How serious an evidential problem is evil for theism?

A way to decide

Anti-theistic arguments from evil come in two general varieties. There are ‘logical’ arguments, which claim that certain evils conclusively disprove theism because their existence is incompatible with it. And there are ‘evidential’ (or ‘probabilistic’ or ‘inductive’) arguments, which merely claim that certain evils disconfirm theism – that is, decrease its epistemic probability.

There are several different ways of putting both logical and evidential arguments, and several different evils which could be appealed to, ranging from particular evil acts to the total amount of disvalue in the world. There is no chance of considering all these arguments. So I shall focus on one particular form of the evidential argument, which is sufficiently broad to let me consider how serious an evidential problem evil is for theism. This argument mirrors the following popular version of the logical problem of evil, differing only in claiming that premise (2) is probable rather than certain, rendering the conclusion equally probable:

(1) Were there a God, there would be no unjustified evils.

(2) There are unjustified evils.

∴ There is no God.

Here, an ‘unjustified’ evil is one which God can increase the expected amount of value in the world by preventing. Take the example of Oswald shooting Kennedy. If God gave Oswald libertarian free will and significant scope to exercise this freedom without His intervention, He would not know that Oswald would shoot Kennedy, and so would not know the ultimate net value of Oswald’s freedom. He would only know the expected value of this freedom, which might be positive even if the actual value turned out to be negative. Kennedy’s murder would then be a justified evil, because God could only have preventing it by robbing Oswald of the freedom to do evil, which He would have expected to decrease the world’s value. (Atheologians can’t suggest God should have miraculously intervened once Oswald subverted His expectations by firing, for that would mean Oswald’s freedom never had much scope, which we’re currently supposing robs it of great value.)

(1) does not require God to maximise the world’s expected value, which is impossible because He could add goods to any world. So I will neither count the absence of goods as an evil nor require theists to defend Leibniz’s notorious (1710) claim that this is the best possible world. They need only reject either (1) or (2).

(1), however, is highly plausible. Theism’s God is omniscient and benevolent. Being omniscient, He would know how to prevent unjustified evils; being perfectly benevolent, He would exercise his ability to do so. That, at least, is how I understand benevolence, following Swinburne. (Swinburne has apparently strengthened his earlier, overly lax requirement that God merely “does no morally bad action” and “does any morally obligatory action”. As Gale puts it: “Many an S.O.B. satisfies this duty-based requirement.”) So I take God’s essential benevolence to preclude Clark’s (1961) ‘solution’ to the problem of evil, which lets God allow unjustified evil and hence deems Him “evil by our standards”. Equally irrelevant here are solutions like ‘process theodicy’, which reject the theistic definition of God as an omniscient, omnipotent, benevolent Creator.

So theists must reject (2). On the current argument, the evils we see around us provide serious evidence against theism only insofar as they render (2) epistemically probable. To settle this, we must evaluate the probability of these evils being justified in some way. A way in which they could be justified is called a ‘theodicy’. Pretend for now that only two simple theodicies, involving three contestible claims each, have been imagined. Then we can use standard calculations to evaluate the probability of (2) on the basis of the probability of these claims, and the ‘Catch-All’ claim that some unknown theodicy is correct, considered in the shaded box in the (purely illustrative!) chart below.

So, granting (1) in the above argument, we can decide how serious an evidential problem known evils pose for theism by seeing how probable it is that they are justified in some way: i.e. that some possible theodicy is correct. If the two theodicies above were the only possibilities, we could do so by assigning epistemic probabilities to the claims they involve. Ramsey (1926) suggests doing so by judging the fair price of a bet on their truth, as a proportion of the jackpot. The fair price is the highest price we’d reasonably expect to at least break even. If we knew a claim with certainty, we’d count on breaking even after wagering the whole jackpot, so its epistemic probability for us would (appropriately) count as 1. With equal reason to accept or reject it, we’d gamble 0.5 of the jackpot: any more and we’d expect a loss.

No doubt this method will not yield precise values in all cases: we might only agree that the bet’s fair price lies within a certain range. That is why I use the < sign: our numbers should be above this range. They should be high enough for us to be confident that their fair values are lower. For example, suppose a £1 prize is offered if Oswald actually shot Kennedy. Being relatively knowledgeable and reasonable about this, I’d buy this bet for up to around 99p, but I’d be really confident only that the fair price is above 90p. So I’d evaluate a theodical claim that Oswald didn’t shoot Kennedy generously, as below 0.1. Generosity is appropriate, since atheologians have the burden of proof here.

Unfortunately, we cannot always be even this confident about the epistemic probabilities of theodical claims. For example, the existence of unpredictable free will, and its value, are notorious areas of disagreement. I shall show why some consider these and other theodical claims implausible, but to decide how serious the evidence against theism is everyone must ultimately evaluate such claims for themselves, putting the values they arrive at into charts like that above. If they consider themselves unable to evaluate some theodical claims, and if assigning these claims a probability of 1 would make (2)’s probability low, they should not consider evil a serious evidential problem for theism. This is why Van Inwagen’s “extreme modal and moral scepticism” defuses the problem: he claims that some possible theodicies involve modal and moral claims which we cannot know to be improbable. I shall consider moral scepticism soon, and leave his modal scepticism aside, noting only that his claim that it is “epistemically possible” that God could only create a universe which obeyed this universe’s laws may compromise divine omnipotence more than theism can allow. In general, he allows God to be bound by extremely restrictive ‘metaphysical necessities’. Unlike Van Inwagen’s views on modality, mine let God be bound only by logical necessity, and are hence relatively non-sceptical; readers who disagree should reject those parts of my argument which rely on this. But they should not feign scepticism about evaluations of probability they confidently make in other contexts.

For example, if within the context of public policy we judge that Hurricane Katrina’s not occurring would probably have had beneficial consequences overall, we should not accept Wynn’s claim that God might well need such an event to cause greater goods. Even non-consequentialists must occasionally rely on such judgements, such as when charged to set policy solely to maximise beneficial consequences. If we rejected them, we’d have no reason to fear climate change, but not even so-called ‘sceptical environmentalists’ claim this. So a preliminary objection to ‘sceptical theism’––which denies that we are in any position to judge whether Hurricane Katrina was necessary for greater goods––is that it would paralyse public policy. Howard-Snyder (1996:292) defends sceptical theists’ ability to justify moral judgements by noting that they can judge that public servants do not know of these greater goods even if God might, but that does not let these servants make justified consequentialist moral judgements, only blameless ones––which is not much practical use. He cannot simply recommend that they make the best judgement they can in ignorance of these goods, for his position entails that their judgement will be no more accurate than flipping a coin.

I shall consider sceptical theism later, when considering ‘Catch-All’. To explain why the chart above considers Catch-All last, I must first describe the background knowledge on which we should evaluate the probability of such theodical claims, and thereby the seriousness of our evidence against theism. Much of what I’ll describe is counterintuitive, because background knowledge isn’t always real knowledge, but it all follows from confirmation theory.

Firstly, our background ‘knowledge’ must include the truth of theism. This is because a theodicy is an integrated story describing a way in which God and evil could coexist, so includes the existence of God. It doesn’t mean that theists’ typical moral views are presupposed; indeed a major conclusion of this essay is that many non-theists’ moral views appropriately make their evaluations of theodicies, including Catch-All, harsh. But it means that non-theists can’t employ other views which presuppose atheism, and that if theism entails that a specific theodicy is correct, evils provide no evidence against theism. (Theists cannot exploit this obvious concession by arguing that, given premise (1) above, theism entails that some theodicy is correct, because it does so only given the evidence of actual evils. As the very evidence at issue, we must feign ignorance of it in our background knowledge to avoid the notorious ‘problem of old evidence’ – cf. my previous essay. So open-minded theists who reject all known theodicies cannot assume without further argument that Catch-All is correct. O’Leary-Hawthorne (1994) shows that parallel reasoning would, absurdly, let geocentrists ignore evidence against their theory because geocentrism entails that there must be some geocentric explanation of apparent counter-evidence, since otherwise it would not be true.)

Each claim on the chart must also be evaluated taking the truth of earlier claims within the column as background knowledge. (The ordering of claims is irrelevant, since by Bayes’ theorem P(p) × P(q|p) = P(p&q) = P(q) × P(p|q), and repeatedly swapping adjacent premises allows any reordering.) This is intuitive for the ‘Theodicy’ columns, since we’re evaluating the probability of all a theodicy’s required claims being true. But it might seem unfair in the ‘Conclusion’ column, since Theodicy 2 didn’t require Theodicy 1’s incorrectness, and Catch-All doesn’t require known theodicies’ incorrectness. However, it is necessary to prevent double-counting of theodicies, and to account for their probabilities not being independent.

To see this, consider theodicists’ ability to combine elements of old theodicies into new ones. This is often necessary, since theodicies in our chart must cover all known evils, and most traditional theodicies fail for some specific evils. We could combine sceptical theism about natural evil with the rest of Theodicy 1: we shall see that sceptical theists may be right that the only plausible theodicy for natural evil lies within Catch-All. Or we can combine Theodicy 1 and Theodicy 2 into a new comprehensive ‘Theodicy 3’, dropping the weakest premise of each (see for yourself). But this theodicy’s probability will obviously depend on theirs. That is why, at least in the ‘Conclusion’ column, the probability of its failure should be conditionalised on their failure. That will increase it slightly, though less than if it had recycled their least plausible claims, for if some of lower-numbered theodicies’ claims fail these are the most likely candidates. Only this conditionalisation prevents theists from winning by attrition, recombining implausible claims into endless new theodicies and claiming that at least one of these is likely correct without considering their lack of independence. It won’t disadvantage then, for they need only show that one theodicy is correct.

We can finally see the full difficulty of evaluating the seriousness of the evidential problem of evil. Theists need not be modal or moral sceptics (or even ‘sceptical theists’) to question our ability to do so. They need only suggest that some theodicies are hard to dismiss while properly accounting for the influence of background knowledge. That is philosophy, however: difficult, but worth attempting, for the only alternative is extreme agnosticism about not only theism but also theories like geocentrism. We should evaluate theodicies as best we can, while maintaining a proper level of humility: one consistent with our attitudes to other theories.

Now, I allowed that people’s evaluations of theodical claims differ. In particular, they will depend upon their different moral systems: these often determine whether they consider the problem of evil serious. Atheologians sometimes accept subjective-state consequentialism, an ethic which hinders theodicy, because positive subjective-states are plausibly logically independent of negative ones (the relatively unhelpful exceptions being certain ‘higher goods’ we’ll consider later). Theodicy 1, by contrast, relies on retributive justice being intrinsically good, which atheologians often deny. To his credit, Swinburne (1998:xii-xiii) recognises that theodicies often need moral claims many people reject, and that neither side can easily convince the other of their moral error. Reactions as different as Mackie’s (1955) claim that evil is logically incompatible with theism and Howard-Snyder’s (2001) claim that it provides no evidence against theism whatsoever can be vindicated given different moral views, and seem mystifying before we realise their ethical presuppositions. For example, Mackie’s views involved the acceptability of compatibilist free will, and we’ll see that Howard-Snyder’s sceptical theism itself presupposes the falsity of certain ethical perspectives, and so requires ethical evaluation itself. Rather than being carried out in isolation, as it too often is, this evaluation should be informed by atheologians’ objections to known theodicies: only then will it engage with the specific difficulties which have afflicted theodicy, and which account for the problem of evil’s seriousness. That is why the ‘Conclusion’ column puts Catch-All last, because within that column’s context we should treat the incorrectness of known theodicies as background knowledge when evaluating it.

Although the considerations above obviously preclude any comprehensive or conclusive evaluation of the evidential problem of evil, I shall indicate criticisms of traditional theodicies which, for those who accept them, make it implausible that either they or any other theodicy (including those within Catch-All) can work. I’ll focus particularly on the popular appeal to free will, on which many other theodicies rely, and then indicate an ethical perspective from which other known theodicies are implausible, and the prospects for unknown theodicies are poor.

Why some reject known theodicies

We have already seen the outlines of a free will theodicy for moral evil, and noted its reliance on libertarianism, which many consider implausible. But I shall now describe some of its other vulnerabilities, which constitute theodical claims to be entered into our chart, and further illustrate why some also find them implausible.

First. It is not obvious that even libertarians cannot let God predict human actions with significant confidence. Our nature and circumstances might not determine our choices, but they certainly shape them; psychologists can predict that people with Oswald’s background often become violent, and God’s predictions would be even better. So perhaps He could shape the world––by tempering our natural selfishness, and ameliorating the resource shortages which encourage competition, for example––so that the expected amount of moral evil would be lower than it actually is. If so, its actual amount is probabilistic (though not conclusive) evidence against a free will theodicy. (I leave Plantinga’s suggestion of trans-world depravity aside, since even he doesn’t regard it as significantly probable, instead targeting it against logical problems of evil.)

Second. Even if Oswald’s libertarian freedom was highly unpredictable, this is arguably insufficiently valuable to outweigh the chance of his killing Kennedy. Now, Swinburne thinks that having significant control over the world, with temptations to do evil which have to be overcome and may well not be, is sufficiently valuable to outweigh the disvalue of moral evil. That would justify God’s not shaping the world in the way suggested above, since limiting Oswald’s temptations would remove such value. But Swinburne concedes that many modern ethicists (and some ancient virtue ethicists) would disagree, regarding freely chosen good as equally valuable even when it does not require overcoming temptation, but follows almost inevitably from the feeling of moral compulsion expressed so well in Luther’s “Here I stand; I can do no other”. They will evaluate free will theodicies harshly, protesting that a benevolent God would spread this feeling of compulsion by improving our moral knowledge and moral nature.

Note that the predictability of Luther’s action given his particular character did not stop it being an intrinsically valuable exercising of free will. This example shows that, pace Hartshorne 1962, predictability would not rob us of the power to shape the world, which according to Swinburne accounts for freedom’s worth. It would just mean we would never exercise that power unwisely, and Swinburne’s equating that with not having it is equivocation: we don’t attribute powers only to those who use them in every way they could. Perhaps the term ‘freedom’ implies not just the logical possibility but the statistical probability of its misuse, but then, unless significant probabilities of our misusing our freedom are valuable, we should not obviously want ‘freedom’.

Third. The special value of libertarian freedom is itself debatable: many philosophers think compatibilist freedom would be equally valuable, allowing control of our fate, free from external constraints. I myself doubt that libertarianism provides enough value over-and-above compatibilism to justify moral evil, but others may share Kant’s view that compatibilism affords “all the freedom of clockwork”; I raise this point simply to further illustrate how some perspectives preclude free will theodicies.

Fourth. Even if libertarian freedom is especially valuable, we may not have it. There is as yet no good way to fit it into our (admittedly incomplete) scientific account of the mind, appeals to quantum mechanics notwithstanding. And Hobart (1934) argues that it is impossible.

These considerations suggest that the second premise in Theodicy 1 requires that the first premise deem free will not only indeterministic but also highly unpredictable. This threatens the plausibility of both premises, while the third premise––an Augustinian extension of the theodicy to natural evil––adds implausibility to implausibility. As noted, it will tempt only those with a strict retributivist ethic: the ‘punishment’ is often capital, offering no opportunity for reform, except perhaps in an afterlife. Even many Christian evaluators will deny that this is just, particularly after considering the distribution of natural evil, which often afflicts the innocent while letting the wicked prosper.

Plantinga suggests another possible extension of the free will theodicy to natural evil: it is really moral evil, committed by fallen angels. This is one of the few attempts to justify all natural evil (even hard cases like animal suffering), so if plausible could usefully supplement other theodicies which cover moral evil. But unfortunately it faces many of the criticisms levelled at free will theodicies above, and more besides. For instance, Swinburne (2004:240) notes that the existence of such angels is not very probable even given theism, and is an additional hypothesis which adds complexity and is thus by his measure less probable. Furthermore, the creation of angels would have to be a very great good to justify a chance of the horrendous natural evils we see. And unless freedom is as unpredictable as discussed above, this chance will not be very great unless God gave angels utterly perverse desires which could overcome their direct awareness of Him. So theodicists will be in trouble if they need fallen angels to explain natural evil; they’d be better off suggesting there are unknown justifications for some of this.

Now, someone who accepts the above criticisms of free will theodicies may well deny that other known theodicies can take their place, if she shares those ethical views we saw Swinburne allow were incompatible with his theodicy. For example, Swinburne’s suggestion is that there are there are ‘higher goods’ which presuppose the existence of evils: one cannot nobly endure suffering without suffering, so if noble endurance is sufficiently valuable, suffering might be justified. But she will deny that endurance, or other purported higher goods, are so valuable, suggesting that although suffering provides opportunity for kindness, sympathy, sacrifice and endurance, and these are praiseworthy things, the world would be better without any need for them. She would favour a world without natural evil, and filled with moral ‘Luthers’ (even if they have only compatibilist freedom, or predictable libertarian freedom).

For instance, such worlds are favoured by many subjective-state consequentialisms (including, but not limited to, hedonistic utilitarianism). As promised, I’m not arguing these are correct, merely flagging them as perspectives from which theodicies are particularly implausible. But they underpin many negative evaluations of theodicy: consider protests like “A world without evils like cancer would obviously be better, and that’s so despite the lost opportunity for endurance, so this can’t be a ‘higher good’!” Many do think this “obvious”, which would make (2) equally obvious, but theists often claim it is too ambitious a statement for us to even justifiedly believe. However, for consequentialists it is no more ambitious, and so no less confident, than their judgement that we should eradicate cancer (of which it is a corollary, for they’d see no reason to eradicate cancer would it have been a justified evil for God). That is why they react with such horror to Mother Teresa’s reported claim that “it is very beautiful for the poor to accept their lot... the world is being much helped by the suffering of the poor people.” Unlike her, they cannot accept it yet think charitable relief of poverty a still higher good, for they regard its value as entirely dependent on its consequences’ value (i.e. on a world without poverty being better) rather than on the act of relief’s intrinsic value. Here we have another demonstration of my claim that different evaluations of the problem of evil’s seriousness depend on different ethical perspectives.

Sceptical theism and unknown theodicies

Even many theists accept that known theodicies fail to explain all evil and suggest that unknown theodicies plug their gaps. Such an appeal to mystery is perhaps reasonable from a perspective which differs from that described above, accepting that free will may justify much evil, and that there can be higher goods which justify evils like cancer without this implying we shouldn’t seek their eradication. On this ethical perspective, complex justifications involving goods unrelated to our recognisable quality-of-life might be possible, and known theodicies’ past successes will inspire confidence in unknown ones.

But those who reject known theodicies for the reasons above, or for other similarly fundamental reasons, will take the ways they fail to suggest deep-seated flaws in the project. For then, as Hare (1975:9) says, “the repeated failure, the recurrence and clustering of criticisms, the permutation of basic moves which have been found wanting, and the slight variations of old favourities is evidence that counts heavily against the likelihood of eventual success”, and even against the possibility Hare omits, viz. that some forever-unknown theodicy is correct. The sorts of goods these people allow cannot logically require evils in ways we can’t imagine, which makes millennia of failed attempts to imagine such ways strong inductive evidence that they don’t exist. They typically reject strong ‘recognition-transcendence’ in meta-ethics, which is implicit in Howard-Snyder’s (1996:296-297) claim that millennia of searching aren’t enough in ethics, since we have a limited grasp of ethical truths. That might seem insufficiently humble, for illuminating scientific explanations of previously puzzling phenomena can elude millennia of mental searching (as can many other things, such as mathematical proofs). But, as Tooley (1991:112-113) argues, humility about our knowledge of basic ethical truths on which justifications for evils must be founded is quite different, and some reasonable evaluators reject it: indeed, we saw that their confident consequentialist verdicts on cases like Hurricane Katrina and cancer require them to do so.

So evil constitutes serious evidence against theism from some perspectives. Unless these perspectives can be proved correct, that is a modest conclusion; perhaps all we can expect after 5,000 words, given my earlier description of the problem’s complexity. I find them plausible myself, and more importantly so do many others. That should interest even theists, for it not only shows that justifying a rejection of these perspectives is a precondition of reasonable theism which––given the structure of the argument above––theism’s other justifications cannot outweigh, but also challenges any natural theology which purports to start from religiously neutral premises. Such natural theology is useful in convincing non-believers, and also possibly vital for rational religious belief if this requires Cartesian reconstruction from neutral premises (something Plantinga (1983) and others dispute). So natural theologians must now refute those ethical perspectives which generate serious problems of evil, and do so without undermining our confidence in ethical views on which natural theological arguments rely. (Otherwise ‘sceptical atheists’ can question whether we’d expect fine-tuning from a benevolent God, suggesting fine-tuning might entail unknown ‘higher evils’.) This is a difficult needle to thread, and as Wynn (1999:95-96) recognises, it cannot rely too much on sceptical theism, or scepticism of other forms, such as Van Inwagen’s modal scepticism, which he himself recognises would undermine the fine-tuning argument. Wynn thinks natural theology requires an ability to make a rough evaluation of the world’s overall goodness: this is precisely what sceptical theists doubt, but Wynn thinks that risks “call[ing] into question the point of human choice in accordance with humanly recognised evaluative standards”.

Such considerations show that dealing with evil is challenging even from many theistic perspectives. But, especially in light of the subtleties of evaluation described earlier, evil is not obviously serious counter-evidence to theism from these perspectives, at least in the final judgement. Showing that it is would require demonstrating that inconsistencies and implausibilities afflict known theodicies even from more hospitable perspectives (e.g. those which endorse libertarianism, retributivism, etc.), and that, even from these perspectives, known theodicies’ failings are inductive evidence of unknown theodicies’ falsity. Martin (1990) attempts such a project, but I am less confident of his attempts than of the more restricted, perspective-relative conclusion I’ve defended above. In particular, his inductive argument seems sometimes to miss sceptical theists’ point.

Whither, then, the debate about evidential arguments from evil? If my argument has been correct, their seriousness depends on our fundamental ethical perspective. Our very ability to evaluate this seriousness also depends on that perspective, insofar as it affects our confidence in evaluative judgements like that embodied in the protest about cancer above. Theodicists and atheologians alike must ultimately follow Swinburne in turning their attention to these ethical questions, without uncritically relying on their own perspective. The consequentialist perspectives I’ve described will doubtless strike many as implausible, but this must be argued. That is less easy than typically thought, especially once we realise that they can be more flexible than hedonic utilitarianism. The difficulty of settling such fundamental ethical debates, and the common failure to recognise that other people are working from radically different ethical and meta-ethical views, go far to explain the persistent and radical disagreement about just how serious the evidential problem of evil is.