This is the very article in which I try to chronicle an argument evaluative that’s also ontological

The following sets out an ontological argument directly inspired by Anselm’s original argument from Proslogion 2. I make no claim to be capturing Anselm’s own argument, and my aim is not to defend the soundness of this evaluative argument, but rather to explore what can be said in its favour (and, in the final section, how an atheist may respond to it).

The argument is notable in part because of what it is not. It makes no appeal to modal notions (including God’s possibility or non-contingency). It does not concern what is conceivably, conceptually, or definitionally the case. It does not make use of proper names, definite descriptions, or other potentially ambiguous referential apparatus to refer to God. It does not claim there to be non-existent objects, or things that exist in the mind but not reality, or otherwise make some deep ontological distinction between types of existence. It does not appeal to anything like the idea that God’s properties are closed under entailment, the claim that there are properties that are not God-properties, or any mereological or experiential notions. It is, finally, not Hegelian. All this being so, the argument does not fall under any of the useful labels Graham Oppy provides to taxonomise versions of the ontological argument (i.e., definitional, conceptual, modal, Meinongian, experiential, mereological, higher-order, and Hegelian – see Oppy Reference Oppy1995, extended in Reference Oppy2006). Rather, it is what I shall call an evaluative ontological argument, as it is made up of two relatively straightforward evaluative claims.

Ransacking Anselm’s argument

Anselm’s Proslogion 2 version of the ontological argument reads as follows:

The two ideas in Anselm’s argument crucial to my evaluative ontological argument are the ideas that (a) existence is great-making and (b) God is that than which no greater can be conceived. Both these ideas will be transformed by the end of this discussion, but what remains shows their influence clearly enough.

Let’s start with (a) – existence is great-making. This idea is left implicit in Anselm, but it is there (in the pithy phrase ‘which is greater’, or ‘quod maius est’). Here are a few ways we might initially try to express the idea (I shall read ‘greater’ as ‘better’):

Anselm need not be committed to the claim that existence is great-making for everything, nor to the idea that anything that exists is greater than anything that doesn’t. This whittling leaves two formulations:

I suggest we chop off the front half of the first formulation (‘It is better for God to exist’) and the back half of the second (‘if he exists than if he doesn’t’) and stick these halves together. Doing so gives us:

More fully, it is better for God to exist if he does exist than it is for him to exist if he doesn’t exist. This gives us my initial formulation of the idea that existence is great-making.

We’re not done yet, however. I do not wish my argument to include any proper names for God. As it stands, one might suspect that ‘God’ is supposed to directly refer to something, in which case acceding to the claim that it is better for God to exist if he does than if he doesn’t is already to concede God exists (a similar point may be made about the anaphoric ‘he’). Let’s replace ‘God’, then, with ‘a perfect being’ (or ‘something perfect’; I shall use ‘a perfect being’ and ‘something perfect’ interchangeably), and make similar changes to rid ourselves of any anaphora:

There’ll be no talk of God in my ontological argument!

This is better, but we are still using ‘exists’ as a predicate (at least, grammatically), and Kant taught us to be highly suspicious of this, especially in the context of the ontological argument (see Kant Reference Kant, Guyer and Wood1998; Proops Reference Proops2013; Stang Reference Stang2015 for discussion). In using ‘exists’ as a predicate, one might fairly wonder if we are allowing there to be things that do not exist. Thus, one way of taking the claim that the golden mountain does not exist is that there is such a thing as the golden mountain, but this thing doesn’t enjoy existence. This leads us into a contentious metaphysics of non-existent objects I wish to avoid. Given this, we may state my formulation in explicitly quantificational terms:

This, for now, is my formulation of Anselm’s idea that existence is great-making. Let’s call it the Superiority of Perfect Existence (or SPE).

Turn now to the claim that God is something than which no greater can be conceived. As I am using ‘better’ instead of ‘greater’, this becomes:

We need to make various other changes to bring the idea in line with SPE. In SPE, we are talking not of some thing being better, but rather what it is better for there to be. We may reflect this change as follows:

We should also replace our talk of God with talk of a perfect being:

A perfect being is something than which no better for there to be can be conceived.

Wishing to avoid the idea that there are intentional objects that may or may not exist, I shall interpret the above as saying that, regardless of what type of thing can be conceived, it is no better for there to be something of that type than it is for there to be something perfect. We may express this by utilising a second-order quantifier and variable, which allows us to quantify into predicate position. This gives us the following:

We have now expunged much of what might be considered suspicious from our initial claim that God is that than which no greater can be conceived.

But the work is not yet done. I have no great desire to adjudicate what is and what isn’t conceivable. And if we take Anselm to be offering us a standard for conceivability when he tells us that, should we understand a phrase, a thing falling under that phrase exists in the understanding, the standard he’s offering seems to be that, as long as a phrase isn’t nonsense, it is conceivable for something to fall under it. It is, of course, already understood that our variable ‘F’ can’t be replaced with nonsense, because in doing so we would be making no claim at all. I suggest, then, that we simply drop any talk of conceivability, simplifying our formulation thusly:

This happily avoids difficult questions about conceivability.

Just as soon as we have simplified our formulation, we need to complicate it again. While the above formulation expresses the idea that the unconditional value of there being an F (for any F) is no greater than the unconditional value of there being something perfect, we will need to express the idea that the value of there being an F (for any F) conditional on anything is no greater than the unconditional value of there being something perfect. This is certainly in the spirit of Anselm’s idea – when he tells us that God is that than which no greater can be conceived, he means that we cannot conceive of anything under any conditions that exceeds God’s greatness. We shall use a propositional variable and quantifier to help us express this idea:

Let us call this claim the Absolute Supremacy of Perfection (or ASP). It is my formulation of Anselm’s characterisation of God as something than which no greater can be conceived.

An evaluative ontological argument

We thus have our two evaluative claims, SPE and ASP. These are the premises of my evaluative ontological argument:

1. 1. For any F, and any Q, it is no better for there to be an F, given Q, than it is for there to be something perfect. (ASP)

2. 2. It is better for there to be a perfect being if there is a perfect being than if there isn’t a perfect being. (SPE)

3. 3. Therefore, there is a perfect being.

Note that the above is clearly an ontological argument, but it makes no modal claims (there isn’t a necessity operator at the front of either premise, and all conditionals should be interpreted materially, not strictly), no claims about conceivability or definitions, no mereological or experiential claims, no distinctions between types of existence, and it uses no suspicious referential apparatus (no ‘God’, no definite descriptions, no predicative use of ‘exists’, etc.). Such an argument is what I was groping towards in Kearns (Reference Kearns2019).

To see fully why the argument is valid, we will need to formalise it. Before that, we can gesture towards why as follows. An instance of ASP is the following (where ‘an F’ is replaced by ‘something perfect’ and ‘Q’ is replaced by ‘there is something perfect’):

1. 4. It is no better for there to be something perfect, given there is something perfect, than it is for there to be something perfect.

Following Anselm’s lead, let’s assume, for reductio:

1. 5. There is no perfect being.

Assuming 5, the value of there being something perfect is equal to the value of there being something perfect given there is nothing perfect. According to premise 2 (i.e., SPE), the value of there being something perfect given there is something perfect is greater than the value of there being something perfect given there is nothing perfect. Thus 5 and 2 together entail that the value of there being something perfect given there is something perfect is greater than the value of there being something perfect. Or, in other words, 2 and 5 entail:

1. 6. It is better for there to be something perfect, given there is something perfect, than it is for there to be something perfect.

Claims 6 and 4 contradict each other – together they say that it is both better and no better for there to be something perfect, given there is something perfect, than it is for there to be something perfect. This contradiction reduces our assumption to absurdity. We thus must reject 5 and conclude:

3. There is a perfect being.

ASP and SPE thus entail that there is something perfect.

The essence of the idea, then, is that if there is no perfect being, the value of there being a perfect being conditional on there being one is greater than the unconditional value of there being one. But the value of there being anything conditional on anything is no greater than the unconditional value of there being a perfect being. Thus, there is a perfect being.

To formalise the argument, we shall use variables that quantify over degrees of goodness (‘d’, ‘e’), a constant for such degrees (‘f’), and an operator, ‘G’, which relates degrees of goodness to propositions, and is to be interpreted as ‘It is good to at least degree x for it to be that…’ Thus ‘Gd(Q)’ reads ‘It is good to at least d for it to be that Q’. As introduced previously, we shall also use the second-order variable ‘F’ and the propositional variable ‘Q’ and their respective quantifiers. Finally, we shall use the propositional constant ‘P’, meaning ‘There is a perfect being’.

This understood, we may formulate ASP and SPE as follows:

Thus formulated, ASP says: for all degrees of goodness d, all F, and all Q, if (if Q, then it is good to at least d for there to be an F), then it is good to at least d for there to be something perfect. Or, in slightly more manageable English: for all d, F, and Q, if it’s good to at least d for there to be an F given Q, then it’s also good to at least d for there to be something perfect. SPE says: there exists a degree of goodness, e, such that, if there is something perfect, then it is good to at least e for there to be something perfect, and if there is nothing perfect, then it’s not good to at least e for there to be something perfect. A brief perusal should show that these formulations capture ASP and SPE (it may be helpful to note that we may also state ASP like this: however good it is for there to be an F, given Q, it is at least as good for there to be something perfect).

We proceed with the argument as follows:

Most of this is standard predicate logic. The only slightly exotic inferences concern 7 and 8. Claim 7 is derived from 6 by second-order universal instantiation (thus ‘∀F’ is eliminated and variable ‘F’ is replaced by ‘perfect’, rendering ‘(∃x)(Fx)’ as ‘P’ (for ‘there is something perfect’)). Similarly, 8 is derived from 7 by propositional universal instantiation (thus ‘∀Q’ is eliminated and propositional variable ‘Q’ is replaced by the propositional constant ‘P’).

The ‘it is good for it to be that’ operator

Before exploring the premises of our evaluative ontological argument, it is worth explicating further the crucial ‘it is good (to degree d) for it to be that’ and ‘it is better for it to be that…than that’ operators (thank you to two anonymous referees for urging me to clarify these notions). I shall concentrate on the former, as the latter is analysed in terms of it (henceforth, I shall also often drop the phrase ‘(to degree d)’)). These are two of a series of such operators, including ‘it is (outright) good for it to be that’ and ‘it is bad for it to be that’ (e.g., ‘it is good for it to be that people are happy’; ‘it is bad for it to be that innocents suffer’). In what follows I shall often state claims of the form ‘it is good/bad for it to be that x is F/x Fs’ in more natural English as ‘it is good/bad for x to be F/to F’ (thus ‘it is good for people to be happy’; ‘it is bad for innocents to suffer’).

The first important point to note is that ‘it is good for it to be that’ is not equivalent to ‘it is good that’. The latter operator is factive; necessarily, it is good that people are happy only if people are indeed happy. The former is not factive; it is possible, for instance, that it is good for people to be happy even when no one is happy. Indeed, that ‘it is good for it to be that’ is non-factive is crucial for our purposes. A factive operator such as ‘it is good that’ would render the evaluative ontological argument a damp squib. On such a factive reading, ASP alone entails there is something perfect. For suppose there is nothing perfect; then there is no degree to which it is good that there is something perfect, in which case the consequent of (every instance of) ASP is false. Seeing as there are (necessarily) many instances of ASP that have true antecedents (including any instance that has a false value for ‘Q’), ASP will itself turn out false. Necessarily, then, if nothing is perfect, ASP is false. Contraposing, necessarily, if ASP is true, there is something perfect.

The ‘it is good for it to be that’ operator, in contrast to the ‘it is good that’ operator, provides us with a tool for expressing the value of a state of affairs without asserting that this state of affairs obtains. What’s more, we may understand ‘it is good that’ in terms of ‘it is good for it to be that’ as follows: it is good that P just in case P and it is good for it to be that P. This also helps us grasp ‘it is good for it to be that’ – it is, speaking roughly, equivalent to ‘it is good that’ minus the latter’s factivity.

Is ‘it is good for it to be that’ equivalent, then, to ‘it would be good for it to be that’? The latter operator is counterfactual (at least on the most natural reading) – it would be good for people to be happy just in case, were people happy, it would be good that people are happy. Given my ambitions to avoid appeal to modal notions in the evaluative ontological argument, this reading of ‘it is good for it to be that’ would not serve my purposes well. And, indeed, I reject this equivalence. This is roughly because ‘it would be good for it to be that P’ does not tell us the actual value of a state of affairs, but only what its value would be were the state to obtain.

There are multiple ways we may further bring out the difference between the two operators. First, consider a toy model of value. Let’s say that the goodness or badness of a state of affairs is equivalent to the extent to which it is valued or disvalued on average (such a subjectivist account of value is, of course, implausible; it is only being used for illustrative purposes). We may express this idea using the ‘it is good for it to be that’ operator as follows: it is good for it to be that P just in case its being that P is positively valued on average. Thus, it is good for people to wear orange just in case people’s wearing orange is positively valued on average (this is not to say that people’s wearing orange is valued when it obtains, but simply that this state of affairs is valued). Further, suppose that people’s wearing orange is currently positively valued on average. So it is good for people to wear orange. But our valuers are a fickle lot. Currently people aren’t wearing orange, but if they were to, valuers would no longer positively value people wearing orange. Given all this, though it is good for people to wear orange (as this is positively valued), it is not the case that it would be good for people to wear orange (because, were people to wear orange, this state of affairs would not be positively valued). This model shows how the two operators can come apart – ‘it is good for it to be that P’ informs us of the actual value of its being that P (regardless of whether it is true that P), while ‘it would be good for it to be that P’ informs us only of the counterfactual value of its being that P, were it true that P.

We can reach the same conclusion by considering the (lack of) value of certain necessarily false propositions. It is not good for me to be tortured for eternity and then completely obliterated from existence. Thankfully, given some plausible assumptions, this is also impossible – I cannot be tortured for eternity and then be obliterated if, necessarily, I exist only if I begin to exist (for then I can be tortured for eternity only once I have started existing, in which case there will be no time at which the torture is finished and I am subsequently obliterated). Taking some appropriate essentiality-of-origin thesis as read, there must be a beginning to my existence (if I exist at all), and thus this sorry state of affairs is indeed impossible. But, on the counterfactual reading of ‘it would be good for it to be that’, it nonetheless would be good for me to be tortured for eternity and then completely obliterated from existence. That is, were I to be tortured for eternity and then completely obliterated from existence, it would be good that I am tortured for eternity and then completely obliterated from existence. This is because, on standard and widely accepted semantics of counterfactuals, any counterfactual with an impossible antecedent is trivially true (a thesis I embrace). This difference is particularly pertinent in debates about God, as the matter of God’s existence is standardly assumed to be non-contingent (i.e., God is either necessary or impossible).

Lastly, we may consider another toy model to bring out the difference between the two operators. Let’s say that hedonism (the thesis that (only) pleasure is valuable) is contingently true, and that, in any possible world in which it is false, asceticism (the thesis that pleasure is disvaluable) is true. Also suppose there is no pleasure, and, due to some quirk, were there to be pleasure, hedonism would be false (and asceticism true). Given all this, it is good for there to be pleasure (as hedonism is true), but it is not the case that it would be good for there to be pleasure (for, if there were pleasure, asceticism would be true and thus it would be bad (and not good) for there to be pleasure).

The ‘it is good for it to be that’ operator, then, is used to express the (positive) value of a state of affairs without entailing that this state obtains. It is thus not equivalent to ‘it is good that’. It is also not equivalent to ‘it would be good for it to be that’ on a counterfactual reading of the latter, as the former assigns value to a state of affairs as things are while the latter tells us what value it would have were it to obtain. The ‘it is good for it to be that’ operator thus allows us to talk of the actual value of states of affairs that do not obtain, that cannot obtain, and that would have different value were they to obtain, which we cannot do with either ‘it is good that’ or ‘it would be good for it to be that’ (on its counterfactual reading).

There is, in my estimation, an alternative reading of ‘it would be good for it to be that’ according to which it is a necessitised version of ‘it is good for it to be that’, for some restricted form of necessity. On this reading, it would be good for it to be that P just in case, in all relevant possible worlds (including the actual world), it is good for it to be that P. In contexts where the only relevant possible world is the actual one, ‘it would be good for it to be that’ is equivalent to ‘it is good for it to be that’. While a version of the evaluative ontological argument could be given with this reading of ‘it would be good for it to be that’, it adds a modal dimension to the argument that we can do without.

The absolute supremacy of perfection

Let us now consider the premises of this argument, starting with ASP. We have seen various formulations of ASP, which I take to be equivalent. In rough order of readability:

ASP captures the thought that nothing under any conditions beats total perfection. This idea tallies well with our understanding of what it takes to be truly perfect. It also tallies with what we can infer, given limited information. Let’s say we’re unfamiliar with the words ‘numpty’ and ‘pillock’. We learn that it’s better for there to be a numpty than it is for there to be a pillock. Isn’t obvious that we can infer from this that ‘pillock’ must not mean ‘a perfect being’? Even if we only learn that it’s better for there to be a numpty, given Canberra is the capital of France, than it is for there to be a pillock, it seems we can safely infer that ‘pillock’ does not mean ‘a perfect being’.

Instances of ASP further bear out its plausibility. Consider the following. Regardless of how good it is for there to be ice cream, even if ice cream cures cancer, it is at least as good for there to be a perfect being. It is no better for there to be world peace, on the assumption that war is the most monstrous evil, than it is for there to be something perfect. If it’s totally amazing or better for there to be eternal bliss for all of us, given that eternal bliss is the most desirable state to be in, it’s also totally amazing or better for there to be a perfect being. I have chosen here types of things I take to be very good indeed, conditional on facts that make it even better for there to be these types of things, but, plausibly, none exceeds the goodness of there being something perfect. We might conclude that the value of there being any type of thing, conditional on the truth of any claim, does not exceed the value of there being something perfect.

Someone may judge at least one of these claims to be false, and thus further judge their negations to be counter-examples to ASP:

The first potential counter-example contends that the goodness of there being a perfect being plus something else good (great joy), conditional on there being such, outstrips the goodness of there being a perfect being. The idea is simple – great joy is good, perfect beings are good; together they are even better. The idea behind the second potential counter-example is also simple – isn’t it better for there to be something better than perfect, if there is something better than perfect, than it is for there to be something perfect?

The natural reply to these proposed counter-examples is something like the following. It is perfect for there to be something perfect. If it weren’t, there would be something deficient about there being something perfect, which there isn’t. As such, for any degree of goodness at all, it is good to at least that degree for there to be something perfect. Given this, there is no degree to which it is at least as good for there to be something perfect plus many joyful people, conditional on there being such things, that is not also a degree to which it is at least as good for there to be something perfect. Thus, it is no better for there to be something perfect plus many joyful people, given there are such things, than it is for there to be something perfect. Similarly, there is no degree to which it is at least as good for there to be something better than perfect, conditional on there being such a thing, which is not also a degree to which it is at least as good for there to be something perfect. Thus, it is no better for there to be something better than perfect, given there is something better than perfect, than it is for there to be something perfect. We might see this as further evidence of ASP – if it’s not even true that it’s better for there to be something better than perfect, conditional on there being something better than perfect, than it is for there to be something perfect, it is plausible to suppose there are no counter-examples to ASP.

In light of the above considerations, it is tempting to say that ASP is a conceptual truth. I am sceptical of the notion of conceptual truth, unless all we mean by ‘a conceptual truth’ is ‘synonymous with a logical truth’ (see Williamson Reference Williamson2007, chapters 3 and 4, for further discussion of the notion), but even if you aren’t, it’s not clear to me you should think it is one. ASP is similar to statements such as these:

These three claims strike me as obviously true, but are they conceptual truths? Certainly, their form doesn’t seem to guarantee their truth. Thus consider:

None of these claims strike me as true. It’s probable for there to be improbable things; I wonder whom exactly we are supposed to admire for there being admirable things; and, while urgent things are those there is a pressing need to deal with lest something very bad occur, it is not a pressing matter for there to be such urgent things – indeed, it would be preferable if there weren’t.

Still, even if ASP is not a conceptual truth, it keeps good company. That is, of the six statements mentioned above, ASP seems far closer in spirit to the first three than the second three. This is further evidence that it is true.

The superiority of perfect existence

SPE is perhaps easier to grasp but harder to accept than ASP. We have seen a few formulations:

In essence, SPE tells us that, for some degree of goodness, e, it is good to at least e for there to be something perfect if and only if there is something perfect (SPE thus rules out as true the claim that, for any degree of goodness, d, there is nothing perfect and it is nonetheless still good to at least d for there to be something perfect). In combination with ASP, which entails that, for any degree of goodness, d, it is good to at least d for there to something perfect, SPE allows us to conclude that there is something perfect. Unlike ASP, however, which will likely strike many as true right off the bat, SPE does not wear plausibility on its sleeve.

I see two main arguments for SPE. I shall start with the simpler one. This is the idea that any perfect being is a universal enhancer – it is better for there to be any type of thing if there is a perfect being than if there isn’t. SPE is an instance of this claim:

The idea that a perfect being is a universal enhancer may itself be supported by the following argument. It’s better for any perfect being to be a universal enhancer (as understood above) than not. For any F, if it’s better for any perfect being to be an F than not, then any perfect being is an F. Therefore, any perfect being is a universal enhancer.

The second argument for SPE is based around the idea that value is perfection-centric if and only if there is a perfect being (call this biconditional ‘VIP’ for ‘Value Iff Perfect’). What, then, does the claim that value is perfection-centric, the left-side of this biconditional, come to? In essence, the idea is that perfection (i.e., the property of being perfect) has a special place in the metaphysics of value. Thus, someone who subscribes to the perfection-centricity of value may accept all (or most) of the following claims:

Most of the above claims are relatively self-explanatory. The first, that perfection has lexical priority over any other value, amounts to the idea that, regardless of the degree of a certain value – pleasure, friendship, or what have you – perfection is always greater. Together, these claims paint a picture of perfection as the ultimate value, the source of all other value (and moral reason), identical to central divine properties, unmatched by any other value (no matter the quantity of such value), and tragic for the world to lack. This picture, or at least major elements of it, make up the idea that value is perfection-centric.

VIP, then, is the claim that (most of) the above claims are true if and only if there is a perfect being. This is a plausible hypothesis, by both the theist’s and atheist’s lights – the typical theist will accept both sides of the biconditional as true, and the atheist will accept both sides of the biconditional as false. The biconditional comes out as true either way.

We cannot derive SPE from VIP alone. Consider the following further thesis, then, which will help us out: it’s better for there to be something perfect if value is perfection-centric than if value isn’t perfection-centric. Call this claim ‘SPV’ for ‘Superiority of Perfection-centric Value’. I take SPV to be rather plausible on its face. If value is perfection-centric, then it’s special for there to be something perfect. If it isn’t, it’s not so special for there to be something perfect. This is not to say it’s bad for there to be a perfect being if value is not perfection-centric. It’s just not as good (it may still be very, very, very good). If value is perfection-centric, however, the value of there being something perfect is on a whole other level.

We may think of SPV as making a claim similar to each of the following plausible ideas:

In other words, SPV tells us that the extent of the goodness of there being something perfect depends on the very nature of value. The more perfection plays a central role in the nature of value, the better it is for there to be something perfect. Just as the extent of the value of there being pleasure depends on the role the property of being pleasurable plays in the metaphysics of value (is it the be-all and end-all of value, a worthy but non-crucial contributor to well-being, or a scourge that keeps us distracted from what is truly good?), the degree to which it is good for there to be something perfect turns on whether value is perfection-centric.

VIP and SPV together entail SPE. That is, the following argument is valid:

1. 1. It’s better for there to be something perfect if value is perfection-centric than if value isn’t perfection-centric. (SPV)

2. 2. Value is perfection-centric if and only if there is a perfect being. (VIP)

3. 3. Therefore, it’s better for there to be a perfect being if there is a perfect being than if there isn’t a perfect being. (SPE)

The argument may be formalised as follows:

The formalisations of SPV and VIP should be relatively clear. ‘V’ is the sentential constant ‘value is perfection-centric’. VIP is thus a simple biconditional, while SPV is of the same form as SPE, with ‘V’ and ‘ ∼ V’ replacing ‘P’ and ‘ ∼ P’ as the antecedents of the two respective embedded conditionals. The logic is straightforward.

Parodies

While my evaluative ontological argument avoids many of the objections to existing ontological arguments by avoiding many of their questionable commitments, all ontological arguments must face the parody gauntlet. If a parallel argument can be constructed with premises that are just as plausible, but with a conclusion we cannot accept, this casts considerable doubt on the original argument. In this section, I shall discuss two parodies that might fit the bill.

We begin with a version of Gaunilo’s classic objection that Anselmian reasoning would lead us to embrace the existence of the Lost Island – the island greater than any other. Thus consider the following Gaunilo-inspired argument:

1. 1. For any F, and any Q, it is no better for there to be an island which is F, given Q, than it is for there to be a perfect island.

2. 2. It is better for there to be a perfect island if there is a perfect island than if there isn’t a perfect island.

3. 3. Therefore, there is a perfect island.

We cannot show there to be a perfect island using the above argument. It is tempting to conclude that our original evaluative ontological argument must be faulty in whatever way the above argument is.

I think this parody fails because its first premise is false for the following substitutions for ‘F’ and ‘Q’: ‘accompanied by a perfect being’ and ‘there is an island which is accompanied by a perfect being’. This instance of the first premise thus reads:

This instance of the first premise of the Gaunilo-style parody is, I suggest, not true – it is better for there to be an island accompanied by a perfect being, given there is an island which is accompanied by a perfect being, than it is for there to be a perfect island. Though it is indeed very good for there to be a perfect island, how good it is is nevertheless constrained by what it takes to be an island. In contrast, how good it is for there to be an island accompanied by a perfect being, especially given that there is such an island, is not similarly constrained.

The second parody is more formidable. It is based on the classic parodic ontological argument for the devil (versions of which date back at least as far as Cock 1917–Reference Cock1918):

1. 1. For any F, and any Q, it is no worse for there to be an F, given Q, than it is for there to be a perfectly bad being.

2. 2. It is worse for there to be a perfectly bad being if there is a perfectly bad being than if there isn’t a perfectly bad being.

3. 3. Therefore, there is a perfectly bad being.

Again, we should surely resist the notion that the above argument gives us good reason to accept the existence of something perfectly bad. Insofar as this parody and the others are bad arguments, we may suspect that the original evaluative argument is too, even if we have not put our fingers on exactly where any of these arguments go wrong.

In response, I recommend a divide and conquer strategy on behalf of the theist. It is most natural to read ‘perfectly bad’ as the opposite of ‘perfect’. While something perfect has no flaws of any kind, something perfectly bad has nothing but. We may use the term ‘perfectly deficient’ to capture this meaning of the phrase. This gives us the following version of the parody:

1. 1. For any F, and any Q, it is no worse for there to be an F, given Q, than it is for there to be a perfectly deficient being.

2. 2. It is worse for there to be a perfectly deficient being if there is a perfectly deficient being than if there isn’t a perfectly deficient being.

3. 3. Therefore, there is a perfectly deficient being.

This parody fails because it isn’t so bad for there to be something perfectly deficient – while such a thing might be immoral, it is also ineffective, powerless, ignorant, laughable, and so on. There will thus be many substitutions for ‘F’ and ‘Q’ according to which the first premise is false.

Two other understandings of ‘perfectly bad’ suggest themselves. We may understand badness morally and interpret ‘perfectly bad’ as perfectly evil. Or we might understand badness as harmfulness and interpret ‘perfectly bad’ as perfectly harmful.

These parodies have the same problem as each other. Because we are narrowing our understanding of badness (from the general notion captured by what we are calling deficiency), we have little reason to think that the first premise of either argument is true. Thus, consider the following instance of the first premise of the perfectly evil being parody:

This is false – however bad it is for there to be something perfectly evil, it is worse for there to be such a thing accompanied by huge amounts of suffering. As we have seen, the matter is arguably different with respect to our original evaluative ontological argument – it is part of the driving idea behind ASP that perfection simpliciter is simply not outstripped in any manner.

This suggested asymmetry between the original argument and these devil parodies may, of course, be questioned. Indeed, I do not take what I say about any of these parodies to be conclusive, and there are no doubt other interestingly distinct parodies well worth investigating. To reiterate – my aim is not to defend the evaluative ontological argument, but rather to suggest that the argument is worth exploring. I am confident it avoids many of the pitfalls of various other ontological arguments, but I do not claim that it avoids every such pitfall. Indeed, I take parody arguments to be among the most difficult problems for ontological arguments; some may even provide reasons to reject the existence of God (see Kearns Reference Kearns2021). This said, I know of no devastating parody of the evaluative ontological argument.

Perfection and possibility

Modal ontological arguments typically include the premise that it is possible for there to be something perfect (or something maximally great, etc.). It is unclear, however, how this claim can be justified independently of the claim that there is something perfect. After all, it is the contention of modal ontological arguments that these claims stand or fall together – either God exists or God is not possible. Should an atheist or agnostic accept that God’s possible existence entails God’s necessary existence (or God’s actual existence), she will accordingly reject God’s possible existence. Given this, modal ontological arguments are often judged to be, at best, dialectically ineffective (see, for example, Oppy Reference Oppy1995; Van Inwagen Reference Van Inwagen and Szatkowski2012). Various responses might be given to this worry. One straightforward reply is to offer positive arguments in favour of God’s possible existence (Antognazza Reference Antognazza and Oppy2018 nicely sets out how Leibniz tries to do this).

The state of play in this debate is not my concern here. Rather, I wish to consider whether something like this objection applies to the evaluative ontological argument. Roughly put, is one justified in believing the premises of this argument without having prior justification for believing that it is possible for there to be something perfect? If we cannot be so justified in accepting the argument’s premises, the argument illicitly assumes the possibility of a perfect being and thus must be either supplemented with an independent case for such a possibility or discarded.

Now, clearly there is a sense in which the argument assumes the possibility of something perfect; the two premises entail such a possibility (by dint of entailing that there is something perfect). The argument cannot, of course, be faulted on that score. And there is another sense in which the argument does not assume such a possibility; neither premise claims that a perfect being is possible, nor formally entails it.

One interesting question is whether either premise strictly implies the possibility of perfection on its own. Is it necessary that, if ASP is true, then a perfect being is possible? Is it necessary that, if SPE is true, a perfect being is possible? I am uncertain. Consider the former question. One might answer positively on the grounds that, (necessarily) for any F, if it is impossible for there to be an F, then, for some G, it is better for there to be a G than it is for there to be an F. Indeed, we might think that, for any F, if it is impossible for there to be an F, it is simply of no positive value at all for there to be an F. Thus, if perfect beings are impossible, there are many kinds of things it is better for there to be (perhaps love, life, and art).

It is unclear to me whether this is right. I can certainly see the sense in someone embracing the impossibility of a perfect being while wistfully remarking that it is nonetheless better for there to be such a thing than anything that is possible for there to be. Indeed, such a person might judge the impossible-to-instantiate property of perfection to capture a genuinely worthy ideal to which we can aspire but can never attain. While nothing can be perfect, it is nonetheless better for there to be such a thing than anything else.

Even if ASP does strictly imply the possibility of perfection, however, this does not show that this premise illicitly assumes this possibility in the sense that any justification we might have for believing ASP rests on prior justification that perfection is possible. Strict implication on its own does not tell us anything like this epistemic story. And, indeed, the basic case I made for ASP makes no mention of the idea that it is possible for something to be perfect (and nor, might I add, does the immediate intuitiveness of ASP). Indeed, if we are convinced that the strict implication holds, we may instead take seriously the following argument: ASP is true; ASP strictly implies the possibility of a perfect being; therefore, it is possible for there to be something perfect. Rather than finding the strict implication problematic, that is, some may find it to be an independent reason for accepting that perfection is possible.

The same basic idea applies to SPE. First, it does not obviously strictly imply the possibility of perfection (there is certainly no explicit contradiction in thinking both that it is better for there to be something perfect if there is than if there isn’t and that perfection is impossible). Second, the arguments set out above for SPE make no obvious commitment to such a possibility. And third, even if SPE does strictly imply that it is possible that there be something perfect, this does not mean that any justification for accepting SPE rests on prior justification for believing that perfection is possible (indeed, as with ASP, we might rather use SPE to support such a possibility).

The dialectic, then, is at best unclear. The objection is thus not, as of yet, terribly forceful. Still, we may wonder what those who think perfection impossible should say about the evaluative ontological argument. Should such people simply be swayed by it (and come to accept the possibility of perfection)? Should they reject a particular premise? These are interesting questions, but they are less direct than they can be. Instead of asking what those who reject the possibility of perfection should say about the argument, we should instead ask the broader question of what those who reject the existence of a perfect being should say about the argument (this latter question obviously encompasses the former). To this we now turn.

Two atheistic takes

In my view, the evaluative ontological argument makes no egregious logical or semantic error, nor does it rest on any deeply idiosyncratic metaphysical or modal claims. Furthermore, ASP and SPE are plausible, even independently of one’s prior commitments concerning the conclusion of the argument. Lastly, the argument is not obviously susceptible to damning parody. The evaluative ontological argument is, to this extent, a success.

Still, we are not rationally compelled to accept either premise. How much either premise resonates with each of us will be a result of our total evidence, other background beliefs, and the ways we are disposed to think about the world. Arguments are tips of icebergs – they are the small visible peaks on public display supported by a much larger array of thought that is often hidden. As such, while the theist might find the premises clearly true, things need not be the same for the atheist. This is not to say that the argument is question-begging (a charge that is anyway often unclear) – one’s justification for believing the premises need not stem from any prior commitment to the conclusion – but that the array of beliefs the theist has may differ from the atheist’s in such a way that the former, but not the latter, is likely to accept the premises. What’s more, the atheist (who I will here understand as someone who rejects the existence of a perfect being) may be being reasonable in rejecting one or more of the argument’s premises. To illustrate this, I shall provide replies to the argument as if written by two types of atheist, the first of whom accepts ASP and denies SPE and the second who does the reverse.

Atheist-1

I agree with the proponent of the evaluative ontological argument that premise 1 (ASP) is true. That is, I accept that, for any F, and any Q, however good it is for there to be an F, given Q, it is at least as good for there to be something perfect. While ASP may not be a logical truth, it strikes me as capturing something about the nature of perfection – the value of there being something perfect cannot be outstripped by the value of there being anything else (even under the most favourable conditions). This is simply part of what perfection amounts to. However, I reject the existence of a perfect being – there is just too much evil in the world for me to countenance such an entity. I thus reject SPE; I do not think it is better for there to be something perfect if there is such a thing than if there isn’t.

I perforce reject the arguments for SPE. Consider first the idea that any perfect being is a universal enhancer – for every F, it is better for there to be an F if there is something perfect than if there isn’t. It is compatible with me rejecting this claim that I accept that, for many kinds of things, the value of there being such things is enhanced by the existence of a perfect being. But the value of there being something perfect is not enhanced by there being something perfect, precisely because it is impossible to enhance the value of there being something perfect – this value is already maxed out, whether or not there is something perfect.

The other argument for SPE suffers from the same error. It tells us, in part, that it is better for there to be something perfect if value is perfection-centric than if it isn’t (this is SPV). But, again, it is of maximal value for there to be something perfect, even while (in my view) value is not perfection-centric (because value is perfection-centric if there is something perfect; so, seeing as there is nothing perfect, value is not perfection-centric). Thus, it is no better for there to be something perfect if value is perfection-centric than if it isn’t – there is just no room for improvement. I concede that this may strike some as misguided, and I feel the pull of SPV even though I reject it. But something’s got to give, and the arguments against the existence of a perfect being (the great evil in the world, the success of naturalistic world views in accounting for the universe, etc.) compel me more than does the case for SPV.

Atheist-2

I agree with the proponent of the evaluative ontological argument that premise 2 (SPE) is true. That is, I accept that it is better for there to be something perfect if there is such a thing than if there isn’t. And my reasons are just those set out above concerning the perfection-centricity of value. It is better for there to be something perfect if value is perfection-centric than if it isn’t (just as it is better for there to be pleasure if hedonism is true than if asceticism is true). Because it is also true that value is perfection-centric if there is something perfect, SPE follows. However, I reject the existence of a perfect being – there is just too much evil in the world for me to countenance such an entity. I thus reject ASP; I do not think that, for any F, and any Q, however good it is for there to be an F, given Q, it is at least as good for there to be something perfect.

What’s more, we need not search for exotic counter-examples to ASP. The combination of SPE and atheism provides its own counter-example to ASP. That is, the following instance of ASP is false:

Seeing as there is nothing perfect, the unconditional value of there being something perfect is equal to the value of there being something perfect conditional on there being nothing perfect. But, as SPE makes clear, the value of there being something perfect conditional on there being something perfect is greater than the value of there being something perfect conditional on there being nothing perfect. Thus, the value of there being something perfect conditional on there being something perfect is greater than the unconditional value of there being something perfect. This is just to say:

The above thus stands as a counterinstance to ASP – it isn’t perfect for there to be something perfect because it being so requires there to be something perfect, which there isn’t.

We might say, then, that ASP captures something true about perfection on the assumption that there is something perfect. If there is something perfect, then the metaphysics of value centres perfection. And if the metaphysics of value centres perfection, then there’s nothing it’s better for there to be than a perfect being. But there isn’t anything perfect (at least in my view – the great evil in the world, the success of naturalistic world views in accounting for the universe, etc., all speak strongly against the existence of a perfect being) and, as such, the metaphysics of value does not centre perfection. The actual (unconditional) value of there being something perfect, then, is surpassed by the value of there being something perfect conditional on a perfection-centric metaphysics of value (one that does not in fact obtain). The proponent of the evaluative ontological arguments errs in thinking that ASP captures an unconditional truth about perfection when it only captures a truth about perfection conditional on a controversial metaphysics of value, and one that I reject.

Atheist-1 and Atheist-2 have deep disagreements about how best to respond to the evaluative ontological argument, but each plainly represents a coherent and reasoned stance that takes the argument seriously and incorporates what they each perceive to be its insights. Equally, as we have seen, a theist’s acceptance of both ASP and SPE captures a different coherent and reasoned stance that embraces the existence of a perfect being.

Anselm has inspired a startling variety of arguments (the strongest of which are often nowadays taken to be modal ontological arguments – see Hartshorne Reference Hartshorne1962; Kearns Reference Kearns2022; Plantinga Reference Plantinga1974). Reflection on these arguments prompts us to become clearer on issues in semantics, existence, modality, definition, conceivability, and other difficult topics. This new Anselm-inspired argument may not rest on controversial issues in all these areas, but, as we have seen, it forces us to think about the consistency of our evaluative thinking. Proslogion 2 still has much to teach us.