Nice Gods Finish Last

Why Omnipotence and Omnibenevolence Are Incompatible

Pictured: This delightful, Biblically accurate image that came up when I googled “omnipotence”

There’s been some discussion recently about arguments for atheism that don’t revolve around the problem of evil, and whether any of them are any good. Now, speaking for myself, I really don’t think it matters much; the problem of evil is enough on its own to make theism unworkable, so any other issues would just be icing on the cake. But still, as a big fan of more indie, underground forms of godlessness - I was into reverse modal ontological arguments before it was cool - my answer is yes! There are a bunch of really good reasons to be an atheist that have nothing to do with actual suffering at all, and in this piece, I’m going to take a look of one of my favorites: The argument from the incompatibility of omnipotence and omnibenevolence.

I don’t think it’s always helpful to write out arguments in premise form, but in this case, it’s nice to see how every step in the chain of reasoning is (to me, at least) pretty much undeniable. This particular formulation is my own, but it’s building off two great papers by Wes Morriston that you can read here and here. It goes like this:

1. God is essentially omnibenevolent.

2. If God is essentially omnibenevolent, then there’s no possible world where God performs acts of gratuitous evil.

3. If there’s no possible world where you perform some act, then it isn’t possible for you to perform that act.

4. So, it isn’t possible for God to perform acts of gratuitous evil.

5. If it isn’t possible for you to perform some act, then you don’t have the ability to perform that act.

6. If you don’t have the ability to perform some act, then you don’t have the power to perform that act.

7. So, God doesn’t have the power to perform acts of gratuitous evil.

8. It’s conceivable that some being could do everything that God can do while also lacking omnibenevolence, in which case they would be able to also perform acts of gratuitous evil.

9. If a being has the power to do everything God can do, and also has the power to do some things God can’t do, then that being is more powerful than God.

10. So, there’s a conceivable being more powerful than God.

11. If there’s a conceivable being more powerful than you, then you aren’t omnipotent.

12. So, God isn’t omnipotent.

13. By definition, God is essentially omnipotent.

14. So, God cannot exist.

In short: If God is perfectly good, then it’s not possible for him to (for example) torture a cat. And if it’s not possible for him to torture a cat, then he doesn’t have the power to torture a cat, since you can’t have the power to do things that are impossible for you to do. But we can then imagine Frank, the all-powerful cat-torturing demon, who has all of God’s powers but none of his omnibenevolent scruples. Because Frank can do all the things that God can do, plus torture a cat, Frank is clearly more powerful than God is. And since you can’t be omnipotent if someone else could be more powerful than you, that means God can’t be omnipotent; assuming as you take omnipotence to be an essential quality of God, that means God can’t exist.

It’s honestly hard for me to imagine which of the premises here a theist could reject. The first four seem like something you’d want to affirm if you believe in God, while 5 and 6 seem about as close to analytically true as you could get, even if the conclusion you can draw from them at step 7 might start feeling uncomfortable. And once you’ve accepted premises 1 through 7, premises 8 and 9 should follow pretty easily too. That leaves 11 or 13 as the last two spots to take a stand, but 11 is also bordering on true by definition, and rejecting 13 is really just admitting that the argument works as stated and doing a major theoretical revision to your theism in response. All bad options, it seems!

Still, if I did have to guess where most theists would go, it would probably be to deny the whole chunk of premises at the start and adopt an alternative model of divine ability. The argument goes something like this: Sure, God “can’t” do something evil, but only because his perfect nature ensures he would never want to. Still, if he did somehow end up with a desire to engage in evil, then he could, so the impossibility here isn’t really a limitation on his power - it’s just a self-imposed refusal to exercise that power towards evil ends. So long as there’s nothing logically possible that God could desire and still not achieve, his omnipotence is preserved, even if practically speaking some acts are off limits because of his omnibenevolence.

This sort of response is intuitively appealing, but it has a bunch of glaring issues. The first is just that it depends on a totally novel and poorly defined sense of possibility that diverges sharply from the established possible-worlds model we use to evaluate pretty much everything else. That’s bad! One great sign that an intuition is confused or unreliable is that it requires arbitrarily changing the rules of the game in order to make sense; if we would normally take “There is no possible world in which agent X does Y” to be equivalent to “Agent X can’t do Y,” then we shouldn’t decide to analyze things differently just because that equivalence generates an embarrassing result for God. And similarly, if we’d normally see “Agent X can’t do Y” and “Agent X doesn’t have the power to do Y” as implying each other, then we should stick with the same judgment in this case.

But even if you are willing to build out a whole new sense of possibility just to secure an intuition, the one you’d need in this case doesn’t really make a lot of sense. The claim here is supposed to be that God has the power to engage in evil because, if he wanted to, he could. But everyone should agree that God’s omnibenevolence is an essential feature of his nature, so any claim about what God could or couldn’t do if he had a desire for evil is necessarily an impossible counterfactual; it’s just a claim about the sorts of powers he would have if his necessary qualities were other than what they could ever possibly be. How, then, does that claim respond to my objection, as opposed to just restate it? To say that God could do evil, if not for his omnibenevolence, just is to say that his omnibenevolence limits his power to do evil!

There’s a reason, I think, for why this sort of analysis appeals to people despite being so incoherent. It has to do with the fact that, when it comes to limited creatures like us, our desires are expressed through distinct and independent faculties. For example, when I want to lift something up, that desire is realized by engaging the muscles in my arms, and those muscles are going to be there regardless of whether or not the desire puts them into action. So it might make sense, in our case, to separate those powers from the desires behind them. It would still be wrong, of course, to say that human beings can have the power to do things they never do in any possible world. But since our desires are never actually necessary or essential in that way, it’s not weird that we’re used to thinking about desires and powers separately.

The problem, though, is that God doesn’t have any “evil-doing power” that his desires merely activate. Rather, his power is his desire, in the sense that things happen purely because he wants them to happen. So part of what’s going on here, if I had to guess, is that people are taking an intuitive separation between powers and desires that applies to humans and projecting it illegitimately on God. In reality, what God can or can’t do is exactly equivalent to what he can or can’t desire to do; the conceptual gap between intention and action that characterizes contingent, limited beings like us just doesn’t apply here.

And here’s the kicker: Even if you put those issues aside and just take the impossible counterfactual as it stands, I still don’t think it’s even true! Up until now, we’ve been talking about omnibenevolence solely in terms of desires, as though all it means to be omnibenevolent is that you’d never want to do evil. But maximal goodness, which is all the term exists to pick out, covers more than that; it also requires that all your actions themselves are always perfectly good. So even in some bizarre impossible world where one evil whim “slips through” God’s omnibenevolence, it isn’t as though all bets are suddenly off and he’s free to do whatever he wants. As long as he still possesses his omnibenevolent nature - which you’d need to stipulate in order for the counterfactual to be relevant at all - he’d still be incapable of making the further choice to act on that desire.

As an analogy, you can imagine Debbie the Democrat, who is essentially maximally liberal. That means she always desires to support liberal politicians and causes, sure, but also that she always actually does support those things. (If not, someone else who both desired to support those things and also actually did so would be more liberal than her, and therefore Debbie wouldn’t be maximally liberal.) So if we try to imagine a world where Debbie is essentially maximally liberal but still, impossibly, has a desire to vote for Donald Trump, then even in that impossible world, she still couldn’t do it; choosing to vote for Trump would be another deviation from her maximally liberal nature, over and above merely desiring to do so. In the same way, even if God gets one impossible evil desire inserted into his mind somehow, his choosing to act on it would be an additional unallowable violation of his omnibenevolence. So it still just isn’t the case that, if God wanted to do evil, he could.

I think these objections pretty convincingly demolish the claim that a perfectly good God would have the power to do evil. But as I said, once you’re forced to accept that, I really don’t see how else you’d avoid the conclusion of the argument. It’s trivially easy to imagine a being who shares all of God’s features except his omnibenevolence, in which case they’d be able to perform all of the evil acts that God’s perfect goodness rules out. And if another being can do everything you can do, plus some other stuff, then that being is more powerful than you - and that means you pretty clearly aren’t omnipotent, since there’s no one more powerful than an omnipotent God. So once you accept God’s inability to perform evil acts, there’s really nothing you can do to avoid also accepting that he isn’t truly omnipotent.

The real question to ask, then, is whether a conclusion like this counts as a decisive blow against the entire theist project. And to that question, I would answer: No, not necessarily, but it’s hardly just a parlor trick either. Omnipotence, after all, has been seen as an essential element of the classical theist framework for at least a millennia, so much so that philosophers often identify that deity in particular as a “tri-omni” God (with the other two omni-s being omnipresence and omnibenevolence). So for anyone whose theism is inexorably wrapped up in that particular project, I would say this sort of objection reveals their position to be fundamentally incoherent.

When it comes to less dogmatic theistic models, though, I’ll admit there’s more room for plausible revision. One approach might be to replace the rigidly tri-omni model of classical theism with a neo-Anselmian approach that identifies God with whatever being has the one maximally great combination of goodness and power overall. But how, exactly, are we supposed to know what that combination would be? We might be reasonably confident that a fairly powerful being who’s also perfectly good wins out over a perfectly powerful being who isn’t good at all - or else we’d be stuck with the worry that “God” properly refers to that amoral demon we mentioned before, rather than YHWH or whoever - but things get trickier when you look at an omnibenevolent God and start testing his limits one by one.

Take, for example, the power to make a tiny speck of dust appear in someone’s eye for no reason. Would adding a power like that to God’s repertoire outweigh a minor loss of goodness, or is refusing to inflict trifling annoyances on your creation more great-making than having the power to do so? I certainly don’t see any principled way of deciding one way or the other. But if you don’t have that power-to-goodness exchange rate figured out, then it seems possible to set up a series of trade-offs through which the greatest conceivable being is plausibly transformed, step-by-step, into someone pretty awful. All you need to do is list out every possible evil act and ask, one by one, whether the power to perform it would be greater than the ethical commitment not to; if you can’t be sure for any particular act, or even just for a hefty chunk of them, then it becomes close to impossible to even know what God’s character actually is.

(Or, if you’d like, you could skip past the sorites-style approach and just ask whether theists are generally justified in preserving omnibenevolence over omnipotence in the case of a conflict. Rather than falling back to the view that God has unlimited power, except when it would violate his perfectly good nature, why not assume instead that he’s perfectly good, except when it would limit his power? Unless we have an external reason to think omnibenevolence is guaranteed to win out by default - that “exchange rate” problem we saw earlier - then we end up in the same rough spot: Accepting any contradiction between the “omni” properties means giving up any certainty at all as to which ones God actually has.)

There’s also the broader question of how reconciling this contradiction would impact the intrinsic probability of theism itself, regardless of which attribute you ultimately prioritize. The classic sales pitch for the existence of God tends to make a big deal out of his simplicity, which makes sense if his attributes are just limitless expressions of a few fundamental concepts like power and goodness. But once you have to amend the picture to say that God has one or two unlimited attributes, combined with a complex conjunction of all the limited powers that can “fit in the gaps,” the overall picture is much less appealing; even if you can find a reason to explain why the balance would have to be the way it is, it’s still a clear increase in stipulated complexity.

Of course, there’s a huge amount more you’d need to say about any one of these points in order to really nail down the specifics. But for now, I’ll just settle for saying that the contradiction between omnipotence and omnibenevolence itself is as close as you can get to undeniable, and that it’s enough to at least rule out classical theism and other “tri-omni” models as incoherent. And considering that “tri-omni” models have been the default for centuries now in many of the world’s most important and influential religious communities, I think that’s a pretty big deal!

So while it may not pose the same kind of dire, all-or-nothing challenge to every kind of theism that the problem of evil does, the basic incompatibility of omnipotence and omnibenevolence is still an issue worth taking very seriously, and a great example of a novel objection to theism that doesn’t get enough attention.