Skeptical Theism and Its Costs

The response introduced at the previous rung is now the most widely held theistic answer to evidential arguments from evil, and it is worth being clear at the outset about what makes it attractive. It requires no theodicy. It does not have to say what good justifies the fawn's suffering, or claim that any particular account is correct. It only has to say that we are not in a position to tell.

The question this rung asks is what else that concession commits its holder to. The critics' answer is that the scepticism does not stay where it was put, and that a theist who accepts it loses the ability to reason about morality, about the significance of religious experience, and about whether appearances of any kind are to be trusted. The theist's answer is that this misreads the scepticism as far broader than it is. Sorting out who is right requires stating the thesis precisely, which is what its principal defender did.

The thesis, stated

The name is unfortunate and regularly misleads. Skeptical theism is not scepticism about God; it is theism combined with scepticism about a specific and narrow matter, namely our ability to survey the realm of value well enough to judge what an omniscient being would have reason to permit.

The form of each thesis matters. None of them says that our beliefs about goods and evils are false, or that there are goods beyond our ken. Each says only that we have no good reason to think our sample is representative, which is weaker and harder to attack. A sample can fail to be representative without being wrong about what it contains.

The application to Rowe's argument is direct. The inference from "we can identify no justifying good here" to "there is probably none" requires that the goods we can identify be a decent sample of the goods there are. If the theses hold, that requirement is not met, and the noseeum inference is blocked at exactly the point CORNEA identified.

Why it is not simply ad hoc

The obvious first reaction is that the theses are invented to escape an argument, and would be believed by nobody who was not already committed to the conclusion. The skeptical theist's response is that the theses are independently plausible, and this is worth taking seriously before the costs are counted.

Consider what a representative grasp of the space of goods would require. It would require having surveyed the kinds of value there can be, having grasped the relations of dependence among them, and being able to tell, for an arbitrary state of affairs, what it makes possible and forecloses across all of history. Nobody thinks we have that. Our moral concepts have visibly expanded over time, which is evidence that the sample available at any moment is incomplete, and the history of ethics is partly a history of discovering that goods we had not considered were at stake.

Consider also the disparity the hypothesis under discussion posits. If theism is true, the difference in cognitive position between a human and God is not the difference between a novice and an expert. The parent analogy of the previous rung understates it, since a child and a parent are the same kind of thing. On the theistic hypothesis it is closer to the difference between a creature that can hold four things in mind at once and one that grasps every truth there is.

So the theses are not obviously unmotivated. The dispute is not about whether we are limited, which everyone grants, but about whether the specific limitation claimed can be held without the rest of one's beliefs coming apart.

What the theses are aimed at

It is worth being exact about which part of the argument from evil the theses attack, because a good deal of confused criticism comes from missing this.

Rowe's argument had two premises. The first was factual: there exist instances of intense suffering that an omnipotent being could have prevented without losing a greater good or permitting an evil equally bad. The second was normative: a wholly good being would prevent such suffering if it could. The skeptical theist accepts the second premise, and in most cases accepts it enthusiastically, since it expresses what perfect goodness comes to. The attack falls entirely on the support for the first.

That support was the noseeum inference. We survey the goods we can think of, find none that would justify permitting the fawn's death, and conclude that there is probably none. The theses say the survey is not the kind of survey from which such a conclusion follows, because there is no reason to think what we surveyed is a fair sample of what there is.

Two consequences follow, and the second is often forgotten.

The first is that skeptical theism does not require the theist to hold that there are goods beyond our ken. It requires only agnosticism about whether there are. That is a lower bar, and it is deliberately set low.

The second is that the position is purely defensive. It gives no reason to believe that God exists, and it cannot be converted into one. A skeptical theist who has successfully blocked the evidential argument has restored the situation to what it was before that argument was made, and the question of whether theism is supported must then be settled by the arguments of the earlier rungs. Presented as though it were a positive answer to the problem of evil, the position claims far more than it can deliver.

How strong a scepticism is needed

The theses are not a single setting but a dial, and much of the dispute is about where it has to be turned.

Set weakly, they say only that our grasp of the space of goods is incomplete in ways we cannot survey from the inside. Almost nobody would deny this, and almost nobody thinks it is enough. A merely incomplete sample can still support a probabilistic conclusion, in the way that a partial census still tells you something about a population, so a weak reading leaves the noseeum inference weakened rather than blocked.

Set strongly, they say that we have no idea at all what proportion of the relevant goods we have surveyed, and that no probability at all can be assigned. This does block the inference, and it is roughly the setting the position needs.

The costs, however, scale with the setting. Every objection in the sections below draws its force from the strong reading, and each of them would be answerable if the weak reading were enough. That is the structural difficulty at the centre of the debate: the strength required to defeat the argument from evil is the same strength that spreads into moral reasoning, into religious epistemology, and into the reliability of appearances. Showing that some intermediate setting both blocks the inference and stays contained would settle the matter, and no such setting has been produced.

The first cost: moral reasoning

The objection is easiest to see in a case. A child is drowning and I can save it at no risk. Ordinarily this settles the matter. But if I am a skeptical theist, I hold that I have no grip on whether some good beyond my ken depends on the child's dying, and that reasoning of exactly this kind is unreliable when applied to whether suffering is permitted. It seems that I must be similarly uncertain about whether to intervene.

The response has several parts, and they are not all equally strong.

The first is that obligations are indexed to what the agent knows, not to what is true. I am obliged to act on the balance of reasons available to me, and the possibility of unknown countervailing goods does not change what is available to me. This is a real principle, and it is why we do not blame people for unforeseeable consequences.

The critic's reply is that this saves my blamelessness rather than my reasoning. The question was not whether I would be culpable if I got it wrong; it was whether I have any reason to think saving the child is the better outcome. If the answer is that I do not, then the fact that I am excused for acting on appearances is cold comfort, and it certainly does not restore my confidence that the world is better for my having acted.

The second response is that the objection assumes a consequentialist account of obligation, and that most theists do not hold one. On a duty-based account, I am required to save the child because that is what the duty of rescue requires, and the balance of goods across all time is not my business.

This is stronger, but it does not fully escape. Duties are not free-floating; they are usually held to track goods in some way, and a scepticism deep enough to sever any connection between our grasp of goods and the content of our duties leaves the duties without support. And in the specific case at issue the theist has already accepted that God's not intervening is justified by goods beyond our ken. If that consideration can justify a perfectly good being in standing by, the theist owes an account of why the same consideration is unavailable to me.

The third response is the one Bergmann and Michael Rea gave in their 2005 reply in the same journal: that God and human agents are differently placed, since God has no duty of rescue of the kind humans have and stands in a different relation to the whole course of events, so what justifies divine permission does not transfer. This is the most promising line, and it turns the debate into one about the structure of moral obligation rather than about evil.

There is a fourth response, of a different kind, and it deserves separate treatment because it is the one most theists actually rely on.

It runs like this. That I ought to save the drowning child is more certain than any premise in any philosophical argument, including the arguments that generate the paralysis. So if some position appears to imply that I have no reason to save the child, what that shows is that the position has been misapplied, not that I should stand and watch. The fixed point is the obligation; the theoretical claim is what gives way.

This is the pattern of reasoning that was used at rung 14 against Rowe's second premise, run now in the other direction, and it has the same strength and the same weakness. Its strength is that it is honest about the order of certainty: nobody arrives at the duty of rescue by deduction from a theory of value, and a theory that overturns it is more likely to be wrong than the duty is. Its weakness is that it saves the practice without saying which applications of the scepticism are legitimate. The theist is holding that the theses defeat one inference about what a good agent has reason to do and leave another untouched, and pointing at the certainty of the second is not yet a reason why the first differs from it. Until that reason is given, the containment is being asserted rather than shown, which is the recurring shape of every reply in this literature.

The second cost: containment

The clearest instances concern things the theist positively wants to affirm.

Consider divine promises. Many theologies hold that God has made undertakings about the future. Assessing whether an undertaking will be kept requires judging what a perfectly good being would do, which is exactly the judgement the theses say we cannot make. If we cannot tell that a good being would prevent a fawn's suffering absent a defeating reason, we cannot tell that a good being would keep a promise absent a defeating reason either, since a good beyond our ken might depend on its being broken.

Consider the interpretation of religious experience, which rung 10 relied on. Taking an experience as of God requires judging that the character of the experience is the sort of thing God would produce, and that judgement draws on our sense of what God is like and what God would want. The theses undercut the confidence with which such judgements can be made.

Consider, finally, the afterlife, which the theist often invokes as the place where apparent injustices are set right. That invocation is itself a claim about what a good being would arrange, and it is not more accessible to us than the claims the theses deny us.

The skeptical theist's containment strategy is to insist that the theses concern our knowledge of the range of goods and of entailment relations among them, not our knowledge of God's character, which is held to be available through revelation and through the concept of perfection. Whether that line can be drawn where it needs to be drawn is the crux, and critics argue that reasoning from God's character to God's actions requires exactly the value-theoretic knowledge that has been given up.

The third cost: appearances

This is the objection that most alarms theists, because it reaches beyond ethics into the theory of knowledge, and because the theistic tradition has often used the goodness of God as a guarantee of the reliability of our faculties.

The argument is short. We normally suppose we are not systematically deceived. On theism, one important ground for supposing this is that a good God would not permit it without adequate reason. Skeptical theism says we cannot judge what reasons a good God might have. So we cannot judge that God would not permit it. So the ground is removed, and the theist who accepted the theses in order to answer the problem of evil has thereby lost the argument from divine goodness to the reliability of appearances.

Two features make the objection difficult. It is generated entirely from within the theist's own commitments rather than imported from outside. And its structure is the same as the structure of the skeptical theist's own move: in each case, an appearance is neutralised by pointing out that reasons beyond our ken might be operating. A theist who thinks the manoeuvre is legitimate in one case has to say why it is illegitimate in the other.

The historical irony is worth noting. Descartes's use of divine goodness to secure the reliability of clear and distinct perception is the ancestor of a long tradition in which God's not being a deceiver does real epistemological work, and the same appeal underwrites the design-based accounts of proper function that rungs 10 and 13 depended on. Those accounts hold that our faculties are reliable because a good God made them to be. If we cannot judge what a good God would permit, we cannot judge that a good God would make them so, and the theistic route to the reliability of cognition closes at the same moment the theistic escape from evil opens.

The available reply is that permitting suffering and actively arranging deception are different in kind, since deception involves a distinctive violation of a rational agent's trust that mere permission does not. That distinction is genuine, but the objection can be reformulated to avoid it: the question is not whether God would deceive but whether God would permit us to be deceived, and permission was the very category the theses said we cannot assess.

That reply is the strongest available, and it also concedes the shape of the problem. The skeptical theist now has to sort inferences into two classes and show that the ones they need are of the safe kind. Nobody has produced a general criterion for the sorting, and until one exists the containment claim is a promissory note.

What it does to the theist's own case

There is a cost that is often overlooked, and it belongs to the argument of rung 11 rather than to ethics or epistemology.

The comparative framework requires each hypothesis to make some observations more expected than others. That is what allowed fine-tuning, the moral argument, and the argument from religious experience to count as evidence at all. Every one of those arguments claims that some feature of the world is what one would expect if God exists.

Skeptical theism denies that we can form such expectations in the domain of value. It is difficult to hold that we can tell what a perfect being would do when the question is whether to fine-tune a universe for life, and cannot tell when the question is whether to permit a child's death. Both are judgements about what a perfect being would have reason to bring about.

Paul Draper has pressed a version of this, and it is why his argument at rung 12 was constructed as it was: the more thoroughly the theist retreats into inscrutability, the less theism can be supported by anything. The theist who wants both the evidential arguments of the intermediate tier and the skeptical reply of this one is holding that our access to divine reasons is good enough to ground positive arguments and too poor to ground negative ones, and that combination requires a defence.

There is a further consequence, less discussed in the journals but more consequential for actual religious life, and it concerns petition. To ask God for something is to suppose that the thing asked for is worth asking for, which is a judgement about value, and to suppose that it is the sort of thing a good God might grant, which is a judgement about divine reasons. The theses do not forbid the practice, but they make the reasoning behind any particular request opaque to the person making it. The same holds for gratitude, since thanking God for an outcome presupposes that the outcome was good and that its coming about reflects well on the one thanked. A believer who has fully internalised the theses is in a strange position with respect to both, and it is not the position most religious traditions describe.

None of this shows the theses false. It shows that the price of the defence is paid somewhere other than where it is usually looked for.

Where it leaves things

The debate has settled into a recognisable shape, and it is worth stating what has and has not been established.

The core defensive claim has held. Noseeum inferences about justifying goods are weaker than they look, and the burden of defending the detectability assumption has been accepted by those who press evidential arguments. That is a genuine result, and the versions of the argument from evil that ignore it are worse for ignoring it.

The containment claim has not been established. No criterion has been produced that admits the inferences the theist needs and excludes the ones the atheist uses, and each of the three costs above remains a live objection with a substantial literature. The essays collected in Skeptical Theism: New Essays, edited by Trent Dougherty and Justin McBrayer (Oxford University Press, 2014), give a good sense of how unsettled the matter is.

And the position's cost in explanatory reach is real and is not usually acknowledged by its defenders. A theism that generates no expectations about the distribution of value is safe from evidence in both directions.

Next

Both parties to this dispute have been discussing suffering in the aggregate, as a quantity to be justified by some greater good. The next rung takes up a challenge to that whole way of framing the question. Some evils appear to ruin the life they fall on, in the sense that no balance of goods within that life could make it a life worth having lived, and a defence of God's permission that operates by outweighing has nothing to say about them. A further tradition holds that the attempt to justify such evils at all is itself the deeper mistake.