I

Our nullifidian protagonist undertakes to argue as follows:

1. (1): An event has occurred which instantiates F. [Premise]

2. (2): An event has occurred which instantiates G. [1; X]

3. (3): If an event has occurred which instantiates G, then theism is false. [Premise]

4. (4): Theism is false. [2, 3; MP]

That is, she assertively utters or inscribes sentences identical to (1)–(4) except that, in place of ‘F’ and ‘G’, they contain names of properties, and in place of ‘X’ they contain the name of an inference rule. If the named inference rule is not a deductive inference rule, let us say that our protagonist has made an evidential atheological argument. If, moreover, the property named in (1) is, or is sufficiently closely related to, the property of being a prima facie evil, let us say that she has made an evidential atheological argument from evil.Footnote ¹

To refer to our nullifidian as a protagonist is to locate her within a broader narrative structure – one which, if it is to be of interest, must contain an antagonist. Let us introduce, then, a strategy for resisting her argument – a strategy which has come to be known as sceptical theism. The substance of our drama will be the dispute between the sceptical theist and our protagonist, or rather between the sceptical theist and myself, writing on our protagonist's behalf. At its climax, I will argue that the consistent sceptical theist is radically insensitive to certain kinds of evidence about prima facie evils – that she is committed to the claim that not even evidence of extreme and pervasive suffering could justify disbelief in theism – and that this is a serious problem for sceptical theism. But first, we must more fully explore the natures of both the evidential argument and the sceptical theist's response to it.

Let us call the first numbered claim in our protagonist's argument its evidential premise, the second its intermediate conclusion, and the rule of inference for the name of which ‘X’ is a placeholder its epistemic norm. An argument of this form works as follows. Because the conjunction of the intermediate conclusion and (3) entails the falsity of theism, the strength of the evidence for the latter must equal or exceed the strength of the evidence for the former. But (3) is chosen so as to be uncontroversial, so the strength of the evidence for the conjunction of the intermediate conclusion and (3) is approximately equal to the strength of the evidence for the intermediate conclusion. As a consequence, the falsity of theism, being relevantly like an implication of the intermediate conclusion alone, inherits the evidential support conferred on it by the evidential premise in the presence of the epistemic norm.

This cannot be the complete story, of course, for there is no such thing as the evidential support conferred on our protagonist's intermediate conclusion by her evidential premise simpliciter. The evidential significance of a proposition depends on the body of evidence to which it is added. It is possible, for example, for a given proposition to confirm a hypothesis relative to one body of evidence and disconfirm it relative to another. What, then, is the body of evidence relative to which (1) is claimed to support (2)? Here, I think, our protagonist may be excused for being inexplicit. It is a body of evidence, she can reasonably say, which includes much of what we take to be our actual evidence. But it excludes both theism and its negation, as well as propositions about the magnitude and distribution of prima facie evils. This inexplicitness is a typical feature of evidential reasoning and should not be thought to vitiate the evidential argument.

In what follows, I will explicate the informal notion of evidential support in terms of evidential probability. I will thus take for granted the existence of an objective evidential probability function P which satisfies the probability axioms. Following Williamson (Reference Williamson2000, 211) I will think of P informally as a measure of ‘something like the intrinsic plausibility of hypotheses prior to investigation’. The conditional evidential probabilities assigned to various propositions relative to bodies of evidence (representing possible epistemic situations) will then represent the total degrees of evidential support those propositions enjoy relative to those epistemic situations. I will use ‘P(h|b)’ to denote the evidential probability of hypothesis h relative to body of evidence b.

A wide variety of evidential arguments can be generated by considering particular substitutions for ‘F’, ‘G’, and ‘X’. The most influential of these has been one formulated by William Rowe (Reference Rowe1988; Reference Rowe1991; Reference Rowe and Howard-Snyder1996). Considering events E ₁ and E ₂ (E ₁ is the immolation of a fawn; E ₂ is the rape and murder of a five-year-old girl), Rowe argues that:

1. (1′): No good we know of justifies an omnipotent, omniscient, perfectly good being in permitting E ₁ or E ₂.

2. (2′): No good at all justifies an omnipotent, omniscient, perfectly good being in permitting E ₁ or E ₂.

3. (3′): If no good at all justifies an omnipotent, omniscient, perfectly good being in permitting E ₁ or E ₂, then there is no omnipotent, omniscient, perfectly good being.

4. (4′): There is no omnipotent, omniscient, perfectly good being.

It is straightforward to interpret Rowe's argument along the lines of (1)–(4). Doing so yields:

1. (1″): An event has occurred which instantiates the property that no good we know of justifies an omnipotent, omniscient, perfectly good being in permitting it. [Premise]

2. (2″): An event has occurred which instantiates the property that no good at all justifies an omnipotent, omniscient, perfectly good being in permitting it. [1; IR]

3. (3″): If an event has occurred which instantiates the property that no good at all justifies an omnipotent, omniscient, perfectly good being in permitting it, then theism is false. [Premise]

4. (4″): Theism is false. [2″ and 3″; MP]

The event in question in (1″) is the disjunctive event formed from E ₁ and E ₂ the event which occurs just in case either E ₁ occurs or E ₂ occurs. Rowe clarifies that the epistemic norm on the basis of which (1″) confers evidential support on (2″) is inductive reasoning: ‘the inference from [(1″)] to [(2″)] is like an inference from “All the A’s we've observed are B’s’’ to “All A’s are B’s” ’ (Rowe (Reference Rowe and Howard-Snyder1996), 267).

II

We are now in a position to assess the force of our protagonist's argument. The crucial question is, of course, whether the transition from its evidential premise to its intermediate conclusion is licit. Theists have not, in general, been terribly impressed with this transition. They have developed a number of strategies for resisting it, the most plausible of which – sceptical theism – has already been introduced. The sceptical theist alleges that very general considerations concerning the human epistemic position vis-à-vis facts about modality and value show that such transitions are, without exception, illicit.

In what has come to be a standard account, Michael Bergmann (Reference Bergmann2001; Reference Bergmann, Rea and Flint2009; Reference Bergmann, Clark and Rea2012) characterizes sceptical theism as traditional theism conjoined with (at least) the following four theses:

1. (ST1): We have no good reason for thinking that the possible goods we know of are representative of the possible goods there are.

2. (ST2): We have no good reason for thinking that the possible evils we know of are representative of the possible evils there are.

3. (ST3): We have no good reason for thinking that the entailment relations we know of between possible goods and the permission of possible evils are representative of the entailment relations there are between possible goods and the permission of possible evils.

4. (ST4): We have no good reason for thinking that the total moral value or disvalue we perceive in certain complex states of affairs accurately reflects the total moral value or disvalue they really have.

Applied to Rowe's argument above, (ST1-4) function by undermining the evidential connection between evidential premise (1″) and intermediate conclusion (2″); if (ST1-4) are true, then (1″) arguably constitutes negligible evidence for (2″) and therefore for (4″).

Bergmann's (ST1-4) are tailor-made to target evidential arguments from evil which, like Rowe's, make explicit reference to possible goods and evils. It would therefore be a mistake to think that sceptical theism as a response to evidential arguments is limited to Bergmann's claims. More generally, I propose that the following thesis captures what is distinctive about sceptical theism:

• (Sceptical Thesis): What we can come to know or justifiably believe (using our value-directed cognitive faculties alone) about the properties of events relevant to the question of whether God, if he exists, would be justified in permitting those events to occur constitutes only weak evidence about the overall justification-conferring status of said properties.

This is not to say that sceptical theists are not additionally committed to (ST1-4); indeed, I will assume in what follows that the sceptical theist endorses both (Sceptical Thesis) and (ST1-4), with her commitment to the latter based on the more general scepticism contained in the former.

III

Despite its purported obviousness, sceptical theism has been the target of extensive criticism. It has been accused of vitiating standard theistic arguments (e.g. by Beaudoin (Reference Beaudoin1998) ) and making it impossible for believers to love or trust God (e.g. by Gale (Reference Gale and Howard-Snyder1996) and Wielenberg (Reference Wielenberg2010; Reference Wielenberg, Dougherty and McBrayer2014) ). Many have also argued that it has unacceptable epistemic and practical consequences: forcing us to suspend judgement concerning the actuality of sceptical scenarios (see e.g. Russell (Reference Russell and Howard-Snyder1996) and Wilks (Reference Wilks2009) ) and interfering with moral deliberation (see e.g. Almeida & Oppy (Reference Almeida and Oppy2003) and Maitzen (Reference Maitzen, Dougherty and McBrayer2014) ).Footnote ² Some of these criticisms, I believe, do constitute difficulties for the sceptical theist. They will not, however, be the focus of my discussion. Instead, I will be developing an argument against sceptical theism which I believe to be both decisive and underappreciated. The spirit of this argument has been succinctly expressed by Rowe (Reference Rowe2001):Footnote ³

Let us refer to the scenario Rowe imagines, in which human life is nothing more than a series of agonizing moments from birth to death (and, we may suppose, an infinite number of human lives are lived over the course of history), as The Worst Scenario. Rowe's claim is that the consistent sceptical theist must concede that even if she found herself in The Worst Scenario, her evidence would not justify disbelief in theism. To my knowledge, the only attempt to engage substantively with Rowe's argument from the perspective of the sceptical theist is due to Bergmann (Reference Bergmann, Rea and Flint2009), who writes:

Bergmann takes Rowe's argument seriously, and rightly so: its conclusion would serve as a reductio of sceptical theism. In the remainder of this article, I will argue on the nullifidian's behalf that the threat of this kind of reductio is more serious than Bergmann and other sceptical theists have imagined.

Before considering Rowe's argument and Bergmann's response in more detail, however, a common misunderstanding of the sceptical theist's commitments must be addressed. It is often claimed that insensitivity to disconfirmation in the light of evidence about value is an essential feature of sceptical theism and thus cannot constitute an objection to it. This is, however, a distortion of the situation, which is that the sceptical theist is committed to the claim that no evidence about value of which anyone is currently in possession justifies disbelief in theism.Footnote ⁴

Here is a more precise way to characterize matters as they stand. Let ‘e’ be a variable ranging over the set of events which are prima facie evils. For any e, let us call the proposition that e occurs ‘${\rm {\cal P}}\left( e \right)$’. For any possible world w and time t, let us define ${\rm {\frak D}}_{\langle w,t\rangle} $ (pronounced ‘the datum at $\langle w,t\rangle $’) as the proposition which is the conjunction of ${\rm {\cal P}}\left( e \right)$ for all e in the history of w up to and including t. Then, if ‘@’ is a constant which rigidly designates the actual world, any acceptable form of sceptical theism must render plausible:

• (The Criterion): For all times t up to and including the present, ${\rm {\frak D}}_{\langle @,t\rangle} $ is such that, relative to a neutral evidential base, it does not justify disbelief in theism.

(The Criterion) is a strong claim – perhaps even a shocking one – but it is considerably weaker than the claim that not even the datum of The Worst Scenario after countless generations could justify disbelief in theism. As Bergmann's rejection of it shows, this latter claim is not an obvious consequence of sceptical theism. In what follows, I will attempt to show that a closely related claim is an unobvious consequence – that sceptical theists must be radically insensitive to certain kinds of evidence about prima facie evils. If my argument is sound, it constitutes a decisive objection to sceptical theism.

In the remainder of this article, I will argue (i) that Bergmann's response to Rowe's argument is inadequate and (ii) that no adequate sceptical-theistic response to the more general class of insensitivity arguments is forthcoming. Along the way, I will consider and reject a number of moves that might be thought to offer the sceptical theist a way to avoid my conclusions.

IV

If Bergmann's response to Rowe's argument succeeds and can be generalized so as to apply to other insensitivity arguments, the sceptical theist emerges from her exchange with Rowe unscathed. Our first step, then, must be to show that Bergmann's response does not succeed. I will offer two arguments to this effect.

First, according to Bergmann (Reference Bergmann, Rea and Flint2009, 382), we are not justified in endorsing the principle that

• (B1): If God exists and is unsurpassably loving, then for any human subject H and time t, if H is at t capable of relating personally to God, H at t believes that God exists, unless H is culpably in a contrary position at t.

We are not justified in endorsing (B1), Bergmann argues, because (B1) is false if there is a God-justifying reason for permitting the falsehood of (B1), and we are not in a position to determine whether such a reason exits. This is not an idiosyncratic aspect of Bergmann's view – if the sceptical theist's scepticism does not extend to (B1), then theism is undermined by the problem of divine hiddenness. But clearly, if our epistemic position with respect to Swinburne's principle is relevantly like our epistemic position with respect to (B1), the consistent sceptical theist cannot maintain that we are justified in endorsing one but not the other. So the sceptical theist is faced with the problem of showing that our epistemic position with respect to Swinburne's principle is more secure (let us call this the Justification Problem).

Why couldn't God be justified in permitting suffering even if the sufferer's life is not on the whole a good one? Perhaps the sceptical theist will claim that Swinburne's principle is more intuitively compelling than (B1). If she does, it is worth pointing out that the most intuitively compelling principles in this vicinity concern particular prima facie evils – for example, the principle that God would not permit the rape and murder of a five-year-old girl (Rowe's E ₂). The sceptical theist must deny that we are in a position to accept such principles; Bergmann (Reference Bergmann, Clark and Rea2012, 24), for example, advances an error theory according to which we judge such principles plausible because we consider the evils they report without at the same time grasping how they could be modally connected to the prevention of identifiable worse evils. Though Bergmann makes much of the vividness with which particular horrific events are presented in the principles in question, the same sorts of considerations apply straightforwardly to more general principles like (B1), as well. It is difficult to see how the sceptical theist can consistently fall back on the possibility of extenuating circumstances when it comes to all particular prima facie evils and certain general principles (like (B1) ), but then, without further explanation, deny that this is a genuine possibility when it comes to Swinburne's principle. By (ST1-4), after all, the sceptical theist must take seriously the possibility that a violation of Swinburne's principle could be excused because it either instantiates goods of which she is unaware or is a necessary concomitant of a state of affairs which instantiates such goods. To refuse to countenance this possibility on the basis that Swinburne's principle is knowable a priori without according equally plausible theses like (B1) the same epistemic status would be special pleading.

Suppose, however, that the sceptical theist is willing to bite the bullet vis-à-vis Swinburne's principle, holding that she knows it a priori, that she fails to know similar principles like (B1) which threaten her theistic commitments, and that she has no general story about why the two kinds of cases come apart. There remains a further problem for her in determining whether any particular life is on the whole a good one (let us call it the Recognition Problem). There are many cases which are difficult to reconcile with Swinburne's principle on the assumption of theism: children who are kidnapped and forced to work as sex workers, infants born with certain genetic disorders, such as Tay-Sachs disease, that cause them extreme pain for the duration of their lives, and others. To accommodate these cases, the sceptical theist who accepts Swinburne's principle must provide some reason for doubting the intuitive judgement that the lives in question are not on the whole good ones. There is, it seems to me, only one way for her to do so. In typical sceptical-theistic fashion, she must argue that we are not in a position to discover, by the exercise of our value-directed cognitive faculties alone, whether such lives are on the whole good ones. Perhaps, after all, they instantiate good-making properties of which we are unaware.Footnote ⁵

But now consider any non-evaluative description of a person's life; in particular, consider the description which consists in the conjunction, for every moment m of the life in question, of the proposition that the person is in agony at m. How is the sceptical theist to conclude on the basis of this description that the life in question is not on the whole a good one? For all she knows, all or most moments in the life instantiate good-making properties of which she is unaware. So, for all she knows, the life is on the whole a good one. And what goes for one life presumably goes for all lives. So there is no way for the sceptical theist to conclude that The Worst Scenario is one in which any life violates Swinburne's criterion.

V

Perhaps, though, it is Swinburne's principle's focus on the features of individual lives that causes problems for Bergmann's response to Rowe. Perhaps a principle appealing to features of populations rather than individuals would not be subject to the objections I have just advanced. For example, it might be that while we are not in a position to know that God would not permit the occurrence of a single life which was not on the whole a good one, we are in a position to know that God would not create a world in which all lives (or a sufficient proportion of lives) had this character.

But this population-based response is not promising; indeed, it solves neither the Justification Problem nor the Recognition Problem. With respect to the Justification Problem, any putative population-level principle must face the same question as Swinburne's principle: might God not be justified in violating it if a state of affairs which does not satisfy it either instantiates goods of which the sceptical theist is unaware or is a necessary concomitant of a state of affairs which instantiates such goods? It is difficult to see how the sceptical theist could justify a negative answer to this question. With respect to the Recognition Problem, it seems that we can apply a population-based principle to rule out Rowe's scenario only if we are in a position to know that any particular life which is nothing more than excruciating pain from birth to death is not on the whole a good one. But we have already seen that the sceptical theist is unable plausibly to claim that she is in a position to know this.

I conclude that Bergmann's strategy of appealing to a priori constraints on the magnitude and distribution of evil in a theistic world cannot help the sceptical theist avoid problems of insensitivity. But perhaps there is another strategy available to the sceptical theist. Her criterion of adequacy, after all, demands only that the actual world's catalogue of evils fail to justify disbelief in theism. Perhaps, then, she can argue that the actual world's catalogue of evils falls somewhere along a continuum, and that while it in fact fails to justify disbelief in theism – so that (The Criterion) is satisfied – there are many possible catalogues which, while not a priori inconsistent with theism, would serve as decisive evidence against it. In what follows, I argue that this strategy also fails, and that its failure demonstrates that there is no plausible way for the sceptical theist to escape arguments from insensitivity.

Before proceeding, a note: although I hope to have shown that there can be no a priori sceptical-theistic argument for the impossibility in a theistic world of human lives which are nothing more than successions of agonizing moments, the sceptical theist will perhaps allege that the probability that such lives occur conditional on theism is sufficiently low as to justify disbelief in theism in The Worst Scenario. Though I believe that this probabilistic claim hangs uncomfortably together with (Sceptical Thesis) and (ST1-4), I am prepared to grant it for the purposes of the arguments below. Instead of focusing on scenarios involving uniformly agonizing lives, therefore, I will focus on scenarios which contain many duplicates of the most miserable lives ever actually lived. This shift in focus allows me to sidestep the imagined sceptical-theistic objection to Rowe's argument, since the actual occurrence of said lives requires the sceptical theist to hold that they are not highly improbable conditional on theism, on pain of conceding that the actual world's current datum justifies disbelief in theism.

VI

My argument begins with the observation that (The Criterion) is not as easily satisfied as the sceptical theist might presume; in order to serve her purposes, it must be strengthened. As Rea points out:

If Rea's claim is correct, the ‘decently principled’ sceptical theist is committed to a stronger principle than (The Criterion). (The Criterion), the reader will recall, is a claim about the history of the actual world up to the present moment. It is thus formally consistent with (The Criterion) to maintain that ‘if the world carries on pretty much as it has to date’ for many generations – that is, if the number and severity of new prima facie evil events per unit of time remains constant as the value of t increases – then ${\rm {\frak D}}_{\langle @,t\rangle} $ will eventually constitute evidence strong enough to justify disbelief in theism. Let us call the imagined version of sceptical theism which endorses this claim de facto sceptical theism. Rea's claim is then that de facto sceptical theism, though formally consistent, is not ‘decently principled’.

The thought behind Rea's claim here seems to be that if the de facto sceptical theist's evidence supports the claim that the world will carry on ‘pretty much as it has to date’ for many generations, she must violate a version of the principle of Reflection. The principle of Reflection (van Fraassen (Reference van Fraassen1984) ) is a norm on subjective probabilities (hereinafter credences). It states, roughly, that if a rational agent learns that she will soon come to have credence x in some proposition p, she ought now to have credence x in p. More precisely, Reflection requires that

• (Reflection): C ₀(p|C ₁(p) = x) = x,

where C ₀ is the earlier credence function and C ₁ is the later.Footnote ⁶

We have couched our argument in terms of evidential probabilities rather than credences, so it would be a mistake to claim that the de facto sceptical theist violates Reflection. Nevertheless, the idea behind Reflection – that if it is certain on a body of evidence e that the balance of future evidence e′ will support a proposition p to extent x, then e itself supports p to extent x – applies just as forcefully to the probabilities at issue in the evidential argument. And surely it is uncontentious that the conjunction of our neutral body of evidence and the actual world's datum up to the present moment overwhelmingly supports the proposition that the world will carry on pretty much as it has to date for many generations. But then the de facto sceptical theist has every reason to think that, in light of a Reflection principle modified so as to constrain evidential probabilities, her response to the evidential argument does not now succeed. For then, by her lights, our neutral body of evidence (together with the world's present datum and her other commitments) strongly supports the proposition that decisive evidence against theism will eventually accrue and remain undefeated. It follows by our modified Reflection principle that the actual world's present datum strongly supports disbelief in theism (relative to this neutral evidence base). De facto sceptical theism cannot coherently be combined with strong evidence for an extended future which is in relevant respects like the present.

There would be no problem for the de facto sceptical theist, of course, if the evidence supported the claim that the world would end shortly or change for the better markedly enough that its datum would never constitute substantial evidence against theism. If, for example, the evidence pointed to an imminent eschaton or a radical change in the moral character of humankind, our modified principle of Reflection would pose no problem for her position. But I take it that no-one, and a fortiori no sceptical theist, has evidence of the required kind. It seems, then, that de facto sceptical theism must be rejected as a tenable sceptical theistic position. The sceptical theist is committed to the conjunctive claim that things will remain substantially the way they have in the past for many generations and that the world's datum will never constitute strong evidence against theism. Let us refer to the conjunction of sceptical theism with the claim that, even if the world carries on ‘pretty much as it has to date’ for many generations, its datum will never justify disbelief in theism, as de jure sceptical theism.Footnote ⁷

It will be convenient to explore the consequences of de jure sceptical theism in two stages. The most radical form of de jure sceptical theism, which holds that the world will carry on ‘pretty much as it has to date’ indefinitely (let us therefore call it infinitary de jure sceptical theism), is also the version most straightforwardly subject to problems of insensitivity. I will therefore begin by discussing infinitary de jure sceptical theism. The insensitivity which afflicts more moderate versions of the position may then be appreciated by way of comparison with that which afflicts infinitary de jure sceptical theism.

VII

Consider, then, the infinitary de jure sceptical theist. As a consequence of holding that history will continue on ‘pretty much as it has to date’ indefinitely, she must also hold that prima facie evils will continue to accrue at roughly the same rate indefinitely, and therefore that, for any finite number n, there will come a time t such that ${\rm {\frak D}}_{\langle @,t\rangle} $ is a catalogue of at least n prima facie evils. If the infinitary de jure sceptical theist is to coherently combine her theism with her beliefs about the future and our modified Reflection principle, she must therefore hold, for every finite number n, that the actual world's datum can be a catalogue of at least n prima facie evils without justifying disbelief in theism. With a slight simplification involving convergent series, this suggests that, if ‘${\rm {\frak n}}$’ is a constant which rigidly designates the present and b is our neutral body of evidence, the infinitary de jure sceptical theist should endorse:

• (Zero Marginal Disconfirmation): For all prima facie evil events e and times t ≥ ${\rm {\frak n}}$, if e is a token event of a type another token of which has occurred prior to ${\rm {\frak n}}$, and if e occurs at t, then P(theism|${\rm {\cal P}}$(e) & ${\rm {\frak D}}$_〈@,${}_{\rm {\frak n}}$_〉 & b) = P(theism|${\rm {\frak D}}$_〈@,${}_{\rm {\frak n}}$_〉 & b).

(Zero Marginal Disconfirmation) allows only the tokens of any type of prima facie evil which has never previously occurred to disconfirm theism.Footnote ⁸ It leaves open the question of whether such first tokens do in fact disconfirm theism.Footnote ⁹ There are two answers to this question available to the sceptical theist. First, there is:

• (Zero Total Disconfirmation): For all times $ t, P(theism \vert{\rm {\frak D}}_{\langle @,t\rangle} $ & b) = $P(theism \vert b)$.

According to (Zero Total Disconfirmation), at no time in the world's history does its datum at that time constitute evidence against theism. This is a strong thesis, and one which sceptical theists have reason to be hesitant to accept.Footnote ¹⁰ Rejecting it yields a view to which I will refer as

• (Negligible Total Disconfirmation): (Zero Marginal Disconfirmation) is true and $P(theism \vert{\rm {\frak D}}_{\langle @,{\rm {\frak n}}\rangle} \, \& \, b) \lt P(theism \vert b)$, though $P(theism \vert{\rm {\frak D}}_{\langle @,{\rm {\frak n}}\rangle} \, \& \, b) \approx $ P(theism|b).

(Negligible Total Disconfirmation) might be motivated by the idea that once we update our neutral body of evidence with the antecedently improbable (conditional on theism) proposition that the world contains instances of a particular variety of prima facie evil, our augmented body of evidence suggests that if theism is true, tokens of that type must have either intrinsic properties or modal connections to other events which justify God in permitting their occurrence, and so the occurrence of further tokens of the same type is no longer improbable conditional on theism.

Whether the sceptical theist opts for (Zero Total Disconfirmation) or (Marginal Total Disconfirmation), the justification she provides for her position will appeal to (Sceptical Thesis): it is because of the epistemic consequences of our cognitive limitations vis-à-vis value and modality that our continued observation of prima facie evils will never tip the evidential scales away from theism.

VIII

But perhaps we have been too hasty. After all, the argument I have just presented is superficially similar to a class of specious arguments like the following:

• [Coins]: Grimgerde is interested in determining whether a certain coin is fair. She knows that the coin is either fair or biased heavily towards tails, and she thinks these two possibilities are equally probable prior to beginning to flip the coin. She flips the coin one hundred times and notes that it comes up heads on exactly fifty-three of these flips. On this basis, she comes to believe that the coin is fair. She enjoys the experiment so much that she resolves to continue flipping the coin for all of eternity. Since she expects that things will carry on pretty much as they have to date, she expects that the coin will continue to land tails in roughly the same proportion of trials as it did in the first hundred. But this means that she must expect that, for any finite number n, her evidence will eventually entail that the coin has landed tails at least n times. If she is to rationally maintain her belief in the fairness of the coin given that she foresees a future containing infinitely many tails outcomes, she must hold that new tails outcomes fail to disconfirm the hypothesis that the coin is fair.

[Coins] illicitly moves from the claim that Grimgerde rationally expects that her evidence will eventually include an arbitrarily large number of tails outcomes to the claim that she must hold that new tails outcomes fail to disconfirm the hypothesis that the coin is fair. For she can rationally expect eventually to observe an arbitrarily large number of tails outcomes without holding that each one fails to disconfirm the hypothesis that the coin is fair if she also rationally expects to observe an arbitrarily large number of heads outcomes and holds that each heads outcome confirms the hypothesis that the coin is fair. If the sceptical theist can help herself to the same sort of move, she can avoid commitment to (Zero Marginal Disconfirmation).

The natural move for the sceptical theist to make here is to claim that substantial evidence for theism accrues as time passes. For surely if history's evils count as evidence against theism, history's goods also count as evidence for it. If this is true, then the sceptical theist can simply endorse (The Criterion) and confidently predict that, if circumstances unfold much as they have in the past, the total evidence about what has happened will never, on balance, justify disbelief in theism.

Unfortunately, this response is dialectically confused: one who adopts it ipso facto adopts a view of the evidential significance of historical events on which the sceptical theist's approach to the evidential argument loses its appeal. If circumstances relating to the magnitudes and distributions of goods and evils have unfolded in such a way as, on balance, to provide evidence for theism, or to provide no evidence either way, then there is no need to take on the commitments of the sceptical theist to reply to our protagonist's argument. (Sceptical Thesis) merely dampens the evidential impact of our observations vis-à-vis theism; if the balance of the evidence points towards theism given (Sceptical Thesis), then it does so given the falsity of (Sceptical Thesis), as well. So if, on balance, history provides evidence for theism, or no evidence either way, the theist can simply note this and observe that, since this is the case, she has reason to believe that the future will continue to provide such evidence.

Indeed, it is presupposed throughout the literature on the problem of evil that there is an asymmetry between the evidential significance of prima facie evils and the evidential significance of prima facie goods: actual prima facie evils form the basis of a case against theism, while actual prima facie goods do not form the basis of a similar case against atheism.Footnote ¹¹ But to acknowledge this is to acknowledge that, whereas Grimgerde can avoid the charge of maximal insensitivity by appealing to the evidential significance of the heads-outcomes she observes, the sceptical theist cannot make a similar move by appealing to the evidential significance of the prima facie goods she observes.

IX

Appreciating the nature of the infinitary de jure sceptical theist's commitment to (Zero Marginal Disconfirmation) clarifies the susceptibility of her position to problems of insensitivity. We have seen that there are two infinitary-de-jure-sceptical-theistic strategies for dealing with the constant addition of new prima facie evils to the course of history. Regardless of which of these two strategies the infinitary de jure sceptical theist adopts, I claim, she must hold that adding instances of a particular type of prima facie evil to the history of a world does not increase the extent to which its datum disconfirms theism. That is, she must endorse:

• (Necessary Zero Marginal Disconfirmation): For all prima facie evil events e, worlds w, and pairs of times t′, t such that t′ ≥ t, if e is a token event of a type another token of which has occurred prior to t at w, and if e occurs at t′ at w, then $P(theism \vert{\rm {\cal P}}\left( e \right) \; {\rm \&} \, {\rm {\frak D}}_{\langle w,t\rangle} \; {\rm \&} \, b) = P(theism \vert{\rm {\frak D}}_{\langle w,t\rangle} \; {\rm \&} \, b)$.

The idea behind (Necessary Zero Marginal Disconfirmation) is that, regardless of the history of a world, only the first occurrence of a given type of prima facie evil can disconfirm theism. Either sceptical-theistic strategy for motivating (Zero Marginal Disconfirmation) suggests this result. Given (Sceptical Thesis), the infinitary de jure sceptical theist argues, it is reasonable to hold that evidence of prima facie evil does not disconfirm theism at all, or that once we observe the first token of a given type of prima facie evil, we have evidence that God is justified in permitting tokens of that type, and the occurrence of subsequent tokens does not further disconfirm theism. Neither of these sceptical-theistic strategies for motivating (Zero Marginal Disconfirmation) depends for its plausibility on any assumption about the character of the past or the future; each proceeds solely with reference to (Sceptical Thesis) and (ST1-4). It follows that each serves just as effectively to motivate (Necessary Zero Marginal Disconfirmation).Footnote ¹²

Now, it is uncontentious that history has included human lives of a very miserable sort – so miserable a sort that, if every human life were so miserable, and if there were an infinite number of human lives, the misery and quantity of those lives would constitute evidence strong enough to justify disbelief in theism. Let us refer to the scenario just described, in which each of the infinite number of human lives ever lived is as miserable as the most miserable actual human life, as The Awful Scenario. Clearly, the datum of The Awful Scenario after countless generations would justify disbelief in theism relative to a neutral evidence base. This claim is admittedly not quite as obvious as Rowe's claim about The Worst Scenario, but it is obvious enough.

Because the infinitary de jure sceptical theist must endorse (Necessary Zero Marginal Disconfirmation), she must also endorse the claim that the datum of The Awful Scenario after countless generations constitutes evidence against theism no stronger than the actual world's datum. The consistent infinitary de jure sceptical theist is thus maximally insensitive with respect to evidence about miserable lives – she holds that no amount of such evidence could justify disbelief in theism. But maximal insensitivity is an intolerable commitment; it follows that infinitary de jure sceptical theism is not a tenable theistic position.

X

Perhaps the infinitary de jure sceptical theist will seek to sever the link between (Zero Marginal Disconfirmation) and (Necessary Zero Marginal Disconfirmation). In the actual world, her argument might proceed, it is true that newly observed prima facie evils fail to disconfirm theism. But this is because they are distributed as they are in the actual world. Had they been differently distributed – had, for example, every human career been a miserable one – then the world's datum would have constituted decisive evidence against theism. Crucially, moreover, the evidential significance of certain distributions of prima facie evils is not to be explained in terms of the cumulative effect of their constituents; certain distributions of prima facie evils have evidential significance which cannot be predicted on the basis of the evidential significance of the individual prima facie evils of which they are composed.Footnote ¹³

Unfortunately, this suggestion on behalf of the infinitary de jure sceptical theist is subject to versions of both the Recognition Problem and the Justification Problem. Recall that the Recognition Problem is the problem of concluding, on the basis of any non-evaluative description of an individual's life, that that life is not on the whole a good one. The sceptical theist makes much of the unreasonableness of concluding, on the basis of the totality of our evidence about the most miserable lives ever lived, that those lives were on the whole bad ones. If this sort of scepticism is to do the kind of work the sceptical theist needs it to do, it must entail that the probability that the most miserable actual lives are on the whole good ones is not particularly low. But if it is unreasonable to conclude that such lives are on the whole bad ones, it must also be unreasonable to conclude that a world containing any particular distribution of these most miserable lives, such as The Awful Scenario, is on the whole a bad one. How, after all, could the probability that The Awful Scenario is a bad one exceed the probability that the miserable lives lived in the actual world are on the whole bad ones? And if the infinitary de jure sceptical theist must suspend judgement about whether The Awful Scenario is on the whole a bad one, she cannot claim that its occurrence is particularly improbable given theism.

The sceptical theist also urges that God may be justified in permitting the occurrence of the most miserable actual lives to secure some outweighing good or avoid some greater evil. Even supposing that the most miserable actual lives are not on the whole good ones, to the extent that the sceptical theist relies on this appeal to modal connections beyond our ken, she faces a version of the Justification Problem – the problem of defending putative a priori constraints on the magnitude or distribution of ultima facie evils in a theistic world. For suppose she holds that the distribution of lives which are not on the whole good ones in The Awful Scenario is highly improbable conditional on theism. Then it must be highly improbable that each of the miserable lives in The Awful Scenario is modally connected to some outweighing good or greater evil. But why should the modal connections to which the sceptical theist appeals when considering the actual world be sensitive to the distribution of miserable lives? Perhaps, after all, each such life makes possible the flourishing of an unseen angelic person, whose metaphysical and moral perfection contributes more to the world than the sufferer's suffering detracts from it. Or perhaps the prima facie value-neutral events modally connected with miserable human lives, such as the conversion of oxygen into carbon dioxide, instantiate, unbeknownst to us, goods which themselves justify lifetimes of human misery. In either scenario, the modal connections between miserable lives and other value-bearing states of affairs would obtain regardless of the proportion of all human lives which were miserable. To the extent that the infinitary de jure sceptical theist can offer no reason for thinking that such scenarios are less probable than ones on which said modal connections are sensitive to proportional facts – and it is difficult to see how she could, given her endorsement of (Sceptical Thesis) and (ST1-4) – she cannot plausibly claim that facts about the distribution of miserable lives allow her to escape the threat of maximal insensitivity.

XI

The difficulties associated with infinitary de jure sceptical theism do not, of course, suffice to establish a problem for sceptical theism in general. For while the sceptical theist lacks evidence of an imminent eschaton, she has good scientific evidence that life as we know it will not continue indefinitely. Eventually, current theories suggest, the sun will begin to die, and this process will lead to changes which make the earth uninhabitable. If the rate at which the accrual of prima facie evils disconfirms theism is nonzero but so small that the world's datum will not constitute strong evidence against theism within the timeframe imposed by humanity's impending extinction, perhaps the sceptical theist can escape both the charge of insensitivity and the threat of violating our modified Reflection principle. In other words, the sceptical theist may seek refuge in:

• (Negligible Marginal Disconfirmation): For all prima facie evil events e and times t ≥ ${\rm {\frak n}}$, if e is a token event of a type another token of which has occurred prior to ${\rm {\frak n}}$, and if e occurs immediately after t, then $P(theism \vert{\rm {\cal P}}\left( e \right) \; {\rm \&} \, {\rm {\frak D}}_{\langle @,t\rangle} \; {\rm \&} \, b) \approx P(theism \vert{\rm {\frak D}}_{\langle @,t\rangle} \; {\rm \&} \, b)$.

(Negligible Marginal Disconfirmation) claims that no token prima facie evil of a type which has already been instantiated can disconfirm theism to more than a negligible extent. Let us refer to the brand of de jure sceptical theism which embraces (Negligible Marginal Disconfirmation) and holds that the rationality of theism is ‘saved by the bell’ of the eventual extinction of humankind as finitary de jure sceptical theism. I claim that finitary de jure sceptical theism is subject to an insensitivity problem which is only marginally less serious than the one faced by its infinitary congener.

A conservative estimate of the total number of human lives which could be lived within the constraints imposed by the death of the sun is 10¹⁶, or ten million million billion.Footnote ¹⁴ If things carry on ‘pretty much as they have to date’ for the rest of human history, a small but significant proportion of these lives will be comparable in character to the most miserable which have so far been lived. Let us refer to the number of miserable human lives that the finitary de jure sceptical theist must foresee will be lived prior to the extinction of the human species as n. Certainly, n is a very large number – large enough that it would take many generations for a human community the size of a small city to produce n human lives. Consider, then, a world in which the only human lives ever lived are n miserable lives. We can imagine that these n miserable lives are temporally arranged into many generations of inhabitants of a single small city, and that the world is spatially and temporally bounded around them, so that nothing exists outside the city, and time begins with the beginning of the first miserable life and ends with the end of the last. Let us call this world The Bad Enough Scenario.

The Bad Enough Scenario is not quite as bad as The Worst Scenario or The Awful Scenario – but it is Bad Enough. If every human life were as miserable as the most miserable actual life, and if there were many human lives distributed over many generations, this would constitute evidence sufficient to justify disbelief in theism. But the finitary de jure sceptical theist cannot hold that the datum of The Bad Enough Scenario at the end of its history would constitute sufficient evidence to justify disbelief in theism. We have already seen that the de jure sceptical theist cannot appeal to proportional facts about the distribution of prima facie evils to defend herself against the charge of insensitivity. So, if the finitary de jure sceptical theist is willing to claim that the occurrence of n miserable lives prior to the extinction of the human species does not constitute strong evidence against theism, she must be willing to claim that the datum of The Bad Enough Scenario at the end of its history would also fail to constitute strong evidence against theism. That is, she must be radically insensitive with respect to evidence about the occurrence of miserable lives – she must concede that not even evidence of extreme and pervasive suffering could justify disbelief in theism. This commitment is hardly less objectionable than a commitment to maximal insensitivity.

XII

By way of concluding, I would like briefly to consider a common objection to my general strategy. My argument has relied at crucial points on assigning particular conditional probabilities to propositions relative to bodies of evidence. There is a common dialectical move in the sceptical theism literature whereby the critic of the evidential argument claims that such probabilities are ‘inscrutable’ (see, for example, van Inwagen (Reference van Inwagen1998) ). The friend of sceptical theism might hope to fall back on this strategy as a response to my argument.

Unfortunately for the sceptical theist, the best she can do by way of using a notion of inscrutability to escape the argument for radical insensitivity is to systematically replace the claim that theism has a low probability conditional on certain bodies of evidence with the claim that theism has an inscrutable probability conditional on those bodies of evidence. This is no way of avoiding the charge of radical insensitivity, for the claim that the probability of theism conditional on the observation of The Worst Scenario is inscrutable entails the claim that the observation of The Worst Scenario does not justify disbelief in theism.

If my arguments above are sound, consistent sceptical theists are radically insensitive theists. This consequence ought to be deemed intolerable by theists. As the dust settles, then, our nullifidian protagonist stands vindicated. Her evidential argument remains, at least pro tem, a serious problem for theism.Footnote ¹⁵