Skeptical theism and divine permission: a reply to Anderson

  • John Danaher


Skeptical theism (ST) may undercut the key inference in the evidential argument from evil, but it does so at a cost. If ST is true, then we lose our ability to assess the all things considered (ATC) value of natural events and states of affairs. And if we lose that ability, a whole slew of undesirable consequences follow. So goes a common consequential critique of ST. In a recent article, Anderson has argued that this consequential critique is flawed. Anderson claims that ST only has the consequence that we lack epistemic access to potentially God-justifying reasons for permitting a prima facie “bad” (or “evil”) event. But this is very different from lacking epistemic access to the ATC value of such events. God could have an (unknowable) reason for not intervening to prevent E and yet E could still be (knowably) ATC-bad. Ingenious though it is, this article argues that Anderson’s attempted defence of ST is flawed. This is for two reasons. First, and most importantly, the consequential critique does not rely on the questionable assumption he identifies. Indeed, the argument can be made quite easily by relying purely on Anderson’s distinction between God-justifying reasons for permitting E and the ATC value of E. And second, Anderson’s defence of his position, if correct, would serve to undermine the foundations of ST.


Skeptical theism Moral ignorance Problem of evil  All things considered value 


In Rowe’s classic presentation,1 the evidential argument from evil proposes that from the seemingly gratuitous nature of human and animal suffering we can infer the actually gratuitous nature of that suffering. This inference is common to many well-known versions of the evidential argument.2 Skeptical Theism (ST) challenges this inference. As leading proponents of ST put it,3 given the sorts of creatures that we are, we have no reason to think that the kinds of goods and evils (and entailment relations between them) of which we are aware are representative of the kinds of goods and evils (and entailment relations) there actually are.

Critics of ST have argued that the skeptical principles which undergird it have undesirable consequences.4 These range from paralysing effects on moral reasoning,5 to undercutting defeaters for arguments from design6 and the reliability of scripture.7 One of the chief problems seems to be that if ST is true, it calls into question our ability to assess the all-things-considered (ATC) value of events and states of affairs.8 ST is then, to quote Hasker, “all too skeptical”.9

In a recent article, Anderson has argued that this critique of ST is mistaken.10 He claims that the implication motivating the critique—viz. if ST is true, we lack epistemic access to the ATC value of events and states of affairs—relies on the assumption that God could only allow an instance of great suffering to occur if it was logically necessary for obtaining a greater good. But this is not the case. God could have justifiable reasons for not intervening to prevent great suffering, without it being the true that the suffering is logically necessary for obtaining a greater good. Since ST only calls into question our epistemic access to God’s reasons, it follows that it need not undercut our ability to assess the ATC value of events and states of affairs.

Anderson’s critique has two admirable qualities: (i) it provides a novel insight into the implications and assumptions at the heart of this debate; and (ii) it offers an ingenious “way out” for defenders of ST. Despite these admirable qualities, this article argues that Anderson’s purported “way out” is little more than a dead end. This is for two reasons. First, the consequential critique need not rely on the assumption identified by Anderson. And second, Anderson’s attempted support for his assumption would destabilise the foundations of ST.

The remainder of the article proceeds as follows. Section “Understanding Anderson’s critique” clarifies the consequential critique of ST as well as Anderson’s response to it. Section “The robustness of the consequential critique” shows that the consequential critique need not rely on the questionable assumption identified by Anderson. Section “Foundations and conclusions” concludes by discussing the foundational problems caused by Anderson’s defence of his thesis.

Understanding Anderson’s critique

The first task is to clarify the consequential critique of ST and Anderson’s response to it. This will be done in three stages. First, the core principles of ST will be identified, and the distinction between its foundations and consequences highlighted. Second, a clear statement of the consequential critique will be given. And third, a brief synopsis of Anderson’s argument shall be presented.

The nature of skeptical theism

Bergmann identifies four basic principles11 which sum up the commitments of the skeptical theist. In essence they hold, that we have no good reason for thinking that the possible goods and evils (and entailment relations between those goods and evils) are representative of all the possible goods and evils (and entailment relations) there are.12 Although there are other statements of ST,13 Bergmann’s will be the focus in this article since they it is also the focus of Anderson’s discussion.

Two clarifications of Bergmann’s position are in order. The first is that the representativeness at stake here is representativeness relative to the property of figuring in a God-justifying reason for permitting apparently gratuitous evil.14 The central claim of ST is that we lack epistemic access to potentially God-justifying reasons for allowing evil to exist. So, when we come across an apparently gratuitous case of suffering, we cannot infer that it is actually gratuitous. The inability to do this undercuts the standard version of evidential argument from evil. The second clarification is that ST does not simply serve to reduce the subjective probability we attach to something’s being an instance of gratuitous suffering, rather it completely undermines our ability to assign such probabilities.15 This is for good reason. If ST merely reduced our confidence that an event (E) was actually gratuitous from, say, 0.8 to 0.01, then the fact that there are many similar events would be sufficient for the evidential argument from evil to go through.16

In sum then, the core conclusion of ST (in its Bergmannian form) is:

ST - Given our cognitive limitations, we lack epistemic access to God-justifying reasons for permitting apparently gratuitous evil to such an extent that we cannot assign meaningful subjective probabilities to an event’s being actually gratuitously evil based on the mere fact that it appears to be gratuitously evil.

A critic of ST can approach the doctrine from one of two directions. In the first instance, they can challenge its foundations, i.e. call into question the arguments used to motivate the skeptical principles. The critical evaluation of these foundational arguments can be called the foundational critique (or the foundational debate). This is to be contrasted with the second critical perspective on ST, which focuses on its consequences, not its foundations. The essence of the consequential critique is that ST has unpalatable consequences, and hence cannot be sustained by the reasonable theist. The critical evaluation of the consequences of ST shall be called the consequential critique (or debate) in the remainder of this article. Anderson’s argument tackles the consequential critique, so focus shall primarily be on that in what follows. Nevertheless, Anderson’s argument reveals something interesting about the foundational debate too, and so I return to that in the conclusion.

The consequential critique in its moral form

The consequential critique can take several forms. Lovering17 offers a highly abstract consequential critique of ST, suggesting that since some knowledge about what God would do is assumed by most arguments for his existence, and since ST calls into question such assumptions, most arguments for the existence of God are undercut by ST. In a similar vein, but with a more specific focus, Wielenberg argues that a lack of epistemic access to God’s reasons for action, undermines our confidence in the reliability of scripture.18

Interesting though these forms of the consequential critique are, they are not the direct target of Anderson’s response. He focuses on a slightly different, arguably quite powerful, version of critique. This version of the critique claims that ST casts into doubt our ability to assess the ATC value of events and states of affairs, and that this, in turn, has several unpalatable consequences. It can be stated as follows:
  1. (1)

    If ST is true, then we lack epistemic access to potentially God-justifying reasons for allowing apparently gratuitous evil to exist.

  2. (2)

    If we lack epistemic access to potentially God-justifying reasons for allowing apparently gratuitous evil to exist, then we lack the ability to assess the ATC value of events and states of affair.

  3. (3)

    Therefore, if ST is true, we lack the ability to assess the ATC value of events and states of affairs.

  4. (4)

    If we lack the ability to assess the ATC value of events and states of affairs, a whole slew of unpalatable consequences follow.

  5. (5)

    Therefore, ST has unpalatable consequences.

The first premise of this argument is relatively uncontroversial. As pointed out above, proponents of ST such as Bergmann accept this statement of their view, so there is no reason to question it here. The second premise is rather more doubtful and is, indeed, the primary target of Anderson’s response to the consequential critique. I shall discuss this at length below. The fourth premise is rather sketchy in its current form (what are these unpalatable consequences?) and it is worth giving a specific example to support it before moving on.

The example I will focus on examines the effects of an inability to assess ATC value on moral reasoning, moral obligations and moral knowledge. The example is derived from Maitzen.19 Imagine the following the scenario. You and a friend are out fishing. He falls into the water and starts to drown because he cannot swim. You, on the other hand, are a strong swimmer with several qualifications in life-saving. Should you intervene to save him? The obvious answer appears to be “yes”. Indeed, most people would hold that there is a strong obligation to intervene in those circumstances. To get rid of the belief in that obligation would be unpalatable.

But what grounds the belief in such an obligation? Two answers seem likely. The first assumes a deontological ethical framework and holds there is a duty to save the life of another that is independent of the consequences of so doing. The second assumes a consequentialist ethical framework and holds that you ought to intervene because doing so is, ATC, better. The deontological answer is not implausible, but would seem to be ruled out in this particular dialectic because ST itself presupposes a consequentialist framework.20 And this is problematic because the consequentialist answer requires an ability to assess the ATC value of the events in question: only if your friend’s death is ATC bad (or, rather, worse than his continued existence) does the obligation arise. But according to the consequential critic, ST undermines our ability make such an assessment. Hence, ST leads to a severe moral ignorance, and thereby undermines our commitment to core moral obligations. This is clearly an unpalatable consequence.21

Anderson’s argument

Anderson now enters the debate. He points out that the critique hinges on the implication stated in premise (2):

(2) If we lack epistemic access to potentially God-justifying reasons for allowing apparently gratuitous evil to exist, then we lack the ability to assess the ATC value of events and states of affair.

And that this implication is in turn being driven by something he calls the “Assumption”:22

Assumption: God’s permission of some instance of suffering E is justified only if the occurrence of E is itself logically required for the attaining of an outweighing benefit.”

But, Anderson argues, this assumption is flawed. We need to understand why he thinks this, and why he thinks this undermines the ATC-value version of consequential critique.

The assumption itself is superficially plausible. According to the classical position, God is omnipotent, omniscient and omnibenevolent. Being omnibenevolent, God desires the good and abjures the bad: he wants to ensure that everything is for the best. Being omniscient, God knows when bad things are happening. Finally, being omnipotent, God is maximally powerful.23 This last characteristic is particularly important in the present context. Since God wishes to ensure that everything is for the best, and since he has the power to achieve this (within the logical limits for a maximally excellent being), it seems to follow that God could only permit instances of suffering to occur if those instances were logically necessary for some outweighing benefit to obtain. From this, it follows that if some intrinsically bad state of affairs is occurring, and God does nothing to prevent it, that state of affairs must be logically necessary for some outweighing benefit. This gives us the assumption.24
  1. (6)

    God is omnipotent: he can do everything that it is logically possible for a maximally excellent being to do.

  2. (7)

    God is omniscient: he knows everything it is possible to know.

  3. (8)

    God is omnibenevolent: he wants everything to be (morally speaking) for the best.

  4. (9)

    Therefore, God’s permission of an instance of suffering E is justified only if E’s occurrence was logically required for the attaining of an outweighing benefit.

The derivation of (9) from the preceding premises is certainly opaque. Anderson rejects it on the grounds that it conflates two distinct notions: (i) whether God’s permitting E to occur is logically required for an outweighing good to obtain; and (ii) whether E’s actual occurrence is logically required for an outweighing good to obtain. God’s permitting E (or non-intervention to prevent E) might be logically necessary for obtaining a good without it also being the case that E is logically necessary for an outweighing good to occur. The occurrence of E may be ATC bad; but God’s non-intervention into E may, nevertheless, be ATC good. In other words, from the conjunction of (6), (7) and (8), it follows that:25

(9*) If God permits E to occur, it must be better, ATC, that he permit it than that he not permit it.26

But not that:

(9**) If God permits E to occur, it must be better, ATC, that E occur than that it not occur.

On a purely formal level, Anderson is correct: (9*) is not logically equivalent to (9**) so if God’s tri-omni properties only imply (9*), the assumption is flawed. But are there good reasons for thinking this is right, that God’s non-intervention to prevent E can be ATC good even if the occurrence of E is not?

Anderson offers three arguments in support of this contention. The first is an obvious one: God may justifiably permit a free agent to engage in acts that cause suffering, without it being the case that the suffering so caused is necessary for some outweighing good. This argument, of course, trespasses on the territory covered by Plantinga’s famous free will defence to the logical problem of evil. Anderson is claiming that the reasons that justify allowing agents to have free will may cover the consequences of their free will too, even when this leads to logically unnecessary suffering. But even then Anderson is somewhat reluctant to go all the way, and reins-in his conclusion somewhat by arguing that if the suffering reaches a certain threshold, God may be required to step in and constrain the free will of the agent. This is a troubling concession, and one to which I shall return.

The second and third arguments are rather more interesting, though, as Anderson notes, more “vague and speculative”. The second argument might be dubbed the Sovereign or Judge Argument. It points to the fact that God, as the creator and sovereign of the world, may face considerations that we, as ordinary agents within the world, do not. Every time he intervenes to prevent suffering, he must interfere with the natural goings on in the world. This could have costs—costs beyond our ability to understand—and these costs could weigh against continual intervention to prevent suffering. The analogy is with a judge deciding important issues of law. He or she may see things, appreciate costs, that we do not. We should adopt a stance of epistemic humility toward them, just as we should with God. The third argument follows very similar lines. God faces a choice when it comes to the intervention into and control of the world. He could adopt a case-by-case approach, choosing to intervene in particular scenarios as he saw fit; he could adopt a principles-based approach, choosing general principles and sticking to them over time; or he could adopt some combination of both. If he adopts a principles-based approach, we might expect non-intervention in at least some cases of suffering, without this implying that the suffering is ATC good. The important point, from the ST perspective, is that we cannot expect to know what kinds of reasons might apply to God when making this kind of decision. Epistemic humility is once again required.

To these three arguments may be added a fourth. One reason why God may permit E to occur, without it necessarily following that E’s occurrence is ATC good, is because it might be necessary for us to intervene to prevent E. Thus, for example, in the drowning friend example given above, one reason why your friend’s drowning may be gratuitous, but God’s permission of it not, is that it may be necessary for you to intervene to prevent it in order to secure a greater good. In keeping with the ST position, one could argue that we should be epistemically humble about knowing whether or not that is true in a particular case. Indeed, this may be the most obvious and compelling reason for thinking God may permit E, even if E’s occurrence is not logically necessary for an outweighing good.

So much for Anderson’s critique of the assumption. What effect does this actually have on the consequential critique? As mentioned above, Anderson’s key claim is that the assumption is what makes the implication (premise (2) of the original argument) plausible. So if he has succeeded in undercutting the assumption, he believes he has succeeded in defeating this particular version of the consequential critique. But why think that? Anderson argues that premise (2) presupposes a tight logical connection between the reasons that justify God’s permitting E to occur and the factors that render an event or state of affairs ATC good. To use a set-theoretical metaphor, premise (2) presupposes that the set of reasons justifying God’s permitting E forms a perfect union with the set of factors determining the overall value of an event or state of affairs. Thus, if we lose epistemic access to the former set—as ST proposes—we also lose epistemic access to the latter. This is traced out in the chain of inferences below:
  1. (10)

    Because of omnipotence, every natural event or state of affairs is permitted by God.

  2. (11)

    Because God is a perfectly moral being, his decisions about which events or states of affairs to permit are motivated by justifying moral reasons.

  3. (12)

    Therefore, the ATC value of a natural event or state of affairs is directly determined by the reasons God has for permitting it to occur.

  4. (13)

    Therefore, the factors determining the ATC value of an event of state of affairs are equivalent to the God-justifying reasons for permitting that event or state of affairs.

If this chain of inferences were correct, the consequential critique would succeed. But Anderson’s argument undermines it. In effect, he claims that (10) and (11) do not necessarily imply (12) and (13). The tight logical connection between God-justifying reasons and the factors determining the ATC value of an event or state of affairs does not exist. The set of reasons does not overlap perfectly with the set of factors. The unpalatable consequences alluded to by the critics of ST are thereby avoided.

The robustness of the consequential critique

By arguing that the assumption underlying the consequential critique conflates permission with occurrence, Anderson has, in effect, called for a replacement assumption that respects this distinction:

Anderson’s Assumption: God’s permission of some instance of suffering E is justified only if the permission of E is itself logically required for the attaining of an outweighing benefit.

My argument is that this assumption is itself sufficient for a version of the consequential critique to go through. I defend this is in three stages. First, by developing an analogy that distinguishes between the relevant forms of divine permission. Second, by using this distinction to develop an argument in favour of the consequential critique. And third, by fending off a variety of objections.

I start with an analogy:

Playground Bullying: You are a parent, watching your child play with his/her peers in the playground. You see several children approach your child, take one of his toys away, and start teasing him. Although you could intervene, retrieve your child’s toy, and call off the bullies, you decide it is best not to. Your child looks askance at you and wonders why you are doing nothing.

It is an obvious point, but I will say it anyway lest there be any doubt: in this analogy, you as the parent are filling the God-like role and your child is filling the role of ordinary human beings.27 You have decided not to intervene to prevent your child’s suffering, and your child is left wondering why. Applying Anderson’s assumption, but dropping the logical necessity requirement, we say that your decision to permit this instance of suffering is justified only if there are reasons demonstrating that non-intervention is necessary (to the best of your knowledge) in order to achieve some outweighing benefit.

What kinds of reasons for permission could supply the necessary justification? Let me suggest that there are two, general, types:

Empowering Permission: You do not intervene in the hope that your child will resolve the situation for themselves. In other words, you hope that by stepping back, your child will resolve the situation themselves by standing up to the aggressors. In this case, although you do not prevent the suffering yourself, you are hoping it will not actually occur because of the actions of your child.

Non-Preventive Permission: You do not intervene in the hope that your child will learn some important moral lesson about the world and the kinds of people within it through this experience. Although this may require some short-term suffering, you feel the long-term lesson will more than outweigh it.

Both kinds of reasons for non-intervention could be justifiable, but they are very different in nature. If the first reason is operative, then it is better that the suffering (E) does not occur but only if your child prevents it; if the second reason is operative, then it is better that the suffering occurs. From the perspective of your child, who is staring askance at you from across the playground, uncertainty as to which reason applies would seem to undermine their knowledge about what they ought to do.

The application of this analogy to the God-humanity relationship should be clear. God may have justifying reasons for non-intervention, but those reasons could come in either of the two forms just outlined. God might not intervene because, although it is better ATC that suffering does not occur, it is necessary for some outweighing benefit that we prevent it rather than him (empowering permission). Or, God might not intervene because there are outweighing benefits that actually make it better for the suffering to occur (non-preventive permission).

The fact that the God-justifying reasons for permission could take either of these general forms has serious consequences. Consider once more your drowning friend. God is not intervening to prevent his drowning, so you must make the decision to help him or not. The problem is that God’s non-intervention could be driven by reasons of empowering permission, i.e. because it is best for your friend not to drown but only if you are the one to save him. Or it could be driven by reasons of non-preventive permission, i.e. because it is actually for the best that your friend drown. You cannot know which. This creates a situation of extreme moral ignorance, as the following argument makes clear:
  1. (14)

    There is an observed fact: event E is occurring, which involves a case of seemingly gratuitous drowning, and God is not intervening to prevent it.

  2. (15)

    This could be a case of empowering permission (EP) or non-preventive permission (NPP).

  3. (16)

    If it is a case of EP, you ought to intervene and prevent the drowning.

  4. (17)

    If it is a case of NPP, you ought not to intervene and prevent the drowning.

  5. (18)

    If ST is true, you have no way of knowing whether EP is more likely than NPP.

  6. (19)

    Therefore, if ST is true, you cannot know whether you ought to intervene or not.

Premise (14) of this argument is merely a description of the scenario from Maitzen’s thought experiment. Premise (15) sets out the possible forms of permission. I believe it is correct that either reason for non-intervention could exist, though I address objections to this possibility below. Premises (16) and (17) flow from the definitions of EP and NPP that I gave earlier. Premise (18) looks to capture what is entailed by the Bergmannian form of ST. As you recall from “The nature of skeptical theism” section, it is not just that ST alters the probabilities we might assign to such reasons, it is that it completely undercuts our ability to make any such probability assignments. Thus, the conditional claim in premise (18), which says we cannot assign a greater probability to EP than to NPP, is justified under Bergmannian ST. Still, this putative justification is controversial and will be the focus of considerable discussion below. (19) follows logically from the preceding premises. The result is moral ignorance.

The argument could develop in a number of ways from here. Perhaps the simplest and most persuasive development would be to use (19) as the basis for a reductio of ST: since we know we ought to intervene, ST must be false.28 There is thus an incompatibility between ordinary commonsensical moral knowledge and ST. Alternatively, one could stress the decisional paralysis that is created by the lack of knowledge in cases like this.

One consideration weighing against the paralysis argument is that in situations of moral uncertainty (i.e. where moral knowledge is absent), the assumption that we are dealing with a case of EP might be thought to dominate the assumption that we are dealing with a case of NPP. In other words, even though there is no way to tell which of the two possible forms of permission applies in the instant case, we minimise the risk of moral error by acting as if it is a case of EP. If your friend is drowning, you should assume you are obligated to rescue him because if it really is morally better for your friend to die (i.e. if NPP holds) then you can expect God to intervene and prevent you from saving him. Either way, you minimise the risk of making a moral error by assuming you ought to save him.29

There are two points to be made in response to this counter-argument. First, one could, I believe, plausibly argue that even saying this much assumes that we have a kind of moral knowledge that is sufficient for determining where the moral risks lie in cases like this. As the contamination argument sketched below suggests, it is not clear that proponents of ST are entitled to even this much. Second, even if it does stave off decisional-paralysis, problems about moral knowledge remain. One might think that if one intervenes and God does nothing, one thereby acquires a form of knowledge that one can apply to similar future dilemmas. But this is not true: if ST is true, the mere fact that God did not intervene previously says nothing about what holds true in future cases, especially given the fact that there are also historical cases of seemingly gratuitous suffering which God did not prevent. One must, consequently, always act from a state of profound moral uncertainty.30 This, I would submit, is a sufficiently unpalatable consequence.31

All this has to do with the possible developments from (19). What about the premises preceding that conclusion? Anderson might argue that there is something wrong with premise (15); that one of the two possibilities is not really “live” in the instant case. For Anderson, that would presumably have to mean that EP was the live possibility since that is the form of permission that holds it is better, ATC, for E not to occur. The problem is that none of the arguments he offers in support of his replacement assumption are sufficient to demonstrate that NPP is not also a live possibility. For instance, his claim that God, as external judge or sovereign, may face certain costs associated with his intervention into E that we cannot see, and that this may justify his non-intervention, supports the possibility non-preventive permission. It is only if the costs of intervention outweigh the benefits of intervention that he could justifiably refrain from intervening (given his omnibenevolence). This could just as easily be because it is better for E to occur than because it better that it does not to occur. The same holds true when weighing the benefits of a principles-based approach to intervention against the benefits of a case-by-case approach.

The free will argument supplies some support for EP since one of the goods associated with free will is that it empowers genuine moral agency, but it does not do so to the exclusion of NPP. First, the free will defence of EP could not apply to all cases of seemingly gratuitous evil. NPP would remain a possibility in cases of natural evil, which includes cases like those of the drowning friend. Second, Anderson wavers here, saying that “respecting the freedom of creatures might not justify God’s permission of just any evil (if the harm bad enough, it would seem better for God to will...)”.32 But if this is accepted, we are constantly forced into asking questions about the possibility of NPP in cases of free agency. If God can intervene to prevent exercises of free will from going too far, it follows that he implicitly permits all historically evil exercises of free will. Take a case like that of David Rothenberg,33 in which a father, using his free will, set his 5-year-old son on fire. The suffering inflicted in that case was appalling. If we concede that God could intervene when the suffering is bad enough, then why did not he do so in that case? The burden would seem to be on Anderson to suggest to us why God permits the exercise of free agency in historical cases like this. He can do so by pointing to epistemic possibilities, as presumably is his wont under ST. But then the question arises: what sorts of epistemic possibilities could justify permitting such appalling exercises of free agency? NPP seems the only real candidate since EP did not apply (the seemingly evil event occurred). So, once again, following Anderson’s argument, the possibility of NPP has to be considered. If that is right, premise (15) would seem sound.

A second line of attack might be to argue that premise (18), contrary to what I said above, is not supported by ST principles. The argument would be that there are other sorts of evidence that can be relied upon to work out the contents of our moral obligations and thereby avoid moral ignorance. Biblical revelations of divine commands are sometimes mooted in defence of this.34 But this is not a particularly promising argument. As Maitzen points out,35 biblical commands are notoriously difficult to identify, potentially controversial and, in any event, frequently ignored. Further, this source of evidence would not be open to a defender of Anderson’s view. The Bible, on whichever view of divine inspiration you might like to take, has a tight connection to God’s reasons.

Are there any other sources of evidence that could disrupt the implication in premise (18)? Anderson says:

If we reject the assumption, then for all we know at least, the set of potentially God-justifying reasons for permitting evil is not exhausted by the set of factors contributing to the ATC value of those evils. So for all we know the first set is larger (much larger?) than the second. Were [critics of ST] to convince us that the goods/evils/entailments we are aware of give us reliable grasp on the contents of the second set, we would not thereby be forced to admit that they give us good grasp on the contents of the first set. That is, ST does not imply our inability to make certain judgments of ATC value.36

The upshot of this seems to be the following:

Anderson’s Claim: Since the set of God-justifying reasons for permitting evils does not form a perfect union with the set of factors determining the ATC value of an event or state of affairs, we can have reliable epistemic access to the contents of the latter without requiring reliable epistemic access to the contents of the former.

This claim is blocked by the following argument, which invokes a contamination principle of evidence:
  1. (20)

    If the set of God-justifying reasons forms any sort of overlap with the set of factors determining the ATC value of an event or state of affairs then, by the contamination principle, an inability to reliably identify members of the former set leads to an inability to reliably identify members of the latter.

  2. (21)

    The set of God-justifying reasons forms some sort of overlap with the set of factors determining the ATC value of an event or state of affairs.

  3. (22)

    ST implies a complete inability to reliably identify God-justifying reasons.

  4. (23)

    Therefore, if ST is true, we lack the ability to reliably identify the factors determining the ATC value of an event or state of affairs.

Nothing in the premises of the argument denies or contradicts anything that Anderson says. Take premise (21) for example. In his own comments on the matter, Anderson only says that the set of factors determining ATC value “does not exhaust” the set of God-justifying reasons. This implies that he thinks there is some overlap between the two sets. Further, even if others wish to deny this, there are good reasons for thinking that it must be true. If God is the omnipotent creator and sustainer of the world, then there must be at least some occasions on which the shape and content of natural events are determined by his reasons for doing or forbearing from some action. Indeed, ST could not undercut the evidential argument if this were not true: ST assumes that natural evils could be explicable by unfathomable God-justifying reasons for action. Premise (22) also seems wholly consistent with the ST-position, particularly in its Bergmannian form, which is the form adopted by Anderson. Premise (20) is the only really troublesome one here. Certainly, Anderson wishes to deny that uncertainty about the contents of the set of God-justifying reasons has this contaminating effect. It is incumbent on me to show that this is denial is wrong.

I begin by clarifying the nature of the contamination principle invoked by that premise. In legal contexts, the principle covers any situation in which the reliability of some body of evidence is called into question by some contaminating fact. For example, suppose DNA samples were taken from the scene of a crime but, while in transit or while at the lab, they were mixed in with other DNA samples so that, when it came time to see whether they matched with the DNA of a suspect, there was no way of knowing whether the sample being tested came from the crime scene or not. In this scenario, the mixing of the samples is the contaminating fact. Its effect is devastating. Even if the sample that is ultimately tested and matched with the DNA from the suspect is the sample taken from the crime scene, the contamination casts sufficient doubt on it. In other words, even if the results of the test are true, the fact of contamination undermines our confidence in them. We cannot rely on this evidence when trying to prove guilt.

Another analogy might be helpful here. In a somewhat controversial piece, Stephen Law37 has argued that Gospel narratives may be contaminated in such a way that their overall reliability is cast into doubt. I have no particular interest in supporting Law’s claim about the Gospels here, but the hypothetical case he uses to illustrate his version of contamination principle is instructive. He asks us to imagine two interlocutors—Ted and Sarah—who tell us first that a man named Bert paid them a visit the previous evening. They then add to this otherwise mundane story the detail that Bert flew around the room by flapping his arms before dying, coming back to life and turning their sofa into a donkey. Law contends that although the first part of their story is credible, the second part is not. And since our only evidence for Bert’s visit is from Ted and Sarah’s testimony, the incredible nature of the second part of their story contaminates the more reasonable first part. If the sources are the same, once we have reason to believe that one part of the story was produced by a false-belief generating mechanism, we have reason to think that the other parts were too. So once again, the reliability of a body of evidence (Ted and Sarah’s testimony) is undermined by a contaminating fact. As Law notes, once contamination of this sort takes place, we cannot simply adopt a “bracketing strategy” towards the evidence. In other words, we cannot simply bracket the incredible and dubious parts of the narrative off from the more mundane and credible parts. The only way to overcome the problem would be to find some independent source of evidence that would vouch for the credible parts but not the incredible ones, or that could be sufficiently trustworthy to overcome our doubts about the incredible parts.

In essence then, the contamination principle stipulates that whenever a collective body evidence, which is verified or vouched for on the basis of a common source, is called into doubt by a contaminating fact, there is reason to doubt our judgments about that entire collective, unless some independent source of verification can be found.

The application to the argument I am making should be obvious. The claim is that our knowledge of the factors that determine the ATC value of events and states of affairs is contaminated by the fact that those factors could be coextensive with God-justifying reasons for permitting evil to occur. Consider our potential sources of knowledge about ATC value: our faculties of reason and intuition, and our observations of the world around us. The contents of each one those sources could be at least partly determined by God’s justifying reasons for doing or forbearing from some action: the external world and our minds are ultimately the product of his creative act, and our faculties of reason and intuition can occasionally be affected by his actions and interventions. But under ST, we have no way to reliably assess or access the justifying reasons that lie behind his acts, so when he is acting we have no way of knowing why our mental faculties or the external world have the particular shape and content that they do.

This has a profoundly contaminating effect. Whenever we try to assess the ATC value of a state of affairs, we either consult our reason or intuition, or we make observations of the external world. In doing so we are aware that God’s reasons could be shaping the contents of each of those sources. Consequently, even if our reason or intuition or observations seem to yield a particular conclusion about ATC value, there is the possibility that these conclusions emanate from unfathomable God-justifying reasons. The end result is that we cannot reliably make inferences from those sources to conclusions about the ATC value of the state of affairs. To do so would be like relying on the results of DNA-matching test when you know the samples might be mixed; or like trusting in the truth of someone’s testimony when it contains a mixture of credible and incredible things. Indeed, it is even worse than this. In both those cases, there might be tractable probabilities we could use to alter our credence in the particular belief; in the case of God-justifying reasons, at least if proponents of ST are correct, such tractable probabilities do not exist.

The argument can also be understood in set theoretical terms. The dilemma faced by anyone trying to assess the ATC value of a state of affairs (say the prospective death of a fawn in a forest fire) is the following. They are confronted by two possible sets of information. The set of God-justifying reasons, which may or may not require the death of the fawn in order to secure some outweighing benefit, and the set of factors typically thought to determine the ATC value of the event, which might seem to imply that the fawn’s death is an evil that ought to be prevented. These two sets could overlap in this particular case: the set of factors could be expressions of God’s will and hence products of his reasons. If it is possible for the sets overlap in this manner, any judgments the person might like to make about the ATC value of the fawn’s prospective death will be unreliable. The set of fathomable factors will be irremediably contaminated by the set of unfathomable reasons. To use a spatial metaphor, the problem is that the person has no way of knowing which location within of the set of factors determining the ATC value of the state of affairs is guiding their judgment: it could be that they are drawing from the part that is independent of God’s reasons, or it could be that they are drawing from the overlapping part. But because of their uncertainty, their ability to make ATC judgments about the value of the fawn’s death are undermined. Once again, unpalatable consequences seem to follow. This is true no matter how small or large the union of the two sets is. As long as they form any sort of overlap at all, the contamination will arise.

This undercuts any purported attempt to sidestep premise (18). The result is that the consequential critique remains robust, even in light of Anderson’s replacement assumption.

Foundations and conclusions

One final observation is in order. Although the preceding arguments demonstrate some crucial flaws in Anderson’s attempted rescue of ST from the consequential critique, the problems do not end there. There is an additional concern with Anderson’s argument. This concern relates to the foundations of ST, not its consequences and reveals an important connection between the consequences and foundations of the doctrine.

In a recent article,38 Dougherty has argued that philosophers have tended to ignore the foundational arguments for ST. The foundational arguments try to defend the claim that the condition of reasonable epistemic access (CORNEA) is not met in the case of God-justifying reasons for permitting seemingly gratuitously evil events. The most common foundational argument is based on the Parent Analogy. A parent may have to cause their child to undergo suffering in order to obtain some outweighing benefit. For example, when they bring the child for a painful round of vaccinations, the child will lack access to the reasons that justify their suffering because of the epistemic disparity between them and the parent. The parent can understand and appreciate things that the child, given their cognitive immaturity, cannot. This disparity means that CORNEA is not met in cases like this. But, so the analogy goes, the cognitive disparity between ourselves and God is similar to that between the parent and child so that when we are confronted with or forced to undergo seemingly gratuitous suffering we cannot reasonably expect to have access to God-justifying reasons for permitting this.

The debate has gone back and forth on the merits of this analogy.39 Rowe, Russell and Dougherty argue that the analogy fails and propose a counter-analogy: the Loving Parent Analogy. According to this, whenever a loving parent has the means to communicate to and reassure their child that the suffering they must undergo will work out for the best, they will do so. God’s relationship to us is much more like that of the loving parent (with the means to communicate) to their child and so we should expect some guidance and reassurance about the reasons that might justify seemingly gratuitous suffering. Particularly given, as Dougherty points out, that the inability to understand why we are suffering typically amplifies our degree of suffering. Wykstra has responded with a refined version of the analogy which argues that God’s characteristics—maximal intelligence, ability and goodness—are likely to make the world morally obscure to us. Dougherty rejects this on the grounds that those characteristics are actually more likely to make the world morally transparent, perhaps because they make it more likely that God would have created us with greater cognitive abilities in the first place.

The point here is not to judge who has the better of this argument—though my sympathies are probably transparent—the point is to highlight how Anderson’s defence of his position can lead to a similar set of concerns. In effect, Anderson has offered us a refined version of the skeptical thesis, namely: CORNEA is not met in the case of God-justifying reasons for permitting seemingly gratuitous evils. He supports this refined thesis with some interesting analogies, liking God to a judge or policy-maker and us to ordinary citizens. He claims that we should not expect to know what the justifying reasons for a judge’s or policy-maker’s decisions are; that we should adopt a stance of epistemic humility toward them. Hence, for similar reasons, we should not expect to know what God’s reasons are.

The problem is that these analogies seem vulnerable to the same kinds objections as the parent analogy. Indeed, in many ways, they are much worse than the parent analogy. Under most contemporary versions of political morality, the moral legitimacy of governmental authorities flows directly from their ability to respect the moral autonomy and liberty of their citizens.40 One of the key markers of such respect is a government’s ability to provide clear conduct guidelines to their citizens, and to justify any coercive or autonomy–undermining decisions to them in terms of moral reasons. In other words, a government’s moral legitimacy is a direct function of their ability to explain their justifying reasons to their citizenry. Now, certainly, this may be a difficult task to perform. The reasons might be complex, based on the weighting of many different factors, and some people may lack the patience or ability to understand. But in an ideal polity the reasons will be provided and we, as citizens, should expect to have them. Which is to say, in an ideal polity CORNEA should be met.

This applies a fortiori to God as judge and policy-maker. Since he created us to be morally autonomous creatures, and since he has the power to shape our cognitive faculties, we should expect him to be able to explain his justifying reasons to us. Particularly when those reasons permit our suffering, and when uncertainty about those reasons amplifies our suffering or leaves us in doubt as to what we should do. While human governments can be expected to fail to provide us with the moral reasons we are owed, because of time or other resource constraints, God should not be expected to fail since those constraints do not apply to him. Thus, if God is really akin to the judge and policy-maker, and we are really akin to the ordinary citizens, CORNEA should be met. He should communicate his reasons for failing to intervene to prevent suffering.

In conclusion then, even if Anderson’s riposte to the consequential critique were successful—which it is not—his foundational defence of his claim is flawed. It provides no strong reason for thinking that ST is true. Quite the contrary in fact: the analogies he uses give us reason to think that ST is false.


  1. 1.

    Rowe (1979). Perhaps the best single work on Rowe’s argument, in its several guises, is Trakakis (2007).

  2. 2.

    There may be some versions that do not use this inference, but many do. See: Howard-Snyder (2009) at fn 1.

  3. 3.

    This characterisation is drawn primarily from Bergmann (2001) with his three core principles of ST, but see also Alston (1996) and his six principles of cognitive limitation.

  4. 4.

    Piper (2008) and Hasker (2010).

  5. 5.

    Maitzen (2013), Sehon (2010) and Almeida and Oppy (2003).

  6. 6.

    Sehon (2010) and Hasker (2010).

  7. 7.

    Wielenberg (2010).

  8. 8.

    Lovering (2009) argues, more precisely, that the problem is attributable to the fact that each of these things relies on some assumptions about what God would do (or allow), at least if one is a theist. (On atheistic views, moral reasoning should be possible without second-guessing what God would allow—this simply reveals the obvious point that the consequential problems of ST are very much an intra-theist, indeed intra-skeptical theist, problem).

  9. 9.

    Hasker (2010).

  10. 10.

    See Anderson (2012). Of course, he is not the only one to address this kind of argument. For overviews of the literature here, see McBrayer (2010) and Dougherty (2012).

  11. 11.

    Three are most typically presented, with the fourth simply spelling out the implications of the doctrine. See Bergmann (2001, 2009).

  12. 12.

    Bergmann (2001, 2009).

  13. 13.

    Alston (1996).

  14. 14.

    Bergmann (2009), p. 377.

  15. 15.

    Bergmann makes this point in response to an argument by Swinburne. See Hasker (2010) for a discussion.

  16. 16.

    See Bass (2011) for a full statement of this argument, with appropriate mathematical demonstrations. See also Hasker (2010) for a more general discussion.

  17. 17.

    Lovering (2009).

  18. 18.

    Wielenberg (2010). Piper (2008) makes a similar point.

  19. 19.

    Almeida and Oppy (2003) appear to have been the first to proffer this argument in a systematic form. Sehon (n 5) also offers a version of it. Maitzen (2013) has possibly the most up-to-date version.

  20. 20.

    It is possible that God’s obligations are consequentialist while ours are deontological, but I suspect the burden of proof is on the skeptical theist to justify this distinction. Our obligations to animals might be consequentialist while our obligations to our fellow humans might be deontological, but that would seem to be because animals and humans are fundamentally different kinds of moral agents. Animals might be capable of conscious suffering but they are not persons so typical Kantian respect arguments seem not to apply. Humans and God are both persons so, prima facie, one would expect their obligations to have the same character, be it deontological or consequentialist. And since ST assumes they are consequentialist, it is up to the proponent of ST to justify the difference. I am indebted to Steve Maitzen for making this point to me.

  21. 21.

    In addition to the previously cited sources, see also Maitzen (2009) for an argument which pushes the consequential problem in this direction (though it does not style itself as a response to ST, Maitzen’s article incorporates similar ideas).

  22. 22.

    Anderson (2012), p. 31.

  23. 23.

    It is difficult to formulate an adequate definition of omnipotence, but defining it in terms of “maximal power” seems to be the broadest and least contentious way to do it. See Hoffman and Rosenkrantz (2012).

  24. 24.

    As documented by Anderson (2012), both Maitzen and Hasker offer versions of this argument when defending the consequential critique.

  25. 25.

    These two premises are slight reworkings of those found on p. 33 of Anderson (2012). He formulates them in response to Maitzen’s argument.

  26. 26.

    The Open Theist will reject this claim, though there are several versions of that doctrine, and some might be consistent with a revised formulation, e.g. if it is accepted that God can estimate the probabilities of possible branching futures, inclusion of the phrase “must be better or at least highly likely to be better” might work. This depends on the notion that, even if the future is open, God must bet on the course of action most likely to be (ATC) better. Still, what counts as a “good bet” is a problematic notion, and one that will not be explored here (I am indebted to an anonymous reviewer for pointing this out to me).

  27. 27.

    Given that proponents of ST often employ a parent-child analogy, this seems acceptable.

  28. 28.

    Somewhat akin to the line taken by Maitzen (2009).

  29. 29.

    I am indebted to an anonymous reviewer for bringing this objection to my attention.

  30. 30.

    Recall that, following the Bergmannian claim, it is not just that you will have reduced confidence in your moral convictions (e.g a reduction in subjective probability from 0.8 to 0.01); you will have moral convictions with inscrutable probabilities.

  31. 31.

    Indeed, it probably even undermines EP since empowerment can be taken to require some form of moral growth and learning (as in the soul-making theodicy). This would be a serious problem for Anderson since EP is the form of permission that holds that E’s occurrence is not actually necessary for an outweighing good.

  32. 32.

    Anderson (2012), p. 35.

  33. 33.

    Originally introduced by Maitzen, but discussed by Anderson.

  34. 34.

    Bergmann and Rea (2005), in particular p. 244.

  35. 35.

    Maitzen (2007).

  36. 36.

    Anderson (2012), p. 37.

  37. 37.

    Law (2011).

  38. 38.

    Dougherty (2012).

  39. 39.

    This characterisation of the debate is taken from Dougherty (2012).

  40. 40.

    See, for example, Gaus (2003), Chap. 8 and (2010), Chap. 1 discussing the problems of moral authority; Estlund (2008) and Machin (2009), discussing Estlund’s argument. Many other authors make similar arguments.



The author would like to thank Stephen Maitzen and an anonymous reviewer for helpful comments on earlier drafts of this paper.


  1. Almeida, M., & Oppy, G. (2003). Sceptical theism and evidential arguments from evil. Australasian Journal of Philosophy, 81, 496–516.CrossRefGoogle Scholar
  2. Alston, W. P. (1996). The inductive argument from evil and the human cognitive condition. In D. Howard-Snyder (Ed.), The evidential argument from evil. Bloomington: Indiana University Press.Google Scholar
  3. Anderson, D. J. (2012). Skeptical theism and value judgments. International Journal for the Philosophy of Religion, 72, 27–39.CrossRefGoogle Scholar
  4. Bass, R. (2011). Many inscrutable evils. Ars Dispuntandi, 11, 118–132.Google Scholar
  5. Bergmann, M. (2001). Skeptical theism and Rowe’s new evidential argument from evil. Nous, 35, 228–296.CrossRefGoogle Scholar
  6. Bergmann, M. (2009). Skeptical theism and the problem of evil. In T. P. Thomas & M. Rea (Eds.), The oxford handbook of philosophical theology. Oxford: OUP.Google Scholar
  7. Bergmann, M., & Rea, M. (2005). In defence of skeptical theism: A reply to Almeida and Oppy. Australasian Journal of Philosophy, 83, 241–251.CrossRefGoogle Scholar
  8. Dougherty, T. (2012). Recent work on the problem of evil. Analysis, 71, 560–573.Google Scholar
  9. Dougherty, T. (2012). Reconsidering the parent analogy: Unfinished business for skeptical theists. International Journal for the Philosophy of Religion, 72, 17–25.Google Scholar
  10. Estlund, D. (2008). Democratic authority. Princeton: Princeton University Press.Google Scholar
  11. Gaus, G. (2003). Contemporary theories of liberalism. London: Sage.Google Scholar
  12. Gaus, G. (2010). The order of public reason. Cambridge: Cambridge University Press.CrossRefGoogle Scholar
  13. Hasker, W. (2010). All too skeptical theism. International Journal for the Philosophy of Religion, 67, 15–29.CrossRefGoogle Scholar
  14. Hoffman, J., & Rosenkrantz, G. (2012). Ominpotence. In Zalta, E. N. (Ed.), Stanford encyclopedia of philosophy.
  15. Howard-Snyder, D. (2009). Epistemic humility, arguments from evil and moral skepticism. In J. Kvanvig (Ed.), Oxford studies in the philosophy of religion (Vol. 2). Oxford: OUP.Google Scholar
  16. Law, S. (2011). Evidence, miracles and the existence of Jesus. Faith and Philosophy, 28, 129–151.CrossRefGoogle Scholar
  17. Lovering, R. (2009). On what God would do. International Journal for the Philosophy of Religion, 66, 87–104.CrossRefGoogle Scholar
  18. Machin, D. (2009). The irrelevance of democracy to the public justification of political authority. Res Publica, 15, 103–120.CrossRefGoogle Scholar
  19. Maitzen, S. (2007). Skeptical theism and God’s commands. SOPHIA, 46, 235–241.CrossRefGoogle Scholar
  20. Maitzen, S. (2009). Ordinary morality implies atheism. European Journal for the Philosophy of Religion, 1, 107–126.Google Scholar
  21. Maitzen, S. (2013). The moral skepticism objection to skeptical theism. In J. McBrayer & D. Howard-Snyder (Eds.), A companion to the problem of evil. Oxford: Wiley-Blackwell.Google Scholar
  22. McBrayer, J. (2010). Sceptical theism. Philosophy Compass, 5, 611–623.CrossRefGoogle Scholar
  23. Piper, M. (2008). Why theists cannot accept skeptical theism. SOPHIA, 47, 129–148.CrossRefGoogle Scholar
  24. Rowe, W. (1979). The problem of evil and some varieties of atheism. American Philosophical Quarterly, 16(4), 335–341.Google Scholar
  25. Sehon, S. (2010). The problem of evil: Skeptical theism leads to moral paralysis. International Journal for the Philosophy of Religion, 67, 67–80.CrossRefGoogle Scholar
  26. Trakakis, N. (2007). The God beyond belief: In defence of William Rowe’s evidential argument from evil. Dordrecht: Springer.Google Scholar
  27. Wielenberg, E. (2010). Sceptical theism and divine lies. Religious Studies, 46(4), 509–523.Google Scholar

Copyright information

© Springer Science+Business Media Dordrecht 2013

Authors and Affiliations

  1. 1.School of LawKeele UniversityStaffordshireUK

Personalised recommendations