What reasons are there for thinking that skeptical theists would endorse pe_i-skepticism? Here I will incorporate skeptical theist responses to Rowe’s evidential problem of evil to sketch one route by which the skeptical theist would arrive at an endorsement of pe_i-skepticism with the caveat that this is but one of several possible skeptical theist routes to arrive at such an endorsement.7 In order to accomplish this, I will begin by examining the skeptical underpinnings of skeptical theism and then explicate how they are typically brought to bear to undermine Rowe’s evidential problem of evil.

Hendricks (2020) has recently offered an argument for the truth of the skepticism which undergirds the skeptical theist position, which he calls “The Preclusion Argument” (PA):

I will assume, as Hendricks does, that skeptical theists should endorse PA. As to how PA undermines Rowe’s problem of evil, consider Rowe’s (1979) paradigm instance of purportedly gratuitous evil: a fawn being burned alive in a forest fire. Within the dialectic of Rowe’s argument, an evil is considered gratuitous if it is not connected to, and necessary for, some set of states of affairs that outweighs it. God—a morally perfect being—would, per the argument, only be morally justified in allowing an individual instance of evil, E_i, iff the total value of the set of states of affairs that are produced by the E_i is greater than the total value of the set of states of affairs that would have obtained had the E_i not occurred; that is, if the E_i is not gratuitous (Hendricks 2020: 265). If any E_i is gratuitous—as Rowe thinks the fawn’s being burned alive is—then God would not be morally justified in allowing its occurrence and therefore would not be morally perfect and therefore would not exist.

The difficulty for Rowe’s argument, however, is that in order to determine whether any E_i is gratuitous and thereby determine whether God would be morally justified in allowing any E_i’s occurrence, one would need to determine the total value of the set of states of affairs produced by the E_i in question. But per PA, we cannot do this: the total value of the set of states of affairs produced by any E_i is inscrutable to us. And if the total value of the set of states of affairs produced by any E_i is inscrutable, one cannot infer whether any E_i is gratuitous and therefore cannot infer the nonexistence of God from the occurrence of any E_i. I take this to be a standard skeptical theist understanding of how skeptical theism undermines Rowe’s evidential problem of evil.

As to how this conclusion supports an endorsement of pe_i-skepticism, consider the following (what will prove to be problematic) line of thought. Not knowing the set of states of affairs which obtains as a result of any E_i’s occurrence, one cannot determine whether any E_i is gratuitous and therefore cannot determine whether any E_i is such that God would be morally justified in allowing its occurrence. But if one cannot determine whether any E_i is such that God would be morally justified in allowing its occurrence, one cannot determine how likely it is that any E_i would occur given theism, as the likelihood of any E_i occurring given theism very much depends upon whether the E_i in question is such that God would be morally justified in allowing its occurrence. If it is the case—as is supposed within Rowe’s argument—that God would only be morally justified in allowing an E_i if the E_i in question is not gratuitous, then Pr(E_i|T) where E_i is a gratuitous evil is 0.9 But if, as follows from PA, it is ultimately inscrutable as to whether or not any evil E_i is gratuitous, then Pr(E_i|T) for every individual instance of evil E_i is correspondingly inscrutable: we do not know how likely the evil in question is given theism because we do not know whether the evil in question is gratuitous or not.

One might resist this latter inference by arguing that while it may be inscrutable as to whether any E_i is gratuitous, the probability that any E_i is gratuitous—and, correspondingly, Pr(E_i|T) for the evil in question—need not be. Whatever one might think of such a line of argument, it does not seem to be one the skeptical theist is liable to endorse.10 For, following Oliveira (2020: 327), if a) one can assign a likelihood to evils being gratuitous, and b) there are evils which seem more-likely-than-not gratuitous, then the size of a large collection of likely-gratuitous evils itself provides sufficient evidence for thinking that the collection contains at least one gratuitous evil.11 Assuming, as seems plausible, that given (a) and (b), the size of the collection of likely-gratuitous evils would be very large, then the consequence of conceding that the probability of any E_i being gratuitous is not inscrutable is that skeptical theism could be true and yet not undermine the very problem it has developed in response to—Rowe’s evidential problem of evil. However, seeing as skeptical theists do think, contra Oliveira, skeptical theism undermines Rowe’s evidential problem of evil, I take it that skeptical theists would endorse the inference from the inscrutability of any E_i being gratuitous to the conclusion that Pr(E_i|T) for any E_i is inscrutable; that is, pe_i-skepticism.

Before proceeding, it is necessary to guard against misunderstanding by making explicit what pe_i-skepticism does and does not commit one to. pe_i-skepticism amounts to the claim that for any arbitrarily chosen instance of evil, the likelihood of that instance of evil occurring given theism is inscrutable in virtue of one’s lack of knowledge over whether God has a morally justifying reason for allowing the evil’s occurrence. pe_i-skepticism does not commit one to taking a stance on whether every instance of evil is independent conditional on theism, nor does it commit one to taking a stance on whether God would be required to have independent justification for his allowance of each individual instance of evil. Therefore, if Seachris and Zagzebski (2007) are correct12 in thinking that if God had a morally justifying reason for allowing some specific instance of evil, that justification could also apply to relevantly similar instances of evil, that would have no bearing on pe_i-skepticism. pe_i-skepticism follows from a skepticism as to one’s knowledge of whether God would be justified in allowing any specific instance of evil; regardless of whether that justification is for God’s allowance of one individual instance of evil or for God’s allowance of all relevantly similar instances of evil.

While it seems skeptical theists would endorse pe_i-skepticism, clearly they would want to avoid an endorsement of pe-skepticism given the agnostic consequences which, per Climenhaga, follow from such an endorsement. The problem, however, is that if skeptical theists endorse pe_i-skepticism, they are implicitly committed to endorsing pe-skepticism. If, for any individual evil within the total collection of observable evils, the probability of that evil occurring given theism is inscrutable, the probability of the total collection of observable evils occurring given theism cannot but be inscrutable.

To begin, consider the collection of observable evils E which consists of the specific individual evils, E₁ and E₂ and … E_n. It seems that there is an intimate relationship between the probability of, and the knowability of the probability of, a collection of evils, and the individual evils within that collection, such that the probability of individual contents within a collection impacts the probability of the totality of the collection, given theism. For example, consider a possible world with copious amounts of evil and in which all of the evils within it exclusively are brutally violent acts against children. Intuitively, were such a world actual, one would say Pr(E|T) is very low. But one would seemingly make this judgment of Pr(E|T) being low because Pr(E_i|T) where the individual instance of evil E_i is a brutally violent act against a child is intuitively low. In other words, knowledge of the likelihood of individual cases of evil occurring given theism informs our knowledge of the likelihood of collections of such cases occurring given theism: we can intuitively know that a world which contained copious amounts of evil and in which all of the evils within it exclusively are brutally violent acts against children is unlikely given theism because brutally violent acts against children, considered on an individual level, do not at all seem to be the sort of thing an omni-God would seek to include in their world actualization given how bad these evils are. The likelihood of these evils occurring given theism is a function of their, for lack of a better word, badness: the worse an evil seems, the more unlikely its occurrence seems given theism. This is the basic intuition which gives force to Rowe’s evidential problem of evil. Hence why Rowe famously focuses on a fawn being burned alive in a forest fire instead of a toe-stubbing. And hence why skeptical theists deny a judgment as to the net badness of this evil.

Therefore, it seems to be the case that however one assigns a value to Pr(E|T), one will need to do so in such a way as to account for the specific contents which make up the collection, E. Assuming, as seems plausible, that any individual evil within the collection of observable evils is not independent conditional on theism,13 one plausible way of assigning a value for Pr(E|T) is as follows:

Quite clearly, if this understanding of Pr(E|T) is correct, then if the skeptical theist endorses pe_i-skepticism, they are correspondingly committed to the probability of any collection of evils given theism being inscrutable. In which case, an endorsement of pe_i-skepticism commits one to an endorsement of pe-skepticism. Therefore, per Climenhaga, those who endorse pe_i-skepticism should not be theists but agnostics.

Whether or not this particular construal of Pr(E|T) is correct does not seem to matter so much as what this construal correctly captures; namely, that in order to assign a value to Pr(E|T), one seemingly needs to do so in such a way as to account for at least some of the specific contents which make up the collection, E.

In order for the skeptical theist to avoid this agnostic consequence, the intimate relationship between the probability of a collection and the probability of the individual contents within the collection needs to be severed, such that one can make a probabilistic judgment as to the likelihood of a collection of evils occurring given theism, despite their not having a clue as to the likelihood of any individual instances of evil within the collection occurring given theism. To continue the above example, one could know that in a world in which all of the evils within it exclusively were brutally violent acts against children, Pr(E|T) is low given how bad the collection of evils is. This, despite one’s total ignorance—given the inscrutability of Pr(E_i|T) for each of the instances of evil within the collection—as to whether any instance of evil within the collection is the sort of thing an omni-God would seek to include in their world actualization. Quite simply, if such an understanding of Pr(E|T) exists, it remains to be seen what it is.

In the absence of such an understanding of Pr(E|T), in endorsing pe_i-skepticism, skeptical theists are rationally committed to endorsing pe-skepticism and therefore, per Climenhaga, should not be theists, but agnostics. To address a likely objection, skeptical theists cannot avoid this conclusion by appeal to the Paradox of the Preface (PP) (see Makinson 1965). PP is the paradox one encounters in reading a preface of a book in which, after acknowledging how the work benefitted from the insight and feedback of others, the author says something along the lines of, “all remaining mistakes are solely attributable to the author”; this despite the fact that the author, we can presume, believes every claim they made within the book is correct. In other words, the author simultaneously believes that every individual claim they made within the book is correct and also that the collection of claims within the book contains at least one error.

The objection then, would go something like this: PP seems to be a concrete example whereby one is rationally able to make a probabilistic prediction about a whole collection—namely, the collection of claims within the book either definitely or very likely does contain at least one error—which seems to be independent of any probabilistic predictions of individual instances within the collection being errors. So, despite the author’s belief that each individual claim within the book is correct, the author can rationally make a probabilistic prediction about the collection of claims which constitute the book containing an error because they have independent (of the contents of the collection) reason to think the collection contains an error: they know from past corrections of their work that the likelihood their book contains an unforeseen error is, if not certain, extremely high. Given this, why cannot the skeptical theist likewise consistently make a probabilistic prediction about a collection—Pr(E|T)—which is independent of any probabilistic predictions of individual instances within the collection?14

For the sake of argument, let us grant that within PP the probabilistic prediction about the collection containing an error is not a function of the probabilistic predictions of individual instances within the collection being accurate since, absent such a concession, the objection is easy to dismiss.15 Even granting such a concession, there is a salient disanalogy between PP and the position skeptical theists find themselves in; at least as things stand. Within PP, what allows the author to consistently make a probabilistic prediction about a collection which—we are presuming—is independent of any probabilistic predictions about individual instances within the collection is the author having independent (of the contents of the collection) reason to think the collection contains an error: they know from past corrections of their work that the likelihood of error in their new book is extremely high.

Therefore, in order for the analogy with PP to hold, skeptical theists would need to provide independent (of the likelihood of individual instances of evil occurring given theism) reason for thinking Pr(E|T) is whatever value they assign to it. In other words, skeptical theists would need to engage in the task of theodicy: explaining what reasons God may or may not have for creating a world with the potential to have a collection of evils like ours does such that Pr(E|T) can be given a knowable value assignment. If the skeptical theist were to supplement their skeptical claims with theodical claims, they conceivably could consistently maintain Pr(E|T) being knowable, despite Pr(E_i|T) for every individual instance of evil being inscrutable. This is because their theodical claims could provide reason to think Pr(E|T) is whatever value they assign to it which is independent of any probabilistic predictions of individual instances within the collection, E.16 However, the difficulty with this approach is that if skeptical theists’s skepticism also applies to their knowledge of how God would weigh the various possible reasons for permitting the collection of evils, E—as Hendricks (2023a) notably argues it should—then it would seem that even if skeptical theists could consistently employ theodical claims alongside their skeptical claims, they would be significantly restricted in both the theodical claims they could supplement their skeptical theism with, as well as the justification they could provide for them (Hendricks 2023b: ch. 8). In which case, even if such claims were rationally consistent, it is fair to wonder as to their plausibility. And given that what is in question is how skeptical theism could be employed to undermine Bayesian arguments from evil, the plausibility of the theodical claims employed by the skeptical theist to supplement their skeptical claims will be relevant.

Absent of pursuing theodicy, then, it would seem that in order to avoid agnosticism, skeptical theists must reject pe-skepticism and thereby must reject pe_i-skepticism as well. One way to reject pe_i-skepticism is by drawing a distinction between the likelihood of a specific evil occurring given theism and the likelihood that the instance of evil in question results in a set of states of affairs which are axiologically greater than the set of state of affairs that would have obtained had the evil in question not occurred. More concisely, the likelihood of a specific evil occurring given theism should be distinguished from the likelihood that that specific evil in question is gratuitous. While certain evidential problems of evil may lead to a conflation of these likelihoods, they clearly are not necessarily synonymous. One need not know the set of states of affairs that obtains as a result of some heinous evil occurring to think that—at least absent reason to think the contrary—the likelihood of that heinous evil occurring given theism is low. Or, in some instances, that the likelihood of some instance of heinous evil occurring given theism is remarkably high. After all, according to one theistic tradition, Pr(E_i|T) is not inscrutable, but 1, where E_i is the individual evil that was the crucifixion of a certain religious teacher at the hands of an angry mob.

Of course, once it is acknowledged that there are at least some individual instances of evils E_i which have a knowable probabilistic likelihood given theism, this paves the way for the success of Bayesian arguments from individual instances of evil to the nonexistence of God; arguments which skeptical theism cannot undermine. And by “cannot undermine” I simply mean that skeptical theism cannot be employed to prevent one from determining whether an individual evil, E_i, is evidence for or against the existence of God by preventing one from assigning a value to Pr(E_i|T). As to how problematic such a possibility is for skeptical theism, however, the jury is still out. For while skeptical theism may not be able to undermine such arguments from evil, skeptical theism may perhaps yet be employed to render such arguments less problematic for the theist than they otherwise would be.17 Whether skeptical theism can ultimately do so will depend upon the strength of the specific Bayesian argument from evil in question. My purpose here is not to offer such an argument. But it is to note that the prospects for Bayesian arguments from evil to the nonexistence of God—be they arguments from the total collection of observable evils, arguments from some collection of observable evils, or arguments from particular instances of evil—look much more worrisome for skeptical theists than previously thought; at least if skeptical theists wish to remain theists.