HistoricalChristian.Faith

The evidential problem of evil — does gratuitous suffering make God improbable?

Type: argument (off-timeline) · Assessment: contested — the evidential argument grants that God and evil are logically consistent (so the free-will defense is beside the point) and mounts a probabilistic case instead: the sheer variety and apparent pointlessness of suffering — Rowe’s fawn dying unseen in a forest fire — is evidence that no omnipotent, wholly good God exists. The argument’s logic is valid and its data real; its whole weight rests on one inductive step — from “we can see no God-justifying reason for this evil” to “there is none” — which is exactly what skeptical theism denies we are entitled to take. Neither side has carried the field: the exchange is the live centre of the modern problem of evil, and it is genuinely unresolved. · advanced by William Rowe (1979), with roots in Hume’s evidential turn

Related arguments: the modern analytic problem of evil — the logical problem & free-will defense · skeptical theism

See also: the modern-debate survey that sets Mackie, Plantinga, Rowe, and skeptical theism side by side; the skeptical-theist reply that this page folds in only in outline and that has its own full treatment; and the constructive theodicies the fawn is designed to outflank — the greater-good answer of Lactantius and the soul-making answer of Irenaeus (revived by John Hick).

The claim

The evidential (or “inductive,” or “probabilistic”) problem of evil is a different argument from the logical one, and William Rowe is careful to separate them. He concedes the logical problem: “Some philosophers have contended that the existence of evil is logically inconsistent with the existence of the theistic God. No one, I think, has succeeded in establishing such an extravagant claim.” What remains is “the evidential form—as opposed to the logical form—of the problem of evil: the view that the variety and profusion of evil in our world, although perhaps not logically inconsistent with the existence of the theistic God, provides, nevertheless, rational support for atheism.”¹ The charge is not that belief is incoherent but that it is, on the evidence, unreasonable.

Rowe casts it as a short valid argument:

1. There exist instances of intense suffering which an omnipotent, omniscient being could have prevented without thereby losing some greater good or permitting some evil equally bad or worse.
2. An omniscient, wholly good being would prevent the occurrence of any intense suffering it could, unless it could not do so without thereby losing some greater good or permitting some evil equally bad or worse.
3. There does not exist an omnipotent, omniscient, wholly good being.

— William Rowe, “The Problem of Evil and Some Varieties of Atheism” (1979).² Premise 2 is a near-truism about what perfect goodness would do; the whole live question is premise 1, the claim that at least some evils are gratuitous — that they serve no greater good an omnipotent being could not have got otherwise. To make it concrete Rowe offers the example the argument is remembered by:

Suppose in some distant forest lightning strikes a dead tree, resulting in a forest fire. In the fire a fawn is trapped, horribly burned, and lies in terrible agony for several days before death relieves its suffering. So far as we can see, the fawn’s intense suffering is pointless. For there does not appear to be any greater good such that the prevention of the fawn’s suffering would require either the loss of that good or the occurrence of an evil equally bad or worse.

— Rowe (1979).³ The fawn is chosen with care. Its suffering is unobserved, so no human is schooled by it; it exercises no creature’s free will, so it cannot be the price of freedom; it benefits no one. It is picked precisely to slip past every theodicy — free will, soul-making, the schooling of virtue — which all locate the point of suffering in some human good it produces. If even one such evil is truly pointless, premise 1 is true and, with premise 2, the theistic God does not exist.

Who advances it

The evidential turn is Hume’s before it is Rowe’s: in the Dialogues concerning Natural Religion (1779) Philo grants that the mixed world might be reconciled with a good author, then presses that it gives us no ground to infer one — the shift from consistency to probability. Rowe’s 1979 paper gave that intuition its canonical modern form and its stock example, and it became the reference statement of the argument in analytic philosophy of religion. Rowe himself refined it repeatedly: a 1988 restatement pairs the fawn with the case of a child brutally beaten, raped, and murdered (“Sue’s case”) to blunt the reply that animal minds might not really suffer, and by 1996 he had recast the whole argument in explicitly Bayesian terms — not “premise 1 is true” but “the facts of evil make theism less probable than its denial.”⁴ The argument in one form or another is now the standard atheistic case from evil, and the position skeptical theists, theodicists, and “G. E. Moore shift” theists all frame their replies against.

The evidence for

It concedes the free-will defense and loses nothing. The argument’s first strength is strategic: it grants everything Plantinga proved. Consistency is not the issue; a probabilistic argument is untouched by a demonstration of mere possibility. So the whole apparatus of the free-will defense — transworld depravity, feasible worlds, the possibility of fallen non-human agency — is simply irrelevant here, because it trades in what could be while Rowe asks what the evidence makes likely. The defender who thinks the free-will defense answered the problem of evil has answered a question Rowe is not asking.

The example is engineered to defeat the theodicies. The recurrent theistic move is to name a good the evil secures — freedom, moral growth, the deepening of virtue. Rowe’s fawn is built to have no such good available: unseen, it teaches no watcher; a brute, it makes no free choice; alone in the forest, it edifies no one. The point generalizes past the single animal. Rowe’s real datum is the variety and profusion of such cases — not one fawn but the vast mass of suffering that, so far as anyone can tell, purchases nothing. The cumulative weight, not the single instance, is what is meant to move the probability.

The inductive step is an ordinary, normally-reliable one. Rowe grants he cannot prove premise 1 — no one can survey all possible goods — but argues that we are constantly and rationally in the position of believing well-supported claims we cannot prove. We cannot prove that neither of two long-shot candidates will ever be elected President, he notes, yet it is entirely reasonable to believe it; likewise, having looked hard for a God-justifying reason for the fawn and found none, it is reasonable to conclude there is none.⁵ This is the everyday inference from “we have searched and found no X” to “there is no X” — the kind of “no-see-um” reasoning we rely on without embarrassment in ordinary life and in science.

Adversarial assessment

The load-bearing joint is the inference in premise 1, and it is contested by a serious rival. The argument stands or falls on the move from “so far as we can see, the fawn’s suffering is pointless” to “the fawn’s suffering is pointless” — from an appearance of gratuitousness to gratuitousness itself. Skeptical theism (Wykstra, Bergmann, Alston) attacks exactly this step: a no-see-um inference is licensed only where, if the thing existed, we could reasonably expect to detect it, and (the skeptical theist says) we have no such expectation about the goods an infinite mind might weigh — our grasp of the space of possible goods and their entailments is no more representative of the whole than a one-month-old’s grasp of its parents’ reasons. That reply is independently advocated and has its own full page here; it is not a knock-down of Rowe (it draws heavy counter-fire of its own — see that page), but it makes premise 1 a genuinely open question rather than an obvious datum.

Rowe himself conceded ground and retooled. The honest measure of how contested the inference is: its author did not treat it as settled. In direct response to Wykstra, Rowe granted that the bare appearance of pointlessness does not by itself carry the day and reworked the argument — distinguishing the goods we know from the goods there are, and eventually restating the whole case probabilistically rather than as a flat assertion that premise 1 is true.⁶ A live argument that its own proponent keeps rebuilding under pressure is, by that fact, contested.

The “G. E. Moore shift” runs the argument backward. Rowe’s argument is valid, so the theist may take it in reverse: deny the conclusion (there is a God) and infer that premise 1 is false (the fawn’s suffering is not gratuitous, appearances notwithstanding). Which direction is rational depends on the comparative warrant of “this evil looks pointless” versus “God exists” — and a theist with independent grounds for belief may reasonably find the latter the firmer, and so let it defeat premise 1.⁷ The shift does not refute the argument; it shows that the argument moves someone only to the extent that its premise is better-supported than the theism it targets, which is precisely the disputed point.

What the argument does not overreach to claim. To Rowe’s credit, he does not present the argument as a proof of God’s non-existence, nor even as showing the theist irrational. He is a self-described “friendly atheist”: he holds that the evidence makes atheism the more reasonable belief while granting that some theists, given their other grounds, may be rational too.⁸ That modesty is itself a reason not to rate the argument higher than contested: it claims to shift a probability, not to close a case, and even its author concedes the other side may hold its ground.

Assessment

Contested. The evidential argument is the strongest and most durable form of the problem of evil, and it is honestly weighed as genuinely undecided rather than as either a proof or a failure. Its logic is valid, it sidesteps the free-will defense by asking about probability rather than possibility, and its data — the vast mass of apparently gratuitous suffering, the fawn among it — are real and not dissolved by any theodicy that locates the point of evil in a human good. What keeps it short of sound is that its whole force rests on a single inductive step — from “we see no justifying good” to “there is none” — which skeptical theism contests on principled epistemic grounds, which the “Moore shift” lets the theist run in reverse, and which Rowe himself conceded needed reworking. And what keeps it well above weak is that the skeptical-theist reply carries serious costs of its own and has not won consensus either. This is a real philosophical stand-off with no agreed victor — the textbook case of an argument that is strong by one discipline’s lights and resistible by another’s. Rate it, and its reply, as the open dispute they are.

Sources