The Problem of Evil

This is the only argument in the subject that people arrive at without being taught it. Nobody constructs a cosmological argument on their own. Everybody, at some point, looks at something that has happened to someone and formulates a version of what follows.

It usually arrives attached to a particular case rather than as a general claim, and the philosophical treatment can feel like a betrayal of that. Handling the argument carefully is not the same as taking suffering less seriously, and the last section of this article returns to the question of whether the two can be kept apart.

The shape of the problem

The oldest statement circulates as the Epicurean paradox, and the attribution deserves care rather than repetition.

Lactantius, in De Ira Dei, attributes to Epicurus a trilemma: God either wishes to remove evils and cannot, in which case God is not omnipotent; or can and does not wish to, in which case God is not good; or neither wishes to nor can, in which case God is neither; or both wishes to and can, in which case the question is where the evils come from.

No surviving text of Epicurus contains this argument. Scholars have proposed that Lactantius took it from a Sceptical source, with Carneades suggested as a candidate, and have noted that Lactantius had reason as a Christian polemicist to place it in the mouth of a philosopher he regarded as an atheist. So the standard name records an ancient report rather than an established attribution, and the safe statement is that Lactantius attributed it to Epicurus.

David Hume put a version of it in Philo's mouth in the Dialogues Concerning Natural Religion of 1779, following Lactantius' attribution, and that is the route by which it entered modern philosophy.

One structural feature makes this argument unlike everything else in this tier, and it is worth stating before any of the detail.

Every premise comes from the theist. The critic does not have to establish that the universe is eternal, or that the constants could have been otherwise, or that obligations are natural facts. The critic takes the attributes as the tradition states them, adds the observation that suffering exists, which no one disputes, and asks how the set can be held together. An argument built entirely out of the opponent's own commitments cannot be answered by attacking the opponent's evidence, because there isn't any.

Two kinds of suffering

A note on the word first. "Evil" here is a term of art covering bad states of affairs generally, not wickedness. A fawn's suffering is an evil in this sense, and so is a toothache. The vocabulary is unfortunate and it is entrenched.

The distinction between the two kinds carries more weight than any other in this area, because the responses divide along it. The family of responses that appeals to the value of free choice covers moral evil directly and covers natural evil only by some further manoeuvre. Everything about the shape of the debate follows from that asymmetry, and a reader who keeps it in view will follow the next article more easily.

The boundary is blurred in practice, and both sides overstate how cleanly it divides. A famine is a natural event whose lethality is largely a matter of decisions about distribution. An earthquake kills in proportion to who was made to live in badly built housing. A disease is natural and who is treated for it is not.

That said, cases on the clean side of the line exist and are what the argument uses. An animal dying in a fire in an unpopulated forest a hundred thousand years ago involves no human decision at any point.

The logical version

J. L. Mackie gave the influential modern statement in "Evil and Omnipotence", published in Mind volume 64, number 254, in 1955, at pages 200 to 212.

Mackie was careful about what he was claiming. The four propositions, that God is omnipotent, that God is omniscient, that God is wholly good, and that evil exists, are not explicitly contradictory. Deriving a contradiction requires additional principles, and he supplied two: that a good thing eliminates evil as far as it can, and that there are no limits to what an omnipotent being can do. With those added, the set becomes inconsistent, and if the added principles are necessary truths then theism is not merely false but impossible.

He also anticipated the obvious response and argued against it in advance. Suppose the reply is that God permits evil because free agents are valuable and freedom carries the possibility of misuse. Mackie's counter is that there is no contradiction in a person freely doing the right thing on a given occasion, since that happens constantly, and so there is no contradiction in a person freely doing the right thing on every occasion. An omnipotent being could have created agents of that kind, and did not.

The current position is that the logical version is not successful, and the reason is a piece of work by Alvin Plantinga that the next article covers in the detail it requires. Two things about that outcome should be recorded here, because both are regularly misreported.

The reason the response works is structural. A claim of logical inconsistency says there is no possible way for both things to be true. Defeating it requires describing one coherent way, and that is a much lower bar than establishing what is actually the case. Mackie himself acknowledged the force of the reply.

And the defeat is limited in a way that apologetic summaries almost never mention. What has been shown is that theism and the existence of some evil are not formally inconsistent. Nothing whatever has been shown about the quantity of suffering, its distribution, or whether any of it looks like the sort of thing a perfectly good and unlimited being would permit. Announcing that the problem of evil has been solved because the logical version failed is a misdescription of what happened, and the philosophers who did the work do not make that claim.

What a response has to do

The contrasting kind of response claims to identify God's actual reasons, and it goes under its own name, theodicy, which the next article takes up in full.

The distinction is not a piece of terminology. It tracks exactly what each version of the problem requires, and getting it wrong is the most common error in this area.

Against the logical version, a defence is sufficient. The critic has claimed impossibility, and one coherent possibility refutes a claim of impossibility. The possibility does not have to be plausible, or likely, or believed by the person offering it.

Against the evidential version, a defence is not sufficient, and the reason is a point about probability that the beginner tier established. A bare possibility does not raise the probability of anything. If I claim your keys are probably in the kitchen and you reply that they could conceivably be in the car, you have not shown my claim false; you have described an alternative without giving any reason to expect it. So a response to the evidential problem has to make its proposal probable, or at least not improbable, which is a different and much harder task.

Everything that follows in the next article turns on that difference, and the arguments there are best read by asking of each proposal which job it is trying to do.

The evidential version

This definition identifies what theism is committed to denying, and stating that commitment plainly is the first step of the evidential argument.

Theism does not have to deny that terrible things happen. It has to deny that any of them are gratuitous. On the classical account, every instance of suffering that has ever occurred is either necessary for some good that outweighs it or unavoidable without something equally bad or worse. Not most instances. Every instance, including each individual one, taken separately.

That is an extremely strong claim, and it is not a claim the theist has added for the sake of the argument. It follows from the attributes: a being that could have prevented a pointless horror and did not is not perfectly good, and a being that could not is not omnipotent, and a being that did not notice is not omniscient.

William Rowe gave the version that set the terms of the modern discussion, in "The Problem of Evil and Some Varieties of Atheism", published in the American Philosophical Quarterly, volume 16, number 4, in 1979, at pages 335 to 341.

His example is a fawn. Lightning starts a fire in a distant forest. The fawn is trapped, badly burned, and dies over several days in agony. No human being ever knows it happened.

The argument then runs: there exist instances of intense suffering that an omnipotent and omniscient being could have prevented without thereby losing some greater good or permitting some equally bad or worse evil; a wholly good being would prevent any such suffering it could; therefore no omnipotent, omniscient, wholly good being exists.

The example is engineered rather than emotive, and seeing what it is engineered to do is the point of it. There is no free choice involved, so responses appealing to the value of freedom do not reach it. The sufferer cannot develop morally through the experience, and dies. No observer is present to be improved by witnessing it or moved to compassion by it. It is one case among an uncountable number stretching back through the whole history of animal life on the planet, most of it before there were any humans to be affected by any of it. Each of the standard replies is blocked in advance by a feature of the case.

Two features of Rowe's own presentation are almost always dropped when the argument is repeated, and both are worth having.

He did not claim the first premise could be proved. He presented it as a reasonable inference from what we can observe, and acknowledged that someone might reasonably reject it.

And he distinguished varieties of atheism accordingly, holding that a theist with independent grounds for theism could rationally reject the first premise while the atheist rationally accepts it. He called the resulting position friendly atheism. Whatever one concludes, that is how to conduct an argument, and it is a standard the surrounding literature has not always met.

The principal reply, which the next article develops, targets exactly the inference Rowe flagged. Our not being able to see a justifying reason for a particular evil is weak evidence that there is none, given the size of the difference between our cognitive position and that of an omniscient being. Rowe's counter is that inferring from "no such thing is discernible" to "there is no such thing" is how everyone reasons about everything, and that suspending it in this one case has consequences that its proponents may not want.

Draper's version

Paul Draper restructured the argument in "Pain and Pleasure: An Evidential Problem for Theists", published in Noûs volume 23 in 1989 at pages 331 to 350, and his version avoids the point at which Rowe's is attacked.

Draper does not claim that any particular evil is gratuitous, which means the reply about our cognitive limits has nothing to fasten on. Instead he compares two hypotheses. One is theism. The other is a hypothesis of indifference, on which the nature and condition of sentient beings is not the result of any benevolent or malevolent action by a non-human person.

He then asks which hypothesis better predicts the specific facts we observe about pain and pleasure. What we find is that pain and pleasure are systematically tied to biological function, distributed in a pattern that tracks reproductive advantage. Creatures feel pain where feeling it helps them survive and not otherwise, and the correspondence is close. Draper argues that this pattern is very much what one would expect if the process producing sentient life were indifferent, and considerably less what one would expect if it were governed by a being concerned with the welfare of the creatures.

This is the likelihood structure from the fine-tuning article, run in the opposite direction, and it inherits both the strengths and the difficulties described there. It does not require any claim about what God could or could not have reasons for, and it does require a judgement about what theism predicts, which is exactly where theists resist it.

Next

The problem as stated here is an argument. The next article takes up the responses, which fall into two families corresponding to the distinction between a defence and a theodicy drawn above, and which have to be assessed by asking of each one whether it is answering the logical version or the evidential version, since a response adequate to the first is not adequate to the second.