavy lifting is being done by the add-ons rather than by bare theism itself. But once you see that, the move loses much of its force because you can attach analogous add-ons to atheism or naturalism just as easily. Nagel's teleological naturalism is exactly this: atheism plus the thesis that mind and value are fundamental features of reality with an intrinsic tendency toward realization. Aristotelian naturalism is another: naturalism plus a robust account of natural teleology grounded in what kind of thing we are. Both of these predict moral knowledge and agency without God. So if the datum is evidence for anything, it's not evidence for theism over atheism as such for that you'd need to show that bare theism, once updated on the datum, yields a higher posterior than bare atheism does. What the argument actually shows, at best, is that the datum is evidence for versions of theism that already have built into them the intentions and values that make the datum expected.
The second problem concerns the first premise of the argument itself. If we grant that there are objective moral facts, then there are also objective facts about evil: that suffering is bad, that cruelty is wrong, that vast amounts of undeserved harm occur, that a vast amount of evil appears to be gratuitous. These facts are in significant tension with the specific version of theism being invoked here, namely one featuring an omnipotent, omniscient, and perfectly good creator. So even if we agree that P(moral knowledge & agency & facts | atheism) is very low, it's not obvious why P(moral knowledge & agency & facts | theism-with-omni-properties) wouldn't be even lower (it appears to be).
Classical theism just is the hypothesis of a perfectly good, rational creator. Those aren’t ad hoc add-ons.
By contrast, your enriched forms of naturalisms build in teleology and value to do the same explanatory work, just less unified, and in a very blatantly ad hoc manner.
On evil: of course it’s part of the data. But then the question is comparative. You’d need to show it outweighs the explanatory advantage theism has re: moral knowledge and agency.
I think the right move for you is to grant that moral knowledge and agency are evidence for theism but hold that theism is still unlikely overall.
An ad hoc add-on is one invoked specifically to rescue a hypothesis from disconfirming data, with no independent motivation. But Nagel and Aristotelian naturalists aren't responding to the moral knowledge and agenecy datum at all they developed their views on independent grounds. The enriched naturalisms weren't constructed to save atheism from your argument. So whatever their other flaws, ad hoc they are not.
You might want to say they are less unified than classical theism, and that this affects their prior. That may be a legitimate point but it is a completely different argument from the one being made in your original post. The original argument is that moral knowledge and agency constitute evidence for theism over atheism. If the reply is now that classical theism has a better prior than enriched naturalism due to greater unity or parsimony (and thus greater posteriori) we have shifted the dialectic entirely. My critique was directed at the original argument, and that stands on its own terms independent of this.
You say I need to show that evil outweighs the explanatory advantage theism has regarding moral knowledge and agency. But I don't think that's right. You are the one asserting premise four that on theism, moral knowledge and agency are to be expected. But what theism must predict is not just moral knowledge and agency in isolation. It includes objective moral facts and that landscape includes evil. So the burden falls on you.
You say "I think the right move for you is to grant that moral knowledge and agency are evidence for theism but hold that theism is still unlikely overall." but I do not think this is right. It is not evidence for theism but for specific conception of theism and that is not particular interesting. For the moral data is evidence for classical theism over bare atheism in roughly the same way that the existence of cheese is evidence for Cheese Theism over bare atheism both are data in which the enriched part not thei theist part is predicting. The question of whether God exists is supporter is not settled by noting that some elaborated version of theism (whether it is a classical tradition or not) predicts.
You’re still mislocating where the explanatory work is being done.
The issue isn’t whether Nagel or Aristotelian naturalists introduced their views to answer this argument. It’s that once you enrich naturalism with fundamental value and teleology, you’ve moved to a view on which the datum is no longer surprising. That’s fine — but then the comparison is no longer “theism vs atheism,” it’s between competing enriched worldviews. And at that point, unity and theoretical economy are directly relevant to the evidential question, not a shift in dialectic.
On evil: of course the total data matters. But that cuts both ways. You don’t get to treat evil as disconfirming theism without also asking how well atheism accommodates objective moral evil as such. The comparison remains holistic.
Finally, the “cheese theism” analogy just misses the structure of the argument. The claim is not that some arbitrarily enriched theism predicts the data, but that a well-motivated, independently supported conception — classical theism — renders it more expected. That’s exactly how evidential arguments are supposed to work.
At this point the positions are clear, so I’ll leave it there.
On the first point: I want to be precise about what the cheese analogy is actually showing, because I think you're not getting its force correctly. The point is about what the data is evidence for. Suppose God exists. The datum of moral knowledge and agency is then evidence about what kind of God exists one with the goodness and revelatory intentions that classical theism attributes to him. It confirms those attributes given theism. But that is different from being evidence that theism is true in the first place. Analogously, the datum of cheese confirms that this God has cheese-desiring intentions. That is evidence about God's nature given his existence, not evidence that God exists. The parallel holds regardless of how independently or well motivated the enriched hypothesis is, the evidence is for the enriched aspect not the bare-bone parts.
For D (moral data) to be evidence for theism, the proportion of theism-space that predicts D well would need to be significantly larger than the proportion of atheism-space that predicts D. That is whats relevant, not whether some specific corner of a vast theistic possibility space is well and independently motivated, that is orthogonal.
If the argument shifts to comparing classical theism against enriched naturalisms on grounds of theoretical unity and parsimony, that may well be a fruitful discussion but not relevant, its conclusion would be that classical theism is a better worldview than Nagelian/Aristotle naturalism, not that moral knowledge and agency are evidence for theism over atheism, the former can be true while the latter false.
Point about evil I agree that the comparison is holisticbut that does not discharge the burden I identified you have by asserting the forth premise.
Happy to leave it there.
Those are real objections. Briefly:
First objection: The concept of God here just is the classical one — a perfect being. So, it’s not contrived, unusual, or unmotivated. Additionally, there are multiple independent lines of argument for the existence of God so conceived. I’ve discussed a number of them in other posts.
Second objection: This is just the problem of evil — and yes, any perfect being theist needs some kind of response to it. I’ve outlined what I take to be the right way of thinking about that problem here: https://reflectionsonwhatmatters.substack.com/p/pondering-the-problem-of-evil?utm_campaign=post-expanded-share&utm_medium=web. Setting aside whether you think I’m right about that, theists generally don’t take the problem of evil to be decisive (hence their theism), so shouldn’t be expected to reject this argument because of it.
The question isn't whether classical theism is an old tradition or well-motivated independently but whether the moral data itself gives us reason to prefer theism over atheism, and that requires showing that bare theism predicts the datum better than bare atheism. That case hasn't been made. The moral evidence would be better regarded as evidence of God's nature and goals (that the classical thesis associates with him) given that he exists, than as evidence of a God simpliciter.
For the second point the argument. The point is not whether the theist takes EPOE to be decisive we agree that they dont what is at issue is whether evil does not disconfirm this classical theistic thesis to such a degree that it balances out the support that moral fact & agency & knowledge gives. The argument needs an additional premise: that whatever disconfirmation evil imposes (important to remember evil is part of the data invoked) on T* is outweighed by the confirmation provided by moral knowledge and agency to such an extent that the argument remain powerful.
I worry that your second paragraph might be an instance of lumping together things that need not be lumped. Sure, it might the case that the evidential problem of evil disconfirms theism to such a degree that it cancels out the boost it got from the existence of moral facts. But like, those are two separate arguments. We can coherently evaluate whether theism gets confirmed by the existence of moral knowledge and separately evaluate whether it gets disconfirmed by the existence of evil. In other words, we can start by assessing P(mknowledge|theism) vs P(mknowledge|atheism) and then put the existence of moral knowledge in the background (after updating on it) to evaluate P(evil|mknowledge&theism) vs P(evil|mknowledge&atheism). Or we could do it in reverse and start by evaluating P(evil|theism) vs P(evil|atheism) and then move on to compare P(mknowledge|theism&evil) vs P(mknowledge|atheism&evil). Or we could even do a c-inductive argument where our modest goal is to evaluate just P(mknowledge|theism) vs P(mknowledge|atheism) and leave the problem of evil to people who specialize on that area. In other words, I think there are many ways to go about it. It could be that once moral knowledge is accounted for, the existence of evil provides evidence that cancels it out, but that would be a separate argument that doesn't need to trouble this one. In general, the ability to break big chunks of data into smaller ones is what gives us the ability to discuss separate arguments for or against God without having to evaluate literally everything all at once.
The argument doesn't just happen to invoke moral knowledge it invokes a range of moral data that includes objective evil facts as part of the same package. So evil isn't a separate independent datum that comes from a different argument it is entailed by the first premise the argument.
You can evaluate P(d1|T) and P(d2|T) separately. But that works when d1 and d2 are independent data points. Here d1 (moral knowledge) and d2 (evil facts) are not independent. You cannot accept premise one, use it to licence moral knowledge as your datum, and then say evil facts belong to a separate argument that need not trouble this one. The premise that generates the datum also generates the problem.
Perhaps you want to run a seperate argument rather than the one Andrew gave. One just from moral knowledge. But that does not work either. Knowledge is factive to know that P is the case entails that P is the case. So when the new argument invokes moral knowledge as its datum, it isn't just invoking the doxastic fact that humans form moral beliefs. It's invoking the stronger claim that those beliefs are true. But now look at what we actually know morally. The package of moral knowledge we possess isn't limited to pleasant normative facts it includes knowledge about the existence of evil. The datum itself, in virtue of being knowledge entails the existence of the very evil. When two propositions stand in an entailment relation like this, you cannot treat them as probabilistically independent.
Would it be possible to use the formulation where the datum is the doxastic fact that humans have lots of moral beliefs that happen to be true? This would help avoid your worry that there is a strict entailment relationship involved here. If the argument is just that it is surprising that we have lots of moral knowledge (and not that every moral belief we have classifies as knowledge), then the fact that we have a moral belief that P, doesn't need to logically entail that P is true.
We can imagine this by imagining a possibility space for all the worlds under consideration. Some of them will be worlds where lots of their inhabitants' moral beliefs are true and some of them will be worlds where that isn't the case. We update our credence in theism vs atheism relative to our proportions here and then put this in the background. This is basically the full extent to which the imagined moral knowledge argument would go.
Once we do that, we can further enter into the traditional problem of evil territory. Lots of us have a strong belief that there is evil in this world. Our first step here would be deciding whether this belief is true or false. If we decide that it is true, we would then decide the extent to which we think it is evidence against God's existence and update accordingly. But this entire last part seems to be separable from the first part.
The doxastic reformulation is interesting but I think it violates the requirement of total evidence principle, you must conditionalize on your total evidence, not a strategically selected subset of it. The total doxastic datum available to us isn't merely that humans have lots of moral beliefs that happen to be true. It's that humans have lots of moral beliefs that happen to be true and warranted, and a significant portion of those true warranted moral beliefs are specifically about the existence of evil. Both of these are part of our evidence simultaneously.