The Skeptical Challenge of the Theistic Multiverse

1. The Laxity of Consequentialist Requirements in Light of the Multiverse Possibility

In this and the next section, I argue that when we give due consideration to the possibility that God has created infinitely many universes, we have reason to think that the standards of acceptability that a universe must meet in order to be worthy of creation are much laxer than we might initially suppose. This section focuses on “consequentialist” considerations, roughly understood as considerations that are concerned only with the total (actual or expected) value resulting from a given action. In Section 2, I consider the implications of “deontic” requirements that may bear on God’s creative activity.

A universe, as I shall use the term, can roughly be understood as some causally isolated “realm” of concrete reality that is part of the one actual world.² For example, suppose that the world includes two different physical realms, each of which is governed by a distinct set of physical laws, that these physical realms bear no spatial relations with one another, and that no events internal to one physical realm can influence the events internal to the other physical realm. This scenario would certainly qualify as a “multiverse” scenario—i.e., a scenario where the world contains multiple universes. The notion of universe as I have characterized it is admittedly vague, but I think it will suffice for present purposes. Traditional theists hold that God—a supremely perfect personal being—is the creator and sustainer of all concrete reality outside of Godself. So God, on this view, is responsible for every universe that exists, whether there be only one or many.

Since creation is understood to be an intentional act of a being perfect in power, rationality, and goodness, it is reasonable to assume that the created reality must satisfy certain standards of goodness or acceptability. Many philosophers have argued that we can be fairly confident that the standards of acceptability that would guide a perfect God’s creative act would rule out any world that contains the sort of suffering that we experience or observe around us. Whatever the standards of acceptability may be, they are not so lax so as to tolerate the amount and intensity of the suffering that is in fact experienced by humans and non-human animals.

Following others, I will argue that this claim owes much of its plausibility to an implicit assumption that our universe comprises all of concrete reality.³ When we carefully reflect on the possibility that God has created an infinite number of universes, we cannot dismiss the possibility of standards that are much laxer than the standards presupposed by atheistic arguments from evil.

Plausibly, we can think of God’s act of creation as a decision to “actualize” one or more “universe plans,” where a universe plan characterizes God’s “causal” input into the nature and history of any universe described by that plan. Such a plan would presumably specify some set of initial conditions and laws for the universe, but it would also specify policies for divine intervention. The laws might be deterministic, or they could be probabilistic but with some or all of the outcomes specified in the universe plan, or they could be probabilistic with some or all of the outcomes left up to chance. And the universe plan might allow for creatures with libertarian free will who can influence events in the universe in ways not determined by the plan. I do not make any assumptions about whether, for any created universe, God has perfect foreknowledge of all God’s causal inputs into that universe. If God does not have perfect foreknowledge of the course of events in some universe (e.g., one where there are creatures with libertarian free will), the plan for that universe may still include conditional policies specifying how God would act in various possible circumstances. Such policies could range from highly general policies that apply across a wide range of circumstances to highly specific policies that apply to some specific situation.

Let’s consider three categories of universe plans. First, let’s say that universe plan P is a rubbish plan if and only if in a hypothetical situation where God was forced to choose between creating a universe according to plan P or creating nothing, God would have most reason to create nothing. Next, let’s say that universe plan P is faulty if and only if (i) it is not rubbish and (ii) it permits at least one instance of harm to a creature that would be “pointless.” In this context, some potential harm H that would be permitted by plan P1 counts as pointless if and only if there is a plan P2 that is otherwise similar to P1 except that it does not permit H (or any comparable harm in its place) and a P2 universe is expected to be intrinsically better than a P1 universe. So, to borrow an example from William Rowe (1979), consider a universe plan that allows for the possibility of a fawn experiencing a slow and painful death resulting from forest fire burns. Now consider a second universe plan that is just like the first except that it says that God miraculously turns off the conscious experience of any fawn dying from forest fire burns, so that no pain is experienced. If a universe corresponding to this second plan would be intrinsically better than a universe corresponding to the first plan, then the first plan permits harm that is pointless in the relevant sense. Finally, those universe plans that are neither rubbish nor faulty we will call pristine plans. (Note that while a pristine plan does not permit harms that would be pointless in the sense defined above, a pristine plan could permit a great many harms that are not pointless.) I will sometimes speak of universes (and not just universe plans) as being rubbish, faulty, or pristine. A universe is one of these types just in case it results from a plan of the relevant type.

Initially, let’s consider a hypothetical situation where God can create only one universe. Clearly, God would not create a universe corresponding to a rubbish plan. Such plans would be unworthy even if no other creative options were available. We can therefore restrict our focus to the pristine plans and the faulty plans. Plausibly, many of these plans would necessarily be deemed unworthy for the consequentialist reason that the corresponding universes would not be expected to contain sufficient value. For example, it is plausible that God would not (as a matter of necessity) choose a plan that is expected to result in some trivial amount of value. If there is a maximum achievable level of value for a single universe, it is plausible that God would, of necessity, choose a plan that corresponds to some such universe.⁴ Even if there is no maximum achievable value, so that no choice could be supported within a maximizing consequentialist decision theory, it’s still plausible to think that God would necessarily choose a plan that would result in a sufficiently “large” amount of value.⁵ If there are consequentialist reasons or principles that would necessarily rule out God’s choosing certain universe plans in some real or hypothetical scenario, then I will say that there is a consequentialist “constraint” or “requirement.” The language of constraint or requirement is meant to convey the metaphysical necessity that is involved, and is not meant to imply that God is somehow bound by extrinsic requirements.

A consequentialist requirement alone would not, it seems, rule out the possibility of God creating some faulty universe. Suppose U1 is some pristine universe that is not ruled out by the consequentialist requirement. Even if the total value contained in U1 is infinite, it seems that there will be finitely bounded “locations” in U1 where the local value is finite.⁶ (Some possible ways of thinking of the locations might be finite physical locations over some finite duration or perhaps individuals over some finite duration.⁷) Now consider some other universe U2 that has the same locations as U1 but where the value at one location is lower because of some pointless harm H. While U2 has lower value at this location, this could be compensated for by U2 having significantly more value at one or more other locations. Thus, there is no reason to think that U2 must rank lower than U1 in terms of consequentialist considerations. If it is possible to have value-based preferences between universes of infinite value, then it seems possible that U2 could rank more highly than U1. It is plausible, then, that for every pristine universe plan, there is a faulty universe plan that is just as good from a purely consequentialist perspective.

Let’s now consider what universes would be worthy of creation when God is not restricted to creating only one universe. Because God can without cost create an infinite number of universes, the relevant standards of acceptability that a universe plan must meet to be worthy of creation are arguably much more permissive. This would certainly seem to be the case with respect to constraints based on consequentialist considerations. In the single-universe scenario, it seems that it would not be fitting for God to create a universe whose expected value is barely positive. But choosing to create such a universe seems perfectly reasonable when God is able to create a multiverse, since creating such a universe in no way prevents God from also creating universes that would be expected to contain more value. (This is essentially the Augustinian insight described in the introduction.)

If this is right, then the argument from evil arguably fails on the supposition that God is a consequentialist. Consider a consequentialist who is considering whether it is plausible to think that God created our universe. Assuming that the same scientific laws and broad regularities hold throughout this universe, it is likely that living things that arise in other parts of the universe suffer in broadly the same way as living things on Earth. Where there are living creatures, we can expect death, predation, and fierce competition for essential resources to be ubiquitous. Recognizing this, it is tempting to think that a consequentialist God would not have created a universe like this one since there are (it might be thought) alternative universes that are so much better. Even allowing that a maximally good universe may not be possible, it might seem likely that a universe like ours would fail to make the cut. The reasoning is mistaken because it assumes that the character of our universe reflects the character of all created reality. But if our universe is but one member of an infinite multiverse, then the question is not whether God should have created a much better universe (since maybe God has), but only whether our universe might reasonably be expected to add value (or at least not diminish it) when it is added to the ensemble of universes that make up the multiverse. In light of the multiverse possibility, it would seem that the argument from evil cannot be advanced on consequentialist grounds unless one holds that our universe is a rubbish universe that contains more bad than good.

Bradley Monton (2010: 130) argues that this sort of response to the problem of evil fails once we realize that God could create arbitrarily many universes of any given type. Monton’s argument comes toward the end of a paper that is not principally focused on the problem of evil, but rather on the “problem of no best world.” This problem for theism begins with the plausible claim that for every possible created order, there is an even better possible created order. From this it follows that no matter what God creates, God does less well than God could have done. This conclusion is thought to threaten the claim that God is morally perfect and perfectly rational. In response to this problem, some have appealed to the multiverse possibility in order to challenge the “no best created order” premise (O’Connor 2008; Kraay 2010; Hudson 2013): the best (or at least an unsurpassable) created order, it is claimed, is the multiverse which contains every universe whose level of (actual or expected) goodness exceeds some relevant threshold, or perhaps a multiverse that exhibits in a suitable manner every valuable type of universe. Against this multiverse solution to the no best world problem, Monton offers the following argument: for any good universe type, creating an additional token of that type will always add to the value of the multiverse; there is no upper bound on how many duplicates of a given universe God can create (since there is no maximum cardinality for infinite sets); thus, any multiverse God creates could be surpassed in value by a different multiverse that includes more duplicates of good universes.⁸

Even if the multiverse possibility does not help to address the problem of no best world (and I am inclined to agree with Monton that it does not), it may nonetheless help to address the problem of evil if, as I am arguing, the standards of acceptability for a universe are lower in a multiverse scenario than in a single-universe scenario. But Monton thinks that in light of God’s option to create an unlimited number of duplicate universes, multiverse responses to the problem of evil will also fail. Consider the following universes: U is a hypothetical universe whose value would barely be positive, perhaps because many of the goods it would contain would be offset by pointless harms; Nice-U, on the other hand, is a hypothetical pristine universe whose total value would be extremely high. While U is not a rubbish universe, let’s stipulate that in a hypothetical scenario where God was limited to creating only one universe, God would have decisive consequentialist reason to refrain from creating it and to create something better instead. And let’s further stipulate that God would be justified in creating Nice-U in such a single-universe scenario. According to the multiverse theodicy described above, a God who is a consequentialist may reasonably and justifiably create U as part of a wider multiverse, since doing so would add to the total value of creation. Monton (2010: 130) objects to this reasoning as follows:

Instead of creating U, God could simply create a duplicate of Nice-U. It’s true that that God could add to the goodness of reality by creating U as well, but God could add to the goodness of reality even more by creating a duplicate of Nice-U. God would never feel compelled to create U, since he could instead create another duplicate of Nice-U, and since no matter what universes God creates, God will never be able to achieve the greatest possible amount of value.⁹

First, a small quibble: the multiverse theodicist needn’t argue that God would “feel compelled” to create U. All that they need to argue is that God has justification for creating U and that God might do so. So let’s revise Monton’s argument to hold that God would never create U since God “could instead create another duplicate of Nice-U.”

The problem with this argument, in my view, lies in the word “instead.” The claim is that God should duplicate pristine universes instead of creating faulty universes. But it is appropriate to say that someone should choose to do X instead of Y only if X and Y are mutually exclusive choices (at least given certain background conditions) or if choosing to do X without doing Y is for some reason better than choosing to do both X and Y. For example, it is not appropriate to criticize someone for smiling at a toddler in the grocery store by saying that he could have given money to the local orphanage instead. Because there is normally no need to choose between such actions, the merits of one these actions typically has no bearing on whether the other action should be done. Granted, there are unusual situations where one might find oneself choosing between these actions. Suppose I commit to doing exactly one extra thing this year to benefit local children, and settle on smiling at the next toddler I encounter. Given that I have framed my choice in these terms, my smiling at the toddler could aptly be criticized on the grounds that giving to the local orphanage would have been much better. But when we consider God’s decision of what universes to create, there is no reason to think that God’s decision to create some faulty universe is a decision God reaches by choosing between the faulty universe and a pristine one.

Of course, beings who are finite in abilities and resources can only do so many things, and for this reason even apparently unrelated possibilities for action are in some sense pitted against one another as competing options. But given God’s unlimited cognitive and creative resources, the presumption should be that the choice of whether to create a universe of a given type—and how many of that type to create—is completely independent of choices that concern other universe types. Someone who is moved by Monton’s duplication objection may implicitly be imagining that God creates a universe “one at a time” in some sort of ordered sequence. In this case, at every stage in the sequence, God must choose between creating U and various other creative possibilities, including Nice-U. But a God with unbounded power and knowledge would not have to execute the creative activity in this sequential way. A more adequate mental model is one where, in a single decision, God specifies for every possible universe plan how many universes of that type will be created. On this way of framing the creative act, it is clear that creating U-type universes in no way trades off with creating universes of any other type.

2. The Laxity of Deontic Requirements in Light of the Multiverse Possibility

Thus far, our discussion has attended only to consequentialist considerations pertaining to how a given universe may add to or detract from total value. I have argued that absent any upper limit on the number of universes that can be created, God has justification to create universes that are not valuable enough to be creation-worthy in a hypothetical single-universe scenario.

While the multiverse possibility might give God justification to create universes that in a single-universe scenario would be excluded on the basis of their low degree of value, one might think that universes that are excluded in a single-universe scenario for non-consequentialist moral reasons would continue to be excluded in a multiverse scenario. In other words, even if the consequentialist requirements for a universe are more permissive in a multiverse context, it might seem that the deontic requirements are equally stringent. In particular, it is initially plausible to think that because God has a moral reason to prevent gratuitous harms, God would avoid creating faulty universes whether or not God creates just one universe or infinitely many.

I believe that this supposition is mistaken, however. Given how I have defined a “pointless” harm, we know that when a pointless harm is permitted to occur in some universe, that particular universe is expected to be intrinsically worse than some otherwise similar universe where the harm is not permitted. But it does not follow from this that a multiverse containing a pointless harm is worse than an otherwise similar multiverse that lacks the harm. We cannot assume that the intrinsic value of a universe fully determines how its inclusion in the multiverse affects the total value of the entire created order. Let’s say that a potential harm H that is permitted to occur in possible created order O1 (where a “created order” encompasses all of created reality) is gratuitous if and only if there is an otherwise similar possible created order O2 where H is not permitted and that is expected to have greater value than O1. In a scenario where the created order consists of a single universe, any harm that is pointless will also be gratuitous. But as I will now argue, pointless harms might not be gratuitous in a multiverse context. So even if we suppose that God is morally obligated to prevent gratuitous harms and thus cannot justifiably create a faulty universe in a single-universe scenario, God may have justification to create faulty universes in a multiverse scenario.

One way that this could be the case is if God must create faulty universes in order to bring about a world that suitably exhibits the diverse types of value that are worth pursuing (Schrynemakers 2015: 143–45; O’Connor 2008). It is plausible that the value of the created order partly depends on the diversity of kinds exhibited in that order.¹⁰ Historically, a number of prominent theistic thinkers have affirmed some sort of principle of diversity or “plenitude” that says that creation better expresses God’s perfection when it contains a great diversity of kinds of creatures and when there are no gaps in the “chain of being” that spans from higher and more perfect beings to lower and less perfect beings.¹¹ Aquinas, for example, claims that the value of creaturely diversity is significant enough that a world consisting of two angels is worse than a world consisting of one angel and one stone. This is because “the perfection of the universe is attained essentially in proportion to the diversity of natures in it, whereby the diverse grades of goodness are filled, and not in proportion to the multiplication of individuals of a single nature.”¹² Given this sort of view, there is reason to think that a multiverse that lacks gratuitous evils may contain faulty universes that include harms that would be gratuitous in a single-universe context.

To see why, grant for the sake of argument that our universe is faulty and ridden with pointless evils (where “pointlessness” is understood according to the definition in Section 1). Now consider a series of progressively better universe plans that improve upon the plan for our universe by disallowing one or more harms that our universe permits. We might start, for example, by considering a universe plan that is similar to ours except it requires miraculous intervention to prevent pointless end-of-life pain like that experienced by Rowe’s fawn. Such a plan would, presumably, still contain many instances and kinds of pointless suffering. So the series of improvements continues to a third type of universe plan that prevents yet another kind of pointless harm and is even better yet. If this universe plan still permits a pointless harm, then we consider a further improved plan, and so on until we finally reach some pristine universe plan. Assuming that the “better than” relation is transitive, this pristine universe plan will be better than the plan that characterizes our universe. So in a single-universe scenario, God would presumably prefer a universe corresponding to the pristine plan over a universe like ours.

But what should we say about a scenario where these two plans were the only plans available to God and where God could create two universes? Would it be better for God to create two universes corresponding to the pristine plan, or to instead create one pristine universe and one faulty universe like ours? Note that if atheistic proponents of the argument from evil are correct, then our universe is rife with pointless evil, and is thus very different in character from a pristine universe that would be reached through progressive improvements of our universe plan to eliminate pointless evils.¹³ But this means that a God who valued diversity in creaturely kinds might have reason to prefer a universe ensemble consisting of our universe and the pristine universe to an ensemble consisting of two instances of the pristine universe. In our universe, the kinds of lives we live and the kinds of relationships and communities we form are deeply shaped by the realities of evil and suffering around us. Beings who lived in a pristine universe without such suffering would be quite unlike like us, even if they were the same biological kind. A God who cares about diversity not only of natural kinds, but who also cares about diversity in the kinds of narratives that characterize creatures’ lives, the kinds of communities they form, the kinds of virtues they exhibit, and so on, may very well prefer to include some universe like ours in the multiverse even if our universe is intrinsically worse than pristine universes where many of the kinds of harm we encounter in this universe are prevented from occurring.¹⁴

One way of putting the point is that in a multiverse scenario, the harms permitted by some universe plan can have a justifying purpose that they cannot have in the single-universe scenario. While a pristine universe that is better than ours may have as much intrinsic diversity as our universe, such a universe may do less to contribute to overall diversity in a context where there are already pristine universes but no faulty universes exhibiting the kinds of creatures and creaturely narratives that require suffering. Harms that do not yield a diversity benefit in a single-universe scenario may yield such a benefit in a multiverse scenario.¹⁵

Doubts may, of course, be raised about whether diversity in the created order is really valuable in itself. In the final section, I will explain why, even if diversity in universe types is not a good in itself, such diversity might have significant instrumental value as a precondition for maximizing the number of creatures who are genuinely special to God. But even before this explanation is given, I hope that the discussion thus far shows that it is at least somewhat plausible that the standards of acceptability for a universe are significantly more permissive in a multiverse context than in a hypothetical single-universe scenario, and that this could be true even if God’s creative choices are subject to deontic as well as consequentialist requirements. I will now turn to the skeptical worry that arises if indeed the standards are more permissive as I have argued.

3. From Laxity of Standards to Skepticism

Thus far, I’ve argued that in light of a multiverse possibility, the standards of acceptability that a universe must meet to be worthy of creation may be much laxer than is often supposed. In this section, I argue that this alleged laxity of standards generates significant skeptical worries for the theist.

To facilitate my argument, I provide the following definitions:¹⁶

Subject S is a recent internal duplicate of yours if and only if S undergoes some interval of conscious experience that is internally indistinguishable from your recent conscious experience. (For sake of this discussion, you count as your own recent internal duplicate.)

Universe U is epistemically inhospitable for you if and only if you have at least one recent internal duplicate in U and a large portion of your recent internal duplicates are radically mistaken about the past, present, or future (or at least are so mistaken during the interval where their experience mirrors your recent experience).

Universe U is epistemically hospitable for you if and only if you have at least one recent internal duplicate in U and U is not epistemically inhospitable.

It may be helpful to list some examples of epistemically inhospitable universes where most of your recent internal duplicates (or your only recent internal duplicate, if there is only one) are radically deceived about the past, present, or future. In the following descriptions, let S stand for a recent internal duplicate of yours:

• S is a “Boltzmann Brain,” an isolated brain formed by chance (either by quantum fluctuation or gradual particle accretion) in an otherwise empty region of space.¹⁷ In S’s universe, the large majority of observers with human-like experiences are deceived Boltzmann Brains that occasionally come into existence in the eons after all of the stars have died.

• S is in a universe that has existed for only five years, and which came into existence in a state exactly like the state of our universe five years ago.

• S is in a universe where, at the stroke of midnight, the laws of physics will undergo a radical shift, resulting in the immediate disintegration of all material objects in the universe that are more complex than a hydrogen atom.

• S is in a universe that looks just like our own but where most animals and human beings are philosophical “zombies” who lack consciousness.

• S exists in a universe that is composed of conscious “monads” that are causally isolated despite its appearing to them that they interact with other beings. The experiences of these monads are not harmonized in any way; each monad lives out a “story” that bears no regular connection to the story of any other monad.¹⁸

To see the skeptical worry posed by a divinely-created multiverse, consider first the possibility that God is a consequentialist who is justified in creating any universe whose total value is expected to be positive. Arguably, many conceivable universes that are epistemically inhospitable for you would make this cut. For example, the five-year-old universe described above is arguably good on the whole even if it is inferior to a universe where apparent evidence of a distant past is not misleading. Indeed, I think that any of the universes described above could be good universes, at least when the rest of the details are filled out in an acceptable way. The bad features of these universes, including the bad of radically mistaken beliefs, could easily be outweighed by significant goods.

The nub of the skeptical concern, however, is not the sheer number of conceivable universes that are good but epistemically inhospitable. The principal concern is that the theist arguably has reason to think that inhospitable universes predominate among the created universes that contain a recent internal duplicate. To argue for this, let’s make the simplifying assumption that each universe plan fully determines the internal history of any universe created according to that plan. Given this assumption, for each universe plan, it is determinate whether a universe that corresponds to that plan would be epistemically hospitable for you, epistemically inhospitable for you, or neither (if such a universe would not contain a recent internal duplicate).¹⁹ We can thus speak of universe plans as being epistemically hospitable or inhospitable for you (or neither) based on what sort of universe would result from God actualizing the plan in question. If God exists and is a consequentialist, then presumably the plan for this universe must be among those plans that a consequentialist God may justifiably actualize and that are either epistemically inhospitable or hospitable for you. Let’s call these the live plans. Arguably, epistemically inhospitable plans are predominant among the live plans. This is because the range of ways that universe can be inhospitable for you is wider than the range of ways that a universe can be hospitable for you (even when we restrict our focus to overall good universes).²⁰ For a universe to be hospitable for you, your recent internal duplicate in that universe (or a large portion of such duplicates, if there be more than one) must be broadly correct in their basic outlook concerning the character of the past, present, and future. This requirement of correctness places fairly demanding constraints on the character of the universe. For example, it would have to be that all (or nearly all) normally functioning adults are conscious, that the physical laws will remain constant far into the future, that the universe’s history extends billions of years into the past, and so on. The constraint of inhospitableness, however, is nowhere near as demanding. All that is required for inhospitableness is that the universe contain a recent internal duplicate and that the universe fail in some radical way to correspond to the beliefs held by your duplicate (or by the majority of your duplicates, if there are more than one). This constraint allows for significant latitude in the age of the universe, in how long the physical laws will persist, in how consciousness is distributed across human beings, and so on.

In claiming that epistemically inhospitable universe plans predominate among the live plans, I do not take myself to be making a claim that can be justified in mathematical terms. There are, it would seem, an infinite number of live plans that are epistemically hospitable for you, as well as an infinite number that are epistemically inhospitable. And I see no reason for thinking that the inhospitable live plans form a set with a greater cardinality than the set of hospitable ones. Nonetheless, I think we have good reason to affirm the predominance of inhospitable plans even if we do not have a mathematical way of characterizing the relevant notion of predominance. Consider the following example.²¹ Suppose human explorers of distant solar systems have returned from their mission and they are about to escort you somewhere to see an object they have brought back from an alien civilization. You know nothing about the object or the civilization. Before being taken to see the object, you ask the explorers whether it is true that the object either (i) has a surface that is entirely red and aptly described as “furry” or (ii) has a surface that is entirely blue, green, or yellow and not aptly described as “furry.” The answer, you learn, is yes—the object does satisfy one of these two conditions. Having learned this, what should you think is more probable, that the object satisfies condition (i) or that it satisfies condition (ii)? It is plausible that on some natural and suitably fine-grained ways of partitioning the space of possible objects, there will be an infinite number of possible objects that would satisfy condition (i) and an infinite number that would satisfy condition (ii). (This could be true, for example, if there is some physical property of an object whose possible measurements form a continuum.) Moreover, there is no reason to think that, on any such partition, the objects satisfying condition (ii) will form a set with a greater cardinality than that of the set of objects satisfying condition (i).²² Even so, it seems fairly clear that, without other relevant evidence, you should think it more likely that the object satisfies condition (ii). And arguably, you should think this because, in light of the greater variety of ways for an object to satisfy (ii) than (i), you should think that objects satisfying (ii) predominate among the available possibilities. If that is right, then it seems that we should accept the similar “greater variety to predominance” reasoning in support of the conclusion that epistemically inhospitable universe plans predominate among the live plans.

I’ve just argued that among the live universe plans, plans that are epistemically inhospitable for you predominate. Does this conclusion, along with the supposition that a consequentialist God exists, straightforwardly entail that among the created universes containing a recent internal duplicate of yours, inhospitable universes predominate? No, for at least two reasons. The first reason stems from the possibility that a single universe plan could be actualized multiple times (resulting in multiple universes that are qualitative duplicates).²³ If God does create duplicate universes by actualizing some universe plans multiple times, there would seem to be no basis for specifying how many times God would actualize any given live plan. It is natural to postulate that God would actualize every live plan an infinite number of times, but this does not specify the extent of duplication since there are different “sizes” of infinity and there is no highest transfinite cardinal (Monton 2010; Johnson 2014; Rubio 2020). Without any basis for specifying how many times a given plan might be actualized, we cannot rule out the possibility that while God has actualized “more” inhospitable plans than hospitable ones, God has nonetheless created more hospitable universes than inhospitable ones.

Second, some philosophers have offered “Cantorian” arguments that purportedly show that God could not possibly actualize every universe plan that God has justification to actualize. Here is one such argument from Hawthorne and Uzquiano (2011) and Rubio (2020). Suppose that bosons are a particle type that can co-locate and that a good universe could include k-many bosons for any (finite or transfinite) cardinal number k. In this case, for any cardinal k, there is some good universe plan which specifies that there will be k-many bosons in a universe that corresponds to the plan. Suppose, for reductio, that God actualizes all of these universe plans, so that for every cardinal k, there is a created universe with k-many bosons. There is (it is claimed) a set B of all the bosons that exist in the multiverse. Let N be the cardinality of B. Since, by Cantor’s Theorem, N < 2^N (where 2^N is the cardinality of the powerset of B), we know that there is no created universe where the number of bosons is as great as 2^N. But this contradicts our supposition. To avoid such a contradiction, the Cantorian argument says that we should deny that God can actualize all the good universe plans.

I think that the theist should be wary of this argument, especially since similar Cantorian considerations pose problems for God’s omniscience (Grim 1988: 356). For sake of discussion, however, I assume that the Cantorian argument is correct. This means that even if God wanted to actualize all of the live plans, God could not do so. For any set of live universe plans, God has the ability to actualize all plans in the set. But the class of live universe plans is “too big” to form a set. On the assumption that the created universes must form a set, it follows that God must be selective in choosing which of the live plans to actualize. And without any insight into how God would select which of the good plans to actualize, we cannot rule out the possibility that God has more intensively sampled the space of hospitable live plans than the space of inhospitable live plans.

In light of the above considerations, I am happy to grant that facts about which types of universe predominate in a theistic multiverse might not correspond to the facts about which types of universe plan predominate among the collection of plans that God has justification to actualize. Nonetheless, if the theist does not have any clue as to how God would go about choosing which of the overall good universe plans to actualize (or how many times to actualize a given plan), then it seems that their expectations concerning the prevalence of the different types of universe should reflect their beliefs about the prevalence of the corresponding plan types within the collection of plans that God may justifiably actualize. Consider again the alien object example: absent any clues about the alien civilization and the kinds of objects it tends to produce, your judgment about whether the object is more likely to be red and furry (rather than not furry and either blue, green, or yellow) should reflect your beliefs about which type of object predominates among the possible objects. Similarly, if the theist should think that epistemically inhospitable plans predominate among the live plans, then the theist has a (defeasible) reason to think that epistemically inhospitable universes predominate among the created universes that include a recent internal duplicate.

In response, one might contend that there is a straightforward reason to think that God would more intensively sample the space of hospitable live plans than the space of inhospitable live plans: God would do this because a multiverse where epistemically hospitable universes are more abundant than epistemically inhospitable universes is, all else being equal, better than a multiverse where inhospitable universes are more abundant. But on reflection, this response is not persuasive. It would seem that for any possible multiverse M where hospitable universes are more abundant, there is a multiverse that surpasses M in value where inhospitable universes are more abundant. For example, God might combine the set of universes in M with an even larger set of universes that are all valuable but also epistemically inhospitable. No reason has yet been given for thinking that this expanded multiverse would not be better than M. If every multiverse where inhospitable universes are scarce is surpassed in value by some multiverse where inhospitable universes are predominant, there is no reason to think that a consequentialist God, concerned only with total value, would favor multiverses balanced towards epistemic hospitableness.

Lacking any idea of how God chooses which of the live universe plans to actualize, the theist is left with this worrying fact: among the live plans (any of which God may justifiably actualize), plans that are epistemically inhospitable for the theist predominate (in an intuitive but epistemically significant sense of “predominance”). This fact gives the theist prima facie reason to think that, on the supposition that God is a consequentialist, epistemically inhospitable universes predominate among the created universes that contain a recent internal duplicate. And if the theist has a prima facie reason to think this, she thereby has a prima facie reason to think that, supposing God is a consequentialist, she is probably in an inhospitable universe and is probably radically mistaken in her beliefs about the character of her universe. For even if some internal duplicates in a given inhospitable universe are not deceived, most are. And there can be no internally discernible evidence that could provide one with reason to think that they are a member of the fortunate non-deceived minority.²⁴

Do theists avoid this skeptical worry if they suppose that God abides by a deontic requirement that rules out gratuitous harms? Not clearly. First, whether the mistaken outlooks in inhospitable universes invariably qualify as harms in the relevant sense is doubtful. When your internal duplicate in an inhospitable universe suffers from a radically mistaken outlook, they are in some significant way lacking in knowledge and understanding, and they also have a great many false beliefs. But do they thereby suffer harm, harm of the sort that a good God would plausibly have an obligation to prevent? An affirmative answer strikes me as being questionable at the very least. When I consider the possibility that I am in a world that is only five years old, or that will disintegrate tomorrow, I do not naturally describe these as worlds where I am thereby harmed. Such worlds do seem to be ones where I am missing out on something quite important, but it is not especially plausible to think that God is under some requirement to guarantee that no beings anywhere miss out on important goods.

Second, even if radically mistaken outlooks qualify as harms, deception may play a role in securing goods that are worth exhibiting in a multiverse. In the previous section, I suggested that God may want the created order to include kinds of creatures and communities that are shaped by living in a universe with pointless evil. While a universe with such creatures might not be worth creating in a single-universe scenario (which is what makes these evils “pointless,” as I have defined the term), such a universe might be worthy of creation in a multiverse scenario. But there are, in fact, two ways that God might see to it that the universe contains creatures with the relevant sort of narratives and (hopefully) the relevant sort of virtues that are forged through grappling with suffering and evil. First, God could create a universe that contains the relevant forms of suffering. Second, God could create a universe that merely appears to its inhabitants to contain such suffering. In such a universe, creatures would lack the kind of serious responsibility for one another that we take ourselves to have. But the appearance of such responsibility is arguably all that is needed to secure a context that allows creatures to exhibit sacrificial love and that facilitates “soul-making” of the sort famously described by Hick (1966). If the likely degree of suffering in some hypothetical universe would constitute a bad that is worse than the bad of creatures being misled into thinking that they inhabit a universe with such intense suffering, then God might have a reason to pursue a systematic plan of deception. For example, God might miraculously turn off the conscious experience of creatures during intervals when they would otherwise be undergoing pointless suffering. God could then supply false memories of the sort of conscious experiences they would have had absent such divine intervention.²⁵ While the creatures in such a universe would suffer from significant deception, this would allow God to promote certain forms of soul-making in a way that avoids the cost of intense suffering (T. DeRose 2020).²⁶

I suspect that many, like myself, will recoil at the suggestion that God would opt for this deceptive scheme. Surely, one might think, God would find a way to realize great goods in a manner that does not rely on deception. To push back, however, it should be emphasized that a God who creates a multiverse no doubt has created universes with great goods that are not predicated on deception. And God may recognize these universes as being far superior to any deceptive universe God might create. Nonetheless, if the deceptive universes make possible certain goods (particular forms of heroism and virtue, for example) that cannot be realized in a better way, then why would God not choose to include such universes in the created ensemble? If it is plausible that the suffering that appears to take place in this universe would be worse than the deceptive appearance of such suffering, then the theist has an additional reason (beyond the sheer abundance of inhospitable universes) to think that she might be in an epistemically inhospitable universe.

4. Responses to the Worry

The skeptical problem posed to the theist by a divinely created multiverse is in certain respects more threatening than the standard sort of Cartesian worry that is typically the focus of epistemological discussions of skepticism. The Cartesian skeptic identifies scenarios (realistic dreams, deceptive demons, and so forth) that are claimed to be compatible with one’s evidence and that entail radical deception. Our knowledge of the external world is then called into question on the grounds that our evidence does not allow us to rule out these skeptical scenarios. But the Cartesian skeptic typically does not give us any positive reason to think that it is probable that some such skeptical scenario obtains. Such scenarios are presented as mere possibilities, ones that reveal how competing theories concerning the external world are underdetermined by our evidence. In response to such a skeptic, one may therefore attempt to argue that we enjoy a presumptive epistemic self-trust that is defeated only in the face of positive reasons to think that some sort of skeptical scenario obtains.²⁷

Unfortunately, this sort of anti-skeptical maneuver does not help the theist to respond to the skeptical threat developed here. In the present case, it is argued that, on the supposition that theism is true, there is a positive reason to think that a skeptical scenario probably obtains. I have argued that the theist has reason to think that epistemically inhospitable universes predominate among the universes where they have a recent internal duplicate; and this in turn gives the theist reason to think that their empirical outlook is probably radically mistaken. It will not help to insist that the onus is on the skeptic to provide some reason for thinking that we find ourselves in a skeptical scenario. For the theist, at least, such reasons have already been provided.

A second reason why the present skeptical argument is more worrying than standard Cartesian skeptical arguments stems from the modesty of its conclusions. Cartesian skeptical arguments aim to establish an unconditional skeptical conclusion. Such arguments may purport to show, for example, that your ordinary perceptual beliefs are not justified. Following G. E. Moore, one might respond to such an argument by arguing that because the negation of the argument’s conclusion is significantly more certain than the conjunction of its premises, it is not reasonable to accept the argument (1959: 226). Quite simply, it is more plausible that one of the premises is false than that the conclusion is true. Such a Moorean response is not plausible in the present context, however. The skeptical challenge developed above does not aim to establish an unqualified skeptical conclusion. Rather, the conclusion is the conditional claim that, on the supposition that theism is true, it is probable that your empirical outlook is radically mistaken. Denying this conclusion requires you to affirm that on the supposition that theism is true, it is not probable that your empirical outlook is radically mistaken. This latter affirmation hardly seems to be among those obvious beliefs of common sense that one might reasonably think are invulnerable to defeat by philosophical argument. So the Moorean response described above is not applicable.²⁸

Looking beyond Cartesian skeptical arguments, a skeptical problem that is more analogous to the one developed here is the problem posed by the fact that mainstream cosmological models predict (at least given certain plausible assumptions) an eventual superabundance of Boltzmann Brains (Carroll 2020). According to these models, in the eons after the universe has succumbed to “heat death,” complex material entities will on rare occasions be formed through chance processes. Some of these may be isolated Boltzmann Brains; and some of these Boltzmann Brains may pass through a sequence of physical states that perfectly mirrors the sequence of physical states of your own brain over the last several seconds. Indeed, standard cosmological models predict that Boltzmann Brains will eventually form the large majority of the “brains” that are intrinsic physical duplicates of your brain over the past few seconds. If we assume that Boltzmann Brains which are intrinsic physical duplicates of your brain over some interval of time have conscious experience that is phenomenally equivalent to your experience during the relevant interval, then Boltzmann Brains would arguably be recent internal duplicates of yours who are radically deceived in their beliefs. In this case, it would turn out that our universe is epistemically inhospitable.

The possibility of Boltzmann Brain superabundance, like the possibility of a predominance of epistemically inhospitable universes, is not presented as a mere possibility that is (allegedly) compatible with your evidence. Rather, it is presented as a possibility that is made probable by a prominent cosmological outlook together with a common view in the philosophy of mind. And it does not seem reasonable for someone who affirms this combination of views to accept that most of their recent internal duplicates are (or will be) deceived Boltzmann Brains while remaining confident that they are a non-deceived “normal” observer.²⁹ Even if it is not plausible that skepticism is rationally required in light of the Boltzmann Brain worry, it is plausible that in light of this worry one might be required to revise one’s scientific or philosophical views in order to avoid an outlook that predicts that our universe is epistemically inhospitable. Similarly, even if skepticism is off the table, the skeptical challenge developed here may exert significant rational pressure on the theist. Arguably, the theist who wants to maintain epistemic self-trust is rationally required to either abandon theism or identify good reasons for thinking that God is unlikely to create a multiverse where epistemically inhospitable universes predominate among the universes that contain a recent internal duplicate.

How might one reasonably remain a theist while resisting the skeptical challenge raised by the prospect of a theistic multiverse? One option, naturally, would simply be to posit that God has non-consequentialist reasons to avoid creating “deceptive” universes, even when creating deceptive universes would enhance the total value in the multiverse. If we could be assured of some such deontic requirement, then the prospect of a divinely created multiverse would not raise serious skeptical worries.

Countenancing such a requirement, while not utterly implausible, strikes me as a rather unconvincing response. For starters, one might worry that God’s ethics are in fact consequentialist. Additionally, while I entertained the possibility that God might create some universes with a deliberate intent to deceive the creatures in that universe, creating an inhospitable universe would not, as a general matter, require such an intention. And it’s far from obvious that God is obliged to prevent the majority of creatures from having radically mistaken outlooks. Consider, again, a universe plan for a universe that begins in a state exactly like the state of our universe five years ago. Call this plan 5Y. It is initially plausible that multiverse M that lacks a 5Y universe has less value than M+5Y, the multiverse that results from adding a 5Y universe to M. Suppose that for this reason, God decides to create a multiverse that includes a 5Y universe rather than a multiverse that lacks such a universe. Would the creation of a 5Y universe violate some sort of deontic constraint that is not violated when God creates some nondeceptive “Big Bang” universe like the one we presumably inhabit? Consider an inhabitant of the 5Y universe named Jill who, in the afterlife, learns that the evidence in her universe of a distant past was misleading. Could Jill reasonably complain that God should have placed her in a Big Bang universe? Not clearly. In response to such a complaint, God might point out that Big Bang universes were also included in the multiverse and that Jill is not identical to any of her counterparts in those universes (no matter how similar Jill and these counterparts may be). And if God had refrained from creating the 5Y universe, this would not (it seems) have somehow made Jill identical to one of her counterparts in a Big Bang universe; rather, she would not have existed at all. Assuming Jill’s life is worthwhile, she arguably has reason to be grateful that God created a 5Y universe.³⁰ Of course, this does not show that there is no deontic constraint forbidding the creation of epistemically inhospitable universes. But by no means is it clear that there is some such constraint.³¹

A more convincing theistic response to the skeptical challenge, it seems to me, would explain why God might have consequentialist reasons to refrain from creating epistemically inhospitable universes. If creating such universes makes things worse on the whole, then we have good reason to think that God would not create them, whether or not God’s ethics are purely consequentialist. In this final section, I want to briefly consider broadly consequentialist responses to the skeptical worry developed above.

Thus far, I’ve assumed that God’s creating some universe that is intrinsically valuable adds to the total value of the multiverse or at least does not detract from the total value. But this assumption can be questioned. There may be cases where adding an intrinsically valuable universe to a universe ensemble undercuts the value contribution of one or more other universes, so that the addition of the intrinsically valuable universe worsens the multiverse overall.

A natural move here might be to appeal to the aesthetic merits of the multiverse.³² Perhaps the beauty of the entire ensemble of universes requires that it not be marred by the inclusion of universes that, while good overall, are ugly in some way or inelegant in their design. Consider the following analogy. Suppose there is a blank white space of 8.5 inches by 11 inches that can either remain blank or that can instead display some of my doodles recently made while listening to some philosophy talk. The space may have more intrinsic value if it is graced by my doodles than if it remains blank. But now suppose that the space happens to be part of a wall in the Louvre and that the space is immediately adjacent to the Mona Lisa.³³ In this case, it would be better for the space to remain blank. Similarly, it could be bad for God to create some mediocre universe that, while intrinsically quite valuable, undermines the aesthetic merits of the multiverse.

I don’t find this appeal to the aesthetic considerations particularly convincing. First, it is by no means obvious that inhospitable universes are especially ugly or that they would somehow compromise the aesthetic merits of the multiverse. Second, if there is some possible universe ensemble that would be beautiful when considered as an organic whole, then that collection would presumably retain its beauty even if it exists as a proper part of some larger ensemble that, considered as a whole, is not as beautiful. Even if universes were arranged spatially, the ugly universes could be separated from the beautiful ones, just as my doodles can remain in my folders and files and away from the Louvre. And if universes are not in some spatial arrangement, then there is even less reason to think that aesthetic merits of the whole would be important.

Setting aesthetic considerations aside, I think that there are more plausible reasons for thinking that adding an intrinsically valuable universe to the multiverse could detract from the overall value contained in that multiverse. The account I propose begins with the supposition that one of the most valuable things a universe could contain is creatures who are special to God. Arguably, a creature’s specialness to God would be undermined were there sufficiently many other creatures in the world (even if not in the same universe) who are sufficiently similar. Suppose that in parallel universes, I have an infinite number of counterparts whose lives perfectly mirror my own. In such a situation, God may still recognize my value and may still care for me. But could I really be special to God? It is hard to see how I could be, given that there is nothing of any significance that distinguishes me from these counterparts.³⁴

If similarity between two creatures can compromise their specialness to God, then we may be able to explain why God would not want to create epistemically inhospitable universes even when such universes have significant intrinsic value. While a deceived subject in an epistemically inhospitable universe and her internal twin in a hospitable universe differ significantly in their degree of knowledge, these individuals may be alike in respects that compromise their specialness. If God has reason to promote creaturely specialness, then God would want to avoid duplicating universes and would also want to avoid creating creatures who, while perhaps inhabiting very different sorts of universes, are internal duplicates. Given these assumptions, a consequentialist God would have reason to be selective in choosing which universes to create, in order to avoid specialness-compromising similarities. This means that God will normally want to choose between epistemically inhospitable universes and various hospitable universes that contain internally similar people. Since deception is a bad and knowledge a significant good, God presumably would want to choose hospitable universes over inhospitable ones in cases where the deception in the inhospitable universe is pointless.

One advantage of this “specialness” account is that it would explain why God might value diverse kinds of creatures and diverse kinds of universes even on the supposition that diversity is not valuable in itself. Because an overabundance of highly similar universes would eventually compromise specialness, the diversity of universe types is instrumental to the production of more creatures who are special to God. God might want to create universes with apparent suffering not because God cares about filling in every niche in the space of possibilities, but because God must make use of these regions in the space of possibilities to maximize the number of special creatures.

The specialness account gives a consequentialist explanation for why God would not create epistemically inhospitable universes when the deception involved is pointless, but what about cases where such deception is not pointless? If the deceptive appearance of suffering makes possible certain forms of soul-making while avoiding the cost of intense suffering, might not God in some cases choose an inhospitable universe over the hospitable alternative? While the “specialness account” does not block this possibility, perhaps we can rule it out by appealing to other ways that the value in one universe might be compromised by some other universe’s existence. For example, perhaps one of the greatest goods achievable in the created order is true friendship between God and creatures that are special to God. And perhaps the desirable kind of friendship between God and creature necessarily involves well-founded trust in God. Arguably, no creature’s trust in God would be well-founded if God has created epistemically inhospitable universes where the inhabitants are radically deceived. The mere existence of such deceived creatures would arguably undermine well-founded trust, even if individuals in hospitable worlds have no knowledge of such deception.

The position emphasizing specialness and true friendship, if tenable, may allow the theist to both retain the explanatory power of a multiverse theodicy while avoiding the skeptical concerns posed by a theistic multiverse. In light of the multiverse possibility, the theist is not required to hold that our universe is good enough to be creation-worthy in a single-universe scenario. But even if the standards of acceptability are much laxer in a multiverse context, the proposal that God aims to promote true friendship with creatures who are special in God’s eyes can explain why God would avoid creating epistemically inhospitable universes (even ones that have significant intrinsic value).

While I think the “friendship response” to the multiverse skeptical worry is promising and plausible, my primary aim has been to show that the prospect of a divinely-created multiverse raises serious skeptical concerns that deserve careful attention. Whether theism is rationally compatible with epistemic self-trust may crucially depend on what the theist has justification to believe about God’s ethics of creation.