Necessary Suffering and Lewisian Theism

One can readily conceive of worlds of horrendous, gratuitous suffering. Moreover, such worlds seem possible. For classical theists, however, God, amongst other things, is perfectly good. So, the question arises: for classical theists are such evil worlds possible? Many classical theists have said no. This is the modal problem of evil. Herein, I discuss a related problem: the problem of evil worlds for Lewisian theism. Lewisian theism is the conjunction of Lewis’s modal realism and classical theism, and a leading Lewisian theist, Almeida, thinks that Lewisian theists should admit the existence of on-balance evil worlds. I do not. Herein, I present a dilemma for Almeida: either give up God’s sovereignty and the reductionist account of modality or make God blameworthy for evil.


Introduction
We can readily conceive of worlds of appalling misery, whose metaphysical possibility seems, at first glance, eminently plausible. Indeed, we can conceive of a world that contains an infinite number of people who suffer physical agony for eternity. According to classical theism, God necessarily exists. In possible worlds talk, if something necessarily exists, that something exists in all worlds. So, since God exists in all worlds, if worlds of terrible suffering exist, God exists in worlds of terrible suffering. But, according to classical theism, God is taken to be maximally perfect, and so is taken to (necessarily) possess the traditional 'omni-' attributes of omnipotence, omnibenevolence and omniscience. Thus, the question arises: if worlds of terrible suffering exist, and God exists therein, how can a maximally perfect God-who is omnipotent, omnibenevolent and omniscient-permit the existence of such worlds? This is the modal problem of evil. In this paper, I discuss a problem connected thereto, which concerns Lewisian theism-the Preliminaries Possible worlds metaphysically underprop the truth values of modal claims, where such claims are analysed through the following biconditionals for possibility and necessity: (1) Possibly: P iff in some possible world w: P.
(2) Necessarily: P iff in all possible worlds w: P. Now, on Lewis's (1986) modal realism, there exists an infinite plurality of possible worlds, where such worlds are concrete, spatiotemporally and causally isolated individuals. More specifically, worlds are maximal mereological sums of spatiotemporally connected individuals that are of the same kind as the concrete world in which we reside, the actual world. Worlds and their parts are neither spatiotemporally nor causally connected to any world or any part of any world other than themselves and their parts (this is the view that worlds fail to overlap). Worlds, for Lewis, represent possibilities through being composed of parts that instantiate certain qualitative features. Now, modal realism's principle of plenitude entails that every way that a world could be configured is the way that some world is configured. This principle ensures that, in logical space, there are no gaps-that is, the total set of possibilities is accounted for. Such a principle, however, does not tell us what is possible and what is not. Hence, modal realism's principle of recombination: for any individual x, x can co-exist or fail to co-exist with any other individual y. 1 The recombination principle is thence epistemically fruitful: if we are pre-theoretically unsure of the truth of certain possibility claims, calculations guided by this principle may assist us in evaluating the truth of such claims.
Moreover, to explain, supply the truth-conditions for and represent de re possibility predications of individuals (of the form: 'x is possibly F'), modal realism invokes counterpart theory. On counterpart theory, a world, w, represents that 'x is possibly F' by w's having a part, y, which is F, and which sufficiently resembles x. x's being possibly F, then, is represented by a counterpart of x's being F, where counterparts (typically) reside in other worlds. 2 Thus, for de re modal predications, we have the biconditionals: (3) x is possibly F iff some world, w, exists in which x has a counterpart, y, that is F. (4) x is necessarily F iff in all worlds, w, all counterparts of x are F. Now, according to classical theism, God necessarily exists and has all His intrinsic attributes necessarily (that is, essentially), which includes His omnipotence, omnibenevolence and omniscience: so, on Lewisian theism, in all worlds, God exists and is omnipotent, omnibenevolent and omniscient. 3 Moreover, necessarily, God is the creator and sustainer of all that exists (except Himself): so, on Lewisian theism, in all worlds, God is the creator and sustainer of the infinite plurality of worlds. The claim that God is necessarily omnipotent, omnibenevolent, omniscient and the creator and sustainer of all that exists is both a de dicto and a de re modal claim: the claim is de dicto through specifying a necessary condition for membership in the class classical God, and the claim is de re through specifying that the essential (necessary) attributes of God are His omnipotence, omnibenevolence, omniscience and His being the creator and sustainer of all that exists. So, we have the following claim: (5) In all worlds, God exists, is the creator and sustainer of all existence, and is omnipotent, omniscient and omnibenevolent.
With Lewisian theism expounded, let us distinguish between two problems: the modal problem of evil and the problem of evil worlds for Lewisian theism.
The modal problem of evil, owed to Gulserian (1983), can be formulated as follows: The existence of God is incompatible with an intuitively true modal claim: there exist worlds that are, on-balance, evil. 4 Given the intuitive truth of this modal claim, some divine attribute must be revised, rejected or significantly qualified. So, either: classical theists must reject God's necessary existence, and so He does not exist in all worlds-but only in the on-balance good, or perhaps additionally the on-balance neutral worlds -; or, classical theists must revise, reject or qualify omnipotence, omniscience or omnibenevolence. For whichever option classical theists opt, they deny their classical theism.
In more detail, here is Gulserian's argument. By classical theism, in all worlds, God exists and is omnipotent, omniscient and omnibenevolent. Each world is, from its own perspective, actual. By classical theism, God creates all worlds (or, perhaps more moderately: God allows each world to exist). 5 By classical theism, necessarily God creates a world only if its creation is morally permissible. Thus, God's creation of all (metaphysically possible) worlds is morally permissible. But, it is conceivable that there exist evil worlds, and conceivability is standardly taken as being a good guide to metaphysical possibility. And so, it seems metaphysically possible that evil worlds exist, and, by (1), if evil worlds are possible, then there exist evil worlds. However, the existence of evil worlds is incompatible with a necessarily existent God who is necessarily omnipotent, omniscient and omnibenevolent and who is the creator of all worlds. Therefore, at least one divine attribute must be revised, significantly qualified or rejected. Accordingly, Gulserian's argument can be formulated as a hexad of inconsistent propositions: 6 (6) Necessarily, God exists. (7) All worlds are actual from their perspectives. (8) Necessarily, if w is a world, God created w.
(9) Necessarily, if w is a world created by God, it is morally permissible for God to have created w. (10) There exists an evil world, w e . (11) It is not morally permissible for God to create w e when God could have prevented w e 's existence. 7 5 The difference between God's allowing a world to exist and His creating a world is as follows: The former simply supposes that God holds the power to prevent the actualisation (or realisation) of a world, w, through, perhaps, actualising some state of affairs that is inconsistent with w or failing to actualise some state of affairs consistent with w. So, any existent world is allowed to be actual by God. God's creating or actualising a world, however, is the notion that God chooses to causally bring a world into existence. In the literature, it is taken that this is a (semi-) important difference. However, I doubt that there is a morally significant difference between the two notions when considering God and the existence of evil worlds. Whilst it may be true that I am only partially culpable for evil if I knowingly allow it to happen when it is within my power to prevent it, and I am fully culpable for evil if I am directly responsible for it, this difference in scale of culpability does not matter when it comes to assessing God and evil worlds. If God is even partially responsible for the existence of on-balance evil worlds through simply allowing them to be actual, then His omnibenevolence seems denied, and any scale of denial of God's omnibenevolence is enough to render God non-omnibenevolent. Indeed, no divinely preventable evil world-it seems-is compatible with God's omnibenevolence. Thus, to make life easier, I simply speak of God's actualisation or creation of worlds. 6 This hexad is partly owed to Kraay (2011: 362), who partly owes it to Garcia (1984: 379). 7 An anonymous reviewer contends that there is a counterexample to (11). The reviewer writes: …suppose we live in an earthly paradise but freely choose wars, holocausts, you name it, and turn it into a very bad place. God, in giving us real free will, gives us the option of [messing]… things up. The mess we make is our fault, not God's. In securing us the intrinsic good of free will, he has to take his chances. Arguably, then, there are on-balance evil worlds it is permissible for God to create. (11) seems to run into the Free-will Defense.
I can agree with the reviewer here, but I note that the defender of the import of (11) can revise (10) to read: (10*) 'There exists a world, w e , where there is evil in w e that is not subject to a theodicy'. Now the counterexample above does not seem to work against (11), once (10) is revised along (10*)'s lines. I thank the reviewer for pointing this out.
Gulserian contends that classical theists should affirm the truth of (7)-(11), and thence should revise, significantly qualify or reject (6). Now, there is a debate between Gulserian (1983) on one side and Garcia (1984) and Tidman (1993) on another, where both Garcia and Tidman accept (7), (8), (9) and (11), and so believe the battle to lie between rejecting either (6) or (10). Both Garcia and Tidman contend that both classical theists and their critics are equally epistemically justified or rationally entitled in holding their respective modal intuitions, since their respective intuitions are equally supported: for classical theists, this will be the intuition that (6) is true, and for their critics, this will be the intuition that (10) is true. 8 This problem is connected, but is nevertheless somewhat distinct from the problem that this paper concerns. Indeed, the modal problem of evil presents a problem for positing the God in the first place. That is, it can be conceived of as an argument against the existence of God, if generic possible worlds-theory is correct. The problem of evil worlds for Lewisian theism, however, assumes the truth of Lewisian theism, and thus takes for granted the necessary existence of God and thus the truth of (6). This problem, then, concerns whether evil worlds exist on Lewisian theism. The problem, thence, is presented as a battle not between rejecting either (6) or (10), but between rejecting either (10) or (11) on a Lewisian theistic framework. 9 Morris (1987), Oppy (1993), Sheehy (2006), Leftow (2012) and Nagasawa (2017) all reject (10) in favour of keeping the moral intuition that God is incompatible with evil worlds-with Sheehy and Oppy examining this view exclusively on a Lewisian theistic framework. However, Almeida (2008Almeida ( , 2011Almeida ( , 2017b defends the view that Lewisian theists should instead reject (11) in favour of keeping the modal intuition that evil worlds exist. I now present, and thereafter levy an objection against, Almeida's argument for this view.
Almeida's Rejection of (11) Before I start, it is worth noting why the problem of evil worlds is particularly problematic for Lewisian theists. The main reason is that there being real, flesh and blood gratuitous evil seems worse than, say, there being an abstract proposition, as a member of some abstract set of propositions, whose content concerns gratuitous evil. That is, Lewisian ontology is such that if evil is possible, then evil is real-but, conversely, for abstractionist theories of possible worlds (of, say, the Adams (1974) variety or the Plantinga (1974) variety), if evil is possible, then-provided that the actual world is not evil-evil does not affect real, concrete individuals. Perhaps a more perspicuous way to present the point is as follows: On the possible worlds theory where mental images play the world-role and so ground modal truths, there being a mental 8 Kraay (2011: 362) maintains that the appropriate response to the equality of support for (6) and (10) (if equality there be) is that both classical theists and their critics suspend their judgement over (6) and (10). 9 An anonymous reviewer contends that there is no real distinction to be made between both problems mentioned, since 'Both of these problems can just be stated by pointing out that propositions (1)-(11) are mutually inconsistent, so at least one of them must be rejected'. I agree that both problems can be stated by highlighting the mutual inconsistency of (1)-(11); however, the modal problem of evil is taken as a battle between rejecting either (6) or (10), whereas the problem of evil worlds for Lewisian theism is taken as a battle between rejecting either (10) or (11). image of gratuitous evil seems less morally problematic than there being real, concrete evil. Certainly, if my mental images alone ground modal truths, and I conjure an image of gratuitous evil, I therefore make it possible that there be suffering; and, likewise, if my creation of Lewisian worlds grounds modal truths, and I create a Lewisian world of gratuitous evil, I therefore make it possible that there be evil. However, my creating Lewisian worlds that contain real evil seems morally worse than my merely conjuring mental images of evil, even though in both cases I am responsible for grounding the possibility of evil. The same should hold, mutatis mutandis, of God. 10 Notice, also, that one cannot circumvent the problem of evil worlds on Lewisian theism by claiming that all that matters is the actual world, since, on Lewisian theism, other worlds are just as real. Thus, the problem of evil worlds matters even if the actual world is not evil.
Thus, for Lewisian theism, the case of evil worlds is more pressing: whilst some might think that God is not (or, at least, is less) blameworthy for rendering evil worlds possible (provided the actual world is not evil) if their possibility is grounded in His conceiving of evil, most will not be content to not blame God for His rendering evil worlds possible if their possibility is grounded in the existence of real, concrete evil worlds. Almeida, however, thinks that, on Lewisian theism, we should not blame God for the existence of evil worlds.
So, now for Almeida's argument. Almeida (2008Almeida ( , 2011Almeida ( , 2017aAlmeida ( , 2017b thinks that Lewisian theism has the resources to posit evil worlds and do so in a classical theistically benign way. Almeida argues for this view by, firstly, highlighting that logical space (otherwise known, on modal realism, as the pluriverse) is structured the way it is by necessity-it could not have been otherwise. Almeida argues: since logical space is what it is necessarily, evil worlds exist necessarily; and, since evil worlds exist necessarily, and no one can be blamed for the existence of something necessary, God cannot be blamed for their existence. Here is Almeida (2017b: 6): It is false that the totality of God's creation might have included any less evil than it does. It is false as well that any part of the totality of God's creation might have included any less evil than it does. Every instance of evil in every region of the pluriverse-every instance of evil in every possible world-is necessary… There is no gratuitous or pointless evil anywhere in the pluriverse. Indeed, there is no eliminable evil anywhere in the pluriverse.
To raise the plausibility of this view, Almeida (2008: 142;: 7) advances the following lifeguard analogy contra (11): 10 Perhaps one may object and say: there is no significant difference when it comes to God. Perhaps one might say that God's mind should contain no evil images, and so on the mental image modal account, evil should be impossible (although maybe Almeida might object here (see his argument below)), for if His mind were to contain evil images, that would threaten, say, His perfectly beautiful mind. That is, a being with a perfectly beautiful mind should not even be able to conceive of evil. However, if one satisfactorily demonstrated that whichever modal programme classical theists used the problem would remain, using Lewisian theism still provides a vivid and perhaps different way to discuss the problem, and through its ontological machinery, raises Lewisian theistic-specific and not merely general issues for classical theism. And at any rate, this paper concerns Almeida's argument, which is explicitly Lewisian theistic.
(12) A lifeguard, L, has the capacity to rescue from drowning either S or S′, but cannot rescue both. Now, given that it is impossible that L realise a world such that both S and S′ are rescued, it is thereby morally permissible that L rescue S at the necessary price of failing to rescue S′ (and vice versa). Almeida claims that God, on Lewisian theism, is in a predicament analogous to the predicament in which L finds himself: that is, God's creation of evil worlds is morally permissible if their existence is necessary. That is, if evil worlds' nonexistence is impossible, then God cannot be blamed for creating such worlds. To appreciate this, Almeida (2008: 142;2011: 8) proffers the following argument. Suppose that there exists a world, w, that contains a person, Tc, who does not suffer and who is a counterpart of T. Further, let us suppose there exists a world, u, such that, in u, T suffers unremitting, gratuitous pain which God could have prevented without the realisation of either greater good or greater evil (let us call the situation of T's suffering of pain 'Pt'). Let us also, for simplicity, suppose that w and u exhaust logical space for de re predications of T-that us, w and u are the only worlds in which T and his counterparts exist. Now, our intuition is that, in u, God ought to have prevented Pt, since God could have prevented Pt without the realisation of either greater good or greater evil. However, Almeida highlights, Lewisian theism employs the (B)-axiom: 'A→□◊A'. Accordingly, if we apply the (B)-axiom to Pt, we get: '□◊Pt'. But, if '□◊Pt' is true, then it is necessary that T-or Tc-suffer gratuitous, but preventable pain. Thus, were God to prevent, in u, the suffering of T, God could not prevent, in w, the suffering of Tc 11 -by the (B)-axiom, it is necessary that either T or Tc suffer preventable pain. That is, whilst preventing either T's or Tc's suffering is possible, the prevention of the suffering of at least one of them is not. Indeed, argues Almeida, logical space is necessarily shaped how it is. Thus, Almeida contends: our supposition that God ought to prevent Pt is misguided-God's predicament is thence analogous to L's. 12 Now, Almeida thinks that if he is correct, Lewisian theists can preserve both their Lewisian and their classical theistic intuitions: God, on this picture, satisfies classical theism, and the configuration of logical space, on this picture, satisfies modal realism. I now argue, however, that Almeida's argument fails: his argument fails to satisfy the Lewisian theist.

A Dilemma for Almeida
I shall ascend to my dilemma for Almeida by way of considering an objection to his view owed to Kraay (2011): Almeida's argument requires that theists revise their 11 Here, one may note that this counterfactual is a counterpossible, and so is vacuously true. For now, I partition such discussion, but later I discuss a connected problem for Almeida. 12 An anonymous reviewer writes: …it seems to me that Almeida's reasoning is as follows: We ask Almeida, Why does God actualize Evil World? Answer: Because Evil World is necessarily possible. Then we ask: Why is Evil World necessarily possible? Answer: Because, given that it's already actualized, it follows that it's necessarily possible. Circular.
I agree with the reviewer that this is a worry for Almeida's view; indeed, this worry motivated my dilemma for Almeida. understanding of the divine nature. Now, Kraay accepts that if evil worlds exist necessarily, then God cannot be blamed for their existence. However, Kraay (2011: 364) argues: This is a highly surprising and counterintuitive consequence for theism. Theists have generally held that no possible amount of (divinely permitted, divinely preventable) gratuitous suffering is compatible with theism. But on Almeida's view, every possible amount of such suffering is compatible with the existence of the Anselmian God. Accordingly, Almeida's move requires theists to dramatically revise their understanding of God. To the extent, then, that theists are committed to the traditional view concerning gratuitous evil and God, Almeida's view will be considered costly.
To assess Kraay's objection, I shall take 'gratuitous suffering' to mean suffering that lacks teleological qualities; that is, the suffering is 'pointless'. If Kraay's objection concerns gratuitous suffering of this kind, Almeida faces a problem, I argue. Now, generally, God is held to not permit 'pointless', nonteleological suffering. However, on Almeida's view, as Kraay notes, all possible amounts of pointless, non-teleological suffering are consistent with God. This certainly seems-pro Kraay-a revisionary claim. Indeed, Almeida must explain why an omnibenevolent God created and sustains infinite worlds of real, concrete, pointless suffering. Now, there are two responses available to Almeida. The first is as follows: 'There does not exist an infinite amount of pointless suffering. Recall that God could only save from suffering either T or Tc but not both, since it is necessary that one suffer, since logical space is necessarily how it is. And so, the preventable and undeserved suffering of T serves the point of allowing Tc not to suffer'. Certainly, this response could be supported by Almeida (2017b: 6): "There is no… pointless evil anywhere in the pluriverse". The second response available to Almeida is as follows: 'There exists an infinite amount of pointless suffering (like T's), but God cannot be blamed therefor, since logical space is necessarily how it is.' Certainly, this response could be supported by Almeida (2011: 8): '…it's necessary that some morally equal counterpart (or other) of [Tc] endures needless suffering'. Here, 'needless' reads like 'pointless'. So, the first response asserts the non-existence of pointless suffering, and the second concedes the existence of pointless suffering, but asserts that God cannot be blamed therefor.
However, both responses, to prevent necessary revisions of God, I argue, can only be provided once a different, substantial revision of God is made. Kraay, then, is right about one thing: Almeida's view does require theists 'to dramatically revise their understanding of God'. However, their revision of God is not required for the reason Kraay cites. Specifically, Almeidians, who favour either of the two response-types above, are committed to rejecting God's sovereignty.
Take the following intuitively true biconditional concerning existential dependence: (13) Necessarily: x existentially depends upon y iff x would not exist were y not to exist. 13 Now, take a necessary feature of the doctrine of divine sovereignty: 14 (14) Necessarily: God is the being upon whom all existence depends. 15 A consequence of the conjunction of (13) and (14) is that were God not to exist, nothing would. Now, recall (3): x is possibly F iff a world, w, exists such that in w, x has a counterpart that is F. On the assumption that w and u exhaust logical space for de re predications of T, (3) implies that T possibly suffers iff there exists a world in which T or Tc suffers. So, if the right-half of this biconditional were false, it would be impossible that T suffer. Moreover, recall: since all existence depends upon God, and worlds exist, all worlds depend upon God. And, finally, recall: modal truths depend upon worlds-that is, modal truths are parasitic upon, and so are less fundamental than, worlds. So, modal truths depend upon God.
Here is where my objection to Almeida lies. If God failed to create a world in which either T or Tc suffers, then by (3), it would be impossible that T suffer. That is, were T's world, u, one in which T does not suffer and were Tc's world, w, one in which Tc does not suffer, and w and u exhaust de re predications of T, then T would lack the de re modal property of possibly suffering-it would be, by (4), true that T necessarily does not suffer. Certainly, Almeida is right that were God to have created a world, u, in which T suffers and a world, w, in which his counterpart, Tc, does not suffer, then, by the (B)-axiom, it would be true that, necessarily, either T or Tc suffers (that is: Pt, 'A→ □◊A'; □◊Pt). However, it would not be true that, necessarily, either T or Tc suffers if God were to not create a world in which either T or Tc suffers. But, here we see that Almeida's claim that since evil worlds exist necessarily, God cannot be blamed for their existence is problematic. For evil worlds to exist necessarily, God must first create them. Indeed, modal truths, for Lewisian theists, depend upon worlds; and, all existence, for theists, depends upon God; therefore, for Lewisian theists, worlds, and so modal truths, depend upon God. Here, then, we see that God's predicament is morally distinct from a L's. L's position has been foisted upon him, and so L must choose between saving S or S′ in the knowledge that either S or S′ will die. God, however, creates worlds and so determines modal truths-that is, He determines that T possibly suffers or not. L would clearly be morally responsible for the drowning of either S or S′, if L put both S and S′ into a drowning situation, and we would not consider absolving L 13 Some may object that, in (13), if y stood for God, then the counterfactual in (13)'s right-half would constitute a counterpossible, and thus be vacuously true. I think, however, that like with many intuitively meaningful counterpossibles, this result should push us more to reject Lewis-Stalnaker semantics for counterfactuals than to claim that seemingly meaningful counterpossibles are in fact meaningless. See Leftow (2006) and Davis (2006) for discussion of God and counterpossibles, and Lewis (1973) and Stalnaker (1968) for discussion of counterfactuals. 14 For support and/or discussion of divine sovereignty and aseity, see Plantinga (1980), Davison (1991) and Davidson (1999). 15 The type of dependence-relation at play here, in the case of individuals, is causal. For abstracta, some (e.g. theistic activists) take the dependence-relation as causal (see Morris and Menzel 1986 for a defence), whereas others construe the dependence-relation in different ways (see Davidson 2019 for an overview). from blame if L responded: 'but, in the situation, at least one had to die'. No, L, you put both S and S′ into a drowning situation, when there was nothing outside you making you do so. God's position, however, is exactly this: He created T and Tc's respective worlds-He determined the shape of logical space, and there was nothing outside Him that made Him do so. So, if Almeida wishes to retain the sense in which God is not blameworthy for His creation of evil worlds by virtue of their being necessary (that is, if Almeida wishes to not beg the question against those in favour of rejecting (10) instead, and hold that there is an analogy between God's predicament and L's), he is compelled to do the following. Almeida must deny both (14) and the Lewisian theistic claim that modal truth is parasitic upon worlds by taking necessary truths (like those concerning the shape of logical space) to be independent of God and more basic than worlds, and thus to determine which worlds God creates. That is, Almeida must deny both the classical theistic conception of God (specifically, God's sovereignty) and the Lewisian conception of the role of worlds. So, I argue, Almeida is committed to being neither a classical theist (through requiring a violation of God's sovereignty) nor a Lewisian.
And so, we can see that both the above responses to Kraay's objection concerning pointless gratuitous suffering fail. According to the first response, it is necessary that either T or Tc suffer, and so T's suffering serves the point of allowing Tc to not suffer, and thus there exists no pointless suffering in logical space. And according to the second response, the existence of pointless suffering is conceded but God is blameless, since logical space is how it is necessarily. In both responses, we see that we are only permitted to absolve God from blame for creating evil worlds of pointless or nonpointless suffering if God Himself does not structure logical space. That is, we can only absolve God from blame if we think that the shape of logical space comes prior to, and so supplies the 'rule book' for, God's creation of worlds. This, however, incurs a violation both of classical theism's doctrine of divine sovereignty (expressed by (14)) and of modal realism's reduction of modality. So, in the first response, the only 'point' that T's suffering would serve would be to realise the shape of logical space; but, given that classical theists believe that (14) is true and that God is necessarily omnibenevolent, we have very good reason to think that classical theists should not think that God would structure logical space in a way that allows T to suffer, even if T's suffering serves the purpose of allowing Tc to not suffer, which simply serves the purpose of realising the shape of logical space.
So, I argue, Almeida faces a dilemma: either he must violate Lewisian theism's doctrine of divine sovereignty and its reductionist account of modality, and so he absolves God from blame for creating evil worlds; or, he affirms Lewisian theism's doctrine of divine sovereignty and its reductionist account of modality, but allows God to be charged with blame for creating evil worlds. Both horns lead to a revision of God (but not in the way that Kraay thought), and thus a revision (and so strictly speaking a rejection) of classical theism. So, theists should steer clear of Almeida's response to the problem of evil worlds on Lewisian theism.
But, here, Almeida may demur, and contend: 'Evil worlds exist necessarily in the sense in which numbers and the rules of logic exist necessarily: they are existentially dependent upon God, but there is no choice for God as to which numbers or which rules of logic exist. So, we can retain modal realism's claim that modal truth is parasitic upon worlds and classical theism's claim that (14) is true: worlds and modal truths depend upon God, but there is no choice for God as to which worlds exist. That is, similarly with how it is in God's nature that numbers exist, it is in His nature that evil worlds exist-and He has no choice over His nature, and so cannot be blamed for their existence'. Now, one may initially think that a theologically problematic consequence lurks: 'If there is no choice for God over which worlds exist, and all reality is simply the pluriverse, we lose another important divine attribute: God's freedom. God, on this picture, has no freedom with regards to the shape of reality. So, whilst we may think that God is not free with respect to which numbers or which rules of logic exist, and this may not seem so problematic, if we think that God is not free with respect to any part of reality, then God is not free with respect to anything. Lewisian theists, therefore, opt for this strategy on pain of rendering God unfree'.
I think this counter-objection can be rebutted. Certainly, classical theists do not wish to hold that God is unfree regarding the existence of numbers; even if it is not strictly God's choice that numbers exist, they wish to claim that God is nevertheless free. One way that classical theists can divorce freedom from choice amongst options is by eschewing an alternate possibilities model of freedom, wherein agents are free by virtue of their selecting actions from possible alternates (which seems to underpin the above divine freedom counter-objection), and instead adopting a source-hood model of freedom, wherein agents are free by virtue of their actions arising out of or originating in themselves. If Almeida adopts this divine freedom model, he can blunt the above divine freedom counter-objection.
But, there is, however, a more crucial objection to the above response to my dilemma: Whilst it may be true that we cannot blame God for His nature, and so we ought not blame Him for His creation of evil worlds, we can (and should) certainly think that His nature is not maximally perfect. For classical theists, however, God's nature is maximally perfect. If we found in God's nature an aspect that ordained the existence of infinite on-balance evil worlds of terrible suffering, we would most likely think that such a nature was not maximally perfect: we could conceive of a more perfect being-that is, one that did not have a nature such that the existence of terrible suffering is ordained. So, since classical theists believe that God's nature is maximally perfect, they should not adopt the above response to my dilemma.
However, one may think that even though this response is not tenable, it shows that my above consequence for Almeida that modal truth must be prior to God's creation of worlds if God is not to be rendered less than maximally perfect does not follow. 16 However, this is no gain for Lewisian theists, since I can concede this, and present the dilemma in the following way. If Almeida wishes to retain the sense in which God is not blameworthy for evil worlds by virtue of their existence's being necessary, he must conceive of Lewisian theism in one of the two following ways: First: On Lewisian theism, God creates worlds, but the theory is not modally reductionist; modal truth assumes primacy over, is more basic than, or is prior to, worlds and God's creation thereof (that is, modality is primitive). So, '□◊Pt' is true in the absence of God's creation of the pluriverse; and given that the truth of '□◊Pt' assumes primacy over, is more basic than, or is prior to God's creation of a world, u, at which Pt befalls, God does not incur blame for His creating T's suffering-world, u. Or second: On Lewisian theism, modal facts are parasitic upon worlds, and God creates worlds and thus the 16 Indeed, I thank an anonymous reviewer for pushing me on this point. pluriverse such that '□◊Pt' turns out true, and His nature determines that He does so. I have argued, however, that adopting the first strategy towards evil worlds not only commits Lewisian theists to rejecting theism through rejecting the doctrine of divine sovereignty, but also commits Lewisian theists to rejecting modal realism through rejecting modal reductionism. Almeida's Lewisian theism, then, would be neither Lewisian nor classically theistic. And, adopting the second strategy commits Lewisian theists to a denial of the classically theistic view that God is maximally perfect. Whilst it might be Lewisian, Almeida's Lewisian theism would not be classically theistic. Thus, we have the dilemma: either God is blameworthy for evil or He is not-but if not, then, either He is not sovereign and modality is primitive, or He is not maximally perfect. Since God should be maximally perfect and not be blameworthy for evil, and modality should not be primitive, Almeida has a problem.