Problem of evil
A Socratic walk-through of the problem of evil — reasoned out one step at a time, not lectured.
The question we started with
THE QUESTION #Why would a world made by an all-powerful and wholly good being contain suffering that appears to accomplish nothing?
Everyone has to account for suffering; that is not peculiar to religion. The problem here is structural: three claims theology asserts together — that the creator can do anything, wants only the good, and made this world.
Notice first that "the problem of evil" names two arguments with different burdens, and most unproductive versions of this conversation conflate them. One says the three claims cannot all be true, as a square cannot have three sides. The other concedes they might and says the evidence tells against it. Replies that work on the first do nothing for the second.
Reasoning it through
REASONING #Take the logical version, in J.L. Mackie's 1955 paper "Evil and Omnipotence". The three propositions are not formally contradictory as they stand; bridging principles are needed. Mackie supplies two: a good thing eliminates evil as far as it can, and omnipotence has no non-logical limits. Add those and the set is inconsistent — not improbable, but such that one central commitment must go.
What would defeat a claim of that strength? Not a true explanation of suffering. Only a possibility: some proposition consistent with God's existence which, together with it, entails that evil exists. If such a proposition might be true, the charge of inconsistency fails.
That is what Alvin Plantinga supplies in God, Freedom, and Evil and The Nature of Necessity, both 1974. The free will defence runs: a world of significantly free creatures who freely do more good than evil is more valuable than one with none; a being who creates free creatures cannot determine what they freely do; and it is possible that every creaturely essence God could have actualised suffers what Plantinga calls transworld depravity — going wrong at least once in any world where it acts freely. If that is so much as possible, even omnipotence might have been unable to make free creatures who never sin.
Two things here are routinely misread. It is not a theodicy: Plantinga does not claim this is why there is evil, and need not. And it addresses moral evil, leaving natural evil to an extension most find far-fetched — that earthquakes and disease might be the free acts of non-human agents. Not fatal, since only possibility is needed, but fatal to using the defence as an explanation. With that qualification, both sides now largely accept that the logical version has not been sustained; Mackie himself later conceded roughly this much (recalled).
So the argument moved, and the burden changed. William Rowe's evidential formulation, from a 1979 paper (recalled), runs: there exist instances of intense suffering an omnipotent being could have prevented without losing a greater good or permitting an evil equally bad or worse; a wholly good being would prevent any such suffering it could; therefore there is no such being. Rowe does not claim the first premise is certain, only that it is rational to believe.
Look at how his example is built. A fawn is trapped in a lightning fire and dies over several days, unseen. No human freedom is implicated, so the free will defence has no grip; no observer is edified. Every candidate good is designed out.
Now the move everything turns on. From "we see no good that this suffering serves" Rowe infers "there probably is none" — an inference from failure to detect, a noseeum inference, legitimate only under a condition.
Stephen Wykstra named that condition in 1984 and called it CORNEA. You may move from "it appears there is no X" to "there probably is no X" only if, were there an X, you would likely see it. Not seeing a dog in a tent is evidence there is no dog; not seeing a flea is no evidence at all. Wykstra's claim is that we stand to an omniscient grasp of goods roughly as a child to an adult's, so goods that might justify a horror are exactly what we should not expect to see. This is sceptical theism: not a theodicy but a challenge to the inference.
It is a strong reply and an expensive one, because the scepticism does not stay where it was put. Three costs get pressed. Ordinary moral reasoning is undercut: if any horror before me might serve a greater good beyond my ken, my grounds for intervening weaken, and my confidence that it is an evil weakens with them. Natural theology is undercut symmetrically: if I cannot read the moral significance of a burning fawn, I cannot read it in a beautiful world either, and apparent goods stop counting as evidence for God. And revelation is undercut, since a being who permits horrors for reasons I cannot assess might permit deception for the same. Sceptical theists answer that the scepticism can be confined to whether we can survey possible goods. Whether it can be is the live dispute, and I do not think it settled.
A theodicy is bolder: an actual reason rather than a possible one. John Hick's soul-making theodicy, in Evil and the God of Love (1966), takes the world as an environment for the growth of persons — and meets Rowe's case again, since much suffering destroys rather than forms and falls on infants and animals.
The analogy
THE ANALOGY #Imagine watching through a window as a surgeon causes a child obvious agony. Three readings are open: she is not good, she cannot do this painlessly, or you do not understand the operation. The third is reasonable once. The evidential version asks what happens after a thousand operations, many ending with the patient dead and nothing evidently achieved. The reply that you are not a surgeon and cannot judge is true — and taken seriously enough, it also means you cannot tell surgery from assault when you next see a knife.
We have independent evidence the surgeon exists and is qualified — other patients recovered, the certificate is on the wall — whereas whether anyone is at the table is the very thing in dispute, so the analogy quietly grants the theist the conclusion and the atheist the framing.
Clarifying the model
THE MODEL #Three refinements connect these steps.
A defence and a theodicy are different undertakings, and the commonest error here is to attack a defence for being unbelievable. Possibility is all it claims, and an unbelievable possibility still defeats a charge of inconsistency.
Nothing in the free will defence requires libertarian free will to be actual, only possible — which is why the arguments in the free will piece, about whether we have such freedom, do not settle this one either way.
And gratuitousness, not quantity, is what the evidential argument is about. It does not ask why there is so much suffering; a theist can grant enormous suffering that all serves a purpose. It asks whether any instance serves none, which is why Rowe built his case on a single fawn.
A picture of it
THE PICTURE #How to readThe notation is repurposed: boxed requirements are claims an argument must make good, plain elements are the replies offered, and satisfies means "is aimed at this", not that the matter is closed. Follow derives down from the top — the evidential version is what remains once the logical one is answered, weaker by design. Contains marks the premise it rests on, that a justifying reason would be visible to us, the single point sceptical theism attacks. The lone traces arrow is the bill: the scepticism that blocks the inference also reaches the bottom requirement.
What became clearer
WHAT CLEARED #The argument has two forms and they have moved apart. Outright inconsistency is now generally taken to have failed, because defeating it needed only a possible story and Plantinga produced one. What survives rests on a single inference — that we would probably see the reason if there were one — and the defence denies our competence to judge. Each side pays: the atheist for an induction from what we cannot detect, the theist for a scepticism that does not obviously stay confined.
Key terms
TERMS #| Term | What it means |
|---|---|
| CORNEA | Wykstra's condition restricting when "I see no reason" licenses "there is no reason". |
Every term the collection defines is gathered in the glossary.