J. L. Mackie, "Evil and Omnipotence," Mind, New Series, Vol. 64, No. 254. (Apr., 1955), pp. 200-212.
EVIL AND OMNIPOTENCE
By J. L. Mackie
The traditional arguments for the existence of God have been fairly thoroughly criticised by philosophers. But the theologian can, if he wishes, accept this criticism. He can admit that no rational proof of God's existence is possible. And he can still retain all that is essential to his position, by holding that God's existence is known in some other, non-rational way. I think, however, that a more telling criticism can be made by way of the traditional problem of evil. Here it can be shown, not that religious beliefs lack rational support, but that they are positively irrational, that the several parts of the essential theological doctrine are inconsistent with one another, so that the theologian can maintain his position as a whole only by a much more extreme rejection of reason than in the former case. He must now be prepared to believe, not merely what cannot be proved, but what can be disproved from other beliefs that he also holds.
The problem of evil, in the sense in which I shall be using the phrase, is a problem only for someone who believes that there is a God who is both omnipotent and wholly good. And it is a logical problem, the problem of clarifying and reconciling a number of beliefs: it is not a scientific problem that might be solved by further observations, or a practical problem that might be solved by a decision or an action. These points are obvious; I mention them only because they are sometimes ignored by theologians, who sometimes parry a statement of the problem with such remarks as "Well, can you solve the problem yourself?" or "This is a mystery which may be revealed to us later" or "Evil is something to be faced and overcome, not to be merely discussed".
In its simplest form the problem is this: God is omnipotent; God is wholly good; and yet evil exists. There seems to be some contradiction between these three propositions, so that if any two of them were true the third would be false. But at the same time all three are essential parts of most theological positions: the theologian, it seems, at once must adhere and cannot consistently adhere to all three. (The problem does not arise only for theists, but I shall discuss it in the form in which it presents itself for ordinary theism.)
However, the contradiction does not arise immediately; to show it we need some additional premises, or perhaps some  quasi-logical rules connecting the terms 'good', 'evil', and 'omnipotent'. These additional principles are that good is opposed to evil, in such a way that a good thing always eliminates evil as far as it can, and that there are no limits to what an omnipotent thing can do. From these it follows that a good omnipotent thing eliminates evil completely, and then the propositions that a good omnipotent thing exists, and that evil exists, are incompatible.
A. Adequate Solutions
Now once the problem is fully stated it is clear that it can be solved, in the sense that the problem will not arise if one gives up at least one of the propositions that constitute it. If you are prepared to say that God is not wholly good, or not quite omnipotent, or that evil does not exist, or that good is not opposed to the kind of evil that exists, or that there are limits to what an omnipotent thing can do, then the problem of evil will not arise for you.
There are, then, quite a number of adequate solutions of the problem of evil, and some of these have been adopted, or almost adopted, by various thinkers. For example, a few have been prepared to deny God's omnipotence, and rather more have been prepared to keep the term 'omnipotence' but severely to restrict its meaning, recording quite a number of things that an omnipotent being cannot do. Some have said that evil is an illusion, perhaps because they held that the whole world of temporal, changing things is an illusion, and that what we call evil belongs only to this world, or perhaps because they held that although temporal things are much as we see them, those that we call evil are not really evil. Some have said that what we call evil is merely the privation of good, that evil in a positive sense, evil that would really be opposed to good, does not exist. Many have agreed with Pope that disorder is harmony not understood, and that partial evil is universal good. Whether any of these views is true is, of course, another question. But each of them gives an adequate solution of the problem of evil in the sense that if you accept it this problem does not arise for you, though you may, of course, have other problems to face.
But often enough these adequate solutions are only almost adopted. The thinkers who restrict God's power, but keep the term 'omnipotence', may reasonably be suspected of thinking, in other contexts, that his power is really unlimited. Those  who say that evil is an illusion may also be thinking, inconsistently, that this illusion is itself an evil. Those who say that "evil" is merely privation of good may also be thinking, inconsistently, that privation of good is an evil. (The fallacy here is akin to some forms of the " naturalistic fallacy " in ethics, where some think, for example, that " good" is just what contributes to evolutionary progress, and that evolutionary progress is itself good.) If Pope meant what he said in the first line of his couplet, that "disorder" is only harmony not understood, the "partial evil" of the second line must, for consistency, mean "that which, taken in isolation, falsely appears to be evil", but it would more naturally mean "that which, in isolation, really is evil". The second line, in fact, hesitates between two views, that "partial evil" isn't really evil, since only the universal quality is real, and that "partial evil" is really an evil, but only a little one.
In addition, therefore, to adequate solutions, we must recognise unsatisfactory inconsistent solutions, in which there is only a half-hearted or temporary rejection of one of the propositions which together constitute the problem. In these, one of the constituent propositions is explicitly rejected, but it is covertly re-asserted or assumed elsewhere in the system.
B. Fallacious Solutions
Besides these half-bearted solutions, which explicitly reject but implicitly assert one of the constituent propositions, there are definitely fallacious solutions which explicitly maintain all the constituent propositions, but implicitly reject at least one of them in the course of the argument that explains away the problem of evil.
There are, in fact, many so-called solutions which purport to remove the contradiction without abandoning any of its constituent propositions. These must be fallacious, as we can see from the very statement of the problem, but it is not so easy to see in each case precisely where the fallacy lies. I suggest that in all cases the fallacy has the general form suggested above: in order to solve the problem one (or perhaps more) of its constituent propositions is given up, but in such a way that it appears to have been retained, and can therefore be asserted without qualification in other contexts. Sometimes there is a further complication: the supposed solution moves to and fro between, say, two of the constituent propositions, at one point asserting the first of these but covertly abandoning the second,  at another point asserting the second but covertly abandoning the first. These fallacious solutions often turn upon some equivocation with the words 'good' and 'evil', or upon some vagueness about the way in which good and evil are opposed to one another, or about how much is meant by 'omnipotence'. I propose to examine some of these so-called solutions, and to exhibit their fallacies in detail. Incidentally, I shall also be considering whether an adequate solution could be reached by a minor modification of one or more of the constituent propositions, which would, however, still satisfy all the essential requirements of ordinary theism.
1. "Good cannot exist without evil" or "Evil is necessary as a counterpart to good."
It is sometimes suggested that evil is necessary as a counterpart to good, that if there were no evil there could be no good either, and that this solves the problem of evil. It is true that it points to an answer to the question "Why should there be evil?" But it does so only by qualifying some of the propositions that constitute the problem.
First, it sets a limit to what God can do, saying that God cannot create good without simultaneously creating evil, and this means either that God is not omnipotent or that there are some limits to what an omnipotent thing can do. It may be replied that these limits are always presupposed, that omnipotence has never meant the power to do what is logically impossible, and on the present view the existence of good without evil would be a logical impossibility. This interpretation of omnipotence may, indeed, be accepted as a modification of our original account which does not reject anything that is essential to theism, and I shall in general assume it in the subsequent discussion. It is, perhaps, the most common theistic view, but I think that some theists at least have maintained that God can do what is logically impossible. Many theists, at any rate, have held that logic itself is created or laid down by God, that logic is the way in which God arbitrarily chooses to think. (This is, of course, parallel to the ethical view that morally right actions are those which God arbitrarily chooses to command, and the two views encounter similar difficulties.) And this account of logic is clearly inconsistent with the view that God is bound by logical necessities -- unless it is possible for an omnipotent being to bind himself, an issue which we shall consider later, when we come to the Paradox of Omnipotence. This solution of the problem  of evil cannot, therefore, be consistently adopted along with the view that logic is itself created by God.
But, secondly, this solution denies that evil is opposed to good in our original sense. If good and evil are counterparts, a good thing will not "eliminate evil as far as it can". Indeed, this view suggests that good and evil are not strictly qualities of things at all. Perhaps the suggestion is that good and evil are related in much the same way as great and small. Certainly, when the term 'great' is used relatively as a condensation of 'greater than so-and-so', and 'small' is used correspondingly, greatness and smallness are counterparts and cannot exist without each other. But in this sense greatness is not a quality, not an intrinsic feature of anything; and it would be absurd to think of a movement in favour of greatness and against smallness in this sense. Such a movement would be self-defeating, since relative greatness can be promoted only by a simultaneous promotion of relative smallness. I feel sure that no theists would be content to regard God's goodness as analogous to this -- as if what he supports were not the good but the better, and as if he had the paradoxical aim that all things should be better than other things.
This point is obscured by the fact that 'great' and 'small' seem to have an absolute as well as a relative sense. I cannot discuss here whether there is absolute magnitude or not, but if there is, there could be an absolute sense for 'great', it could mean of at least a certain size, and it would make sense to speak of all things getting bigger, of a universe that was expanding all over, and therefore it would make sense to speak of promoting greatness. But in this sense great and small are not logically necessary counterparts: either quality could exist without the other. There would be no logical impossibility in everything's being small or in everything's being great.
Neither in the absolute nor in the relative sense, then, of 'great' and 'small' do these terms provide an analogy of the sort that would be needed to support this solution of the problem of evil. In neither case are greatness and smallness both necessary counterparts and mutually opposed forces or possible objects for support and attack.
It may be replied that good and evil are necessary counterparts in the same way as any quality and its logical opposite: redness can occur, it is suggested, only if non-redness also occurs. But unless evil is merely the privation of good, they are not logical opposites, and some farther argument would be needed to show that they are counterparts in the same way as genuine  logical opposites. Let us assume that this could be given. There is still doubt of the correctness of the metaphysical principle that a quality must have a real opposite: I suggest that it is not really impossible that everything should be, say, red, that the truth is merely that if everything were red we should not notice redness, and so we should have no word 'red'; we observe and give names to qualities only if they have real opposites. If so, the principle that a term must have an opposite would belong only to our language or to our thought, and would not be an ontological principle, and, correspondingly, the rule that good cannot exist without evil would not state a logical necessity of a sort that God would just have to put up with. God might have made everything good, though we should not have noticed it if he had.
But, finally, even if we concede that this is an ontological principle, it will provide a solution for the problem of evil only if one is prepared to say, "Evil exists, but only just enough evil to serve as the counterpart of good". I doubt whether any theist will accept this. After all, the ontological requirement that non-redness should occur would be satisfied even if all the universe, except for a minute speck, were red, and, if there were a corresponding requirement for evil as a counterpart to good, a minute dose of evil would presumably do. But theists are not usually willing to say, in all contexts, that all the evil that occurs is a minute and necessary dose.
2. "Evil is necessary as a means to good."
It is sometimes suggested that evil is necessary for good not as a counterpart but as a means. In its simple form this has little plausibility as a solution of the problem of evil, since it obviously implies a severe restriction of God's power. It would be a causal law that you cannot have a certain end without a certain means, so that if God has to introduce evil as a means to good, he must be subject to at least some causal laws. This certainly conflicts with what a theist normally means by omnipotence. This view of God as limited by causal laws also conflicts with the view that causal laws are themselves made by God, which is more widely held than the corresponding view about the laws of logic. This conflict would, indeed be resolved if it were possible for an omnipotent being to bind himself, and this possibility has still to be considered. Unless a favourable answer can be given to this question, the suggestion that evil is necessary as a means to good solves the problem of evil only by  denying one of its constituent propositions, either that God is omnipotent or that 'omnipotent' means what it says.
3. "The universe is better with some evil in it than it could be if there were no evil."
Much more important is a solution which at first seems to be a mere variant of the previous one, that evil may contribute to the goodness of a whole in which it is found, so that the universe as a whole is better as it is, with some evil in it, than it would be if there were no evil. This solution may be developed in either of two ways. It may be supported by an aesthetic analogy, by the fact that contrasts heighten beauty, that in a musical work, for example, there may occur discords which somehow add to the beauty of the work as a whole. Alternatively, it may be worked out in connexion with the notion of progress, that the best possible organisation of the universe will not be static, but progressive, that the gradual overcoming of evil by good is really a finer thing than would be the eternal unchallenged supremacy of good.
In either case, this solution usually starts from the assumption that the evil whose existence gives rise to the problem of evil is primarily what is called physical evil, that is to say, pain. In Hume's rather half-hearted presentation of the problem of evil, the evils that he stresses are pain and disease, and those who reply to him argue that the existence of pain and disease makes possible the existence of sympathy, benevolence, heroism, and the gradually successful struggle of doctors and reformers to overcome these evils. In fact, theists often seize the opportunity to accuse those who stress the problem of evil of taking a low, materialistic view of good and evil, equating these with pleasure and pain, and of ignoring the more spiritual goods which can arise in the struggle against evils.
But let us see exactly what is being done here. Let us call pain and misery 'first order evil' or 'evil (1)'. What contrasts with this, namely, pleasure and happiness, will be called 'first order good' or 'good (1)'. Distinct from this is 'second order good' or 'good (2)' which somehow emerges in a complex situation in which evil (1) is a necessary component -- logically, not merely causally, necessary. (Exactly how it emerges does not matter: in the crudest version of this solution good (2) is simply the heightening of happiness by the contrast with misery, in other versions it includes sympathy with suffering, heroism in facing danger, and the gradual decrease of first order evil afld increase of first order good.) It is also being assumed that  second order good is more important than first order good or evil, in particular that it more than outweighs the first order evil it involves.
Now this is a particularly subtle attempt to solve the problem of evil. It defends God's goodness and omnipotence on the ground that (on a sufficiently long view) this, is the best of all logically possible worlds, because it includes the important second order goods, and yet it admits that real evils, namely first order evils, exist. But does it still hold that good and evil are opposed? Not, clearly, in the sense that we set out originally: good does not tend to ehminate evil in general. Instead, we have a modified, a more complex pattern. First order good (e.g. happiness) contrasts with first order evil (e.g. misery): these two are opposed in a fairly mechanical way; some second order goods (e.g. benevolence) try to maximise first order good and minimise first order evil; but God's goodness is not this, it is rather the will to maximise second order good. We might, therefore, call God's goodness an example of a third order goodness, or good (3). While this account is different from our original one, it might well be held to be an improvement on it, to give a more accurate description of the way in which good is opposed to evil, and to be consistent with the essential theist position.
There might, however, be several objections to this solution.
First, some might argue that such qualities as benevolence -- and a fortiori the third order goodness which promotes benevolence -- have a merely derivative value, that they are not higher sorts of good, but merely means to good (1), that is, to happiness, so that it would be absurd for God to keep misery in existence in order to make possible the virtues of benevolence, heroism, etc. The theist who adopts the present solution must, of course, deny this, but he can do so with some plausibility, so I should not press this objection.
Secondly, it follows from this solution that God is not in our sense benevolent or sympathetic: he is not concerned to minimise evil (1), but only to promote good (2); and this might be a disturbing conclusion for some theists.
But, thirdly, the fatal objection is this. Our analysis shows clearly the possibility of the existence of a second order evil, an evil (2) contrasting with good (2) as evil (1) contrasts with good (1). This would include malevolence, cruelty, callousness, cowardice, and states in which good (1) is decreasing and evil (1) increasing. And just as good (2) is held to be the important kind of good, the kind that God is concerned to promote, so evil (2) will, by analogy, be the important kind of evil, the kind  which God, if he were wholly good and omnipotent, would eliminate. And yet evil (2) plainly exists, and indeed most theists (in other contexts) stress its existence more than that of evil (1). We should, therefore, state the problem of evil in terms of second order evil, and against this form of the problem the present solution is useless.
An attempt might be made to use this solution again, at a higher level, to explain the occurrence of evil (2): indeed the next main solution that we shall examine does just this, with the help of some new notions. Without any fresh notions, such a solution would have little plausibility: for example, we could hardly say that the really important good was a good (3), such as the increase of benevolence in proportion to cruelty, which logically required for its occurrence the occurrence of some second order evil. But even if evil (2) could be explained in this way, it is fairly clear that there would be third order evils contrasting with this third order good: and we should be well on the way to an infinite regress, where the solution of a problem of evil, stated in terms of evil (n), indicated the existence of an evil (n + 1), and a farther problem to be solved.
4. "Evil is due to human freewill."
Perhaps the most important proposed solution of the problem of evil is that evil is not to be ascribed to God at all, but to the independent actions of human beings, supposed to have been endowed by God with freedom of the will. This solution may be combined with the preceding one: first order evil (e.g. pain) may be justified as a logically necessary component in second order good (e.g. sympathy) while second order evil (e.g. cruelty) is not justified, but is so ascribed to human beings that God cannot be held responsible for it. This combination evades my third criticism of the preceding solution.
The freewill solution also involves the preceding solution at a higher level. To explain why a wholly good God gave men freewill although it would lead to some important evils, it must be argued that it is better on the whole that men should act freely, and sometimes err, than that they should be innocent automata, acting rightly in a wholly determined way. Freedom, that is to say, is now treated as a third order good, and as being more valuable than second order goods (such as sympathy and heroism) would be if they were deterministically produced, and it is being assumed that second order evils, such as cruelty, are logically necessary accompaniments of freedom, just as pain is a logically necessary pre-condition of sympathy. 
I think that this solution is unsatisfactory primarily because of the incoherence of the notion of freedom of the will: but I cannot discuss this topic adequately here, although some of my criticisms will touch upon it.
First I should query the assumption that second order evils are logically necessary accompaniments of freedom. I should ask this: if God has made men such that in their free choices they sometimes prefer what is good and sometimes what is evil, why could he not have made men such that they always freely choose the good? If there is no logical impossibility in a man's freely choosing the good on one, or on several, occasions, there cannot be a logical impossibility in his freely choosing the good on every occasion. God was not, then, faced with a choice between making innocent automata and making beings who, in acting freely, would sometimes go wrong: there was open to him the obviously better possibility of making beings who would act freely but always go right. Clearly, his failure to avail himself of this possibility is inconsistent with his being both omnipotent and wholly good.
If it is replied that this objection is absurd, that the making of some wrong choices is logically necessary for freedom, it would seem that 'freedom' must here mean complete randomness or indeterminacy, including randomness with regard to the alternatives good and evil, in other words that men's choices and consequent actions can be "free " only if they are not determined by their characters. Only on this assumption can God escape the responsibility for men's actions; for if he made them as they are, but did not determine their wrong choices, this can only be because the wrong choices are not determined by men as they are. But then if freedom is randomness, how can it be a characteristic of will? And, still more, how can it be the most important good? What value or merit would there be in free choices if these were random actions which were not determined by the nature of the agent?
I conclude that to make this solution plausible two different senses of 'freedom' must be confused, one sense which will justify the view that freedom is a third order good, more valuable than other goods would be without it, and another sense, sheer randomness, to prevent us from ascribing to God a decision to make men such that they sometimes go wrong when he might have made them such that they would always freely go right.
This criticism is sufficient to dispose of this solution. But besides this there is a fundamental difficulty in the notion of an omnipotent God creating men with free will, for if men's wills  are really free this must mean that even God cannot control them, that is, that God is no longer omnipotent. It may be objected that God's gift of freedom to men does not mean that he cannot control their wills, but that he always refrains from controlling their wills. But why, we may ask, should God refrain from controlling evil wills? Why should he not leave men free to will rightly, but intervene when he sees them beginning to will wrongly? If God could do this, but does not, and if he is wholly good, the only explanation could be that even a wrong free act of will is not really evil, that its freedom is a value which outweighs its wrongness, so that there would be a loss of value if God took away the wrongness and the freedom together. But this is utterly opposed to what theists say about sin in other contexts. The present solution of the problem of evil, then, can be maintained only in the form that God has made men so free that he cannot control their wills.
This leads us to what I call the Paradox of Omnipotence: can an omnipotent being make things which he cannot subsequently control? Or, what is practically equivalent to this, can an omnipotent being make rules which then bind himself? (These are practically equivalent because any such rules could be regarded as setting certain things beyond his control, and vice versa.) The second of these formulations is relevant to the suggestions that we have already met, that an omnipotent God creates the rules of logic or causal laws, and is then bound by them.
It is clear that this is a paradox: the questions cannot be answered satisfactorily either in the affirmative or in the negative. If we answer "Yes", it follows that if God actually makes things which he cannot control, or makes rules which bind himself, he is not omnipotent once he has made them: there are then things which he cannot do. But if we answer "No", we are immediately asserting that there are things which he cannot do, that is to say that he is already not omnipotent.
It cannot be replied that the question which sets this paradox is not a proper question. It would make perfectly good sense to say that a human mechanic has made a machine which he cannot control: if there is any difficulty about the question it lies in the notion of omnipotence itself.
This, incidentally, shows that although we have approached this paradox from the free will theory, it is equally a problem for a theological determinist. No one thinks that machines have free will, yet they may well be beyond the control of their  makers. The determinist might reply that anyone who makes anything determines its ways of acting, and so determines its subsequent behaviour : even the human mechanic does this by his choice of materials and structure for his machine, though he does not know all about either of these : the mechanic thus determines, though he may not foresee, his machine's actions. And since God is omniscient, and since his creation of things is total, he both determines and foresees the ways in which his creatures will act. We may grant this, but it is beside the point. The question is not whether God originally determined the future actions of his creatures, but whether he can subsequently control their actions, or whether he was able in his original creation to put things beyond his subsequent control. Even on determinist principles the answers " Yes" and " No" are equally irreconcilable with God's omnipotence.
Before suggesting a solution of this paradox, I would point out that there is a parallel Paradox of Sovereignty. Can a legal sovereign make a law restricting its own future legislative power? For example, could the British parliament make a law forbidding any future parliament to socialise banking, and also forbidding the future repeal of this law itself? Or could the British parliament, which was legally sovereign in Australia in, say, 1899, pass a valid law, or series of laws, which made it no longer sovereign in 1933? Again, neither the affirmative nor the negative answer is really satisfactory. If we were to answer "Yes", we should be admitting the validity of a law which, if it were actually made, would mean that parliament was no longer sovereign. If we were to answer "No", we should be admitting that there is a law, not logically absurd, which parliament cannot validly make, that is, that parliament is not now a legal sovereign. This paradox can be solved in the following way. We should distinguish between first order laws, that is laws governing the actions of individuals and bodies other than the legislature, and second order laws, that is laws about laws, laws governing the actions of the legislature itself. Correspondingly, we should distinguish two orders of sovereignty, first order sovereignty (sovereignty (1)) which is unlimited authority to make first order laws, and second order sovereignty (sovereignty (2)) which is unlimited authority to make second order laws. If we say that parliament is sovereign we might mean that any parliament at any time has sovereignty (1), or we might mean that parliament has both sovereignty (1) and sovereignty (2) at present, but we cannot without contradiction mean both that the present parliament has sovereignty (2) and  that every parliament at every time has sovereignty (1), for if the present parliament has sovereignty (2) it may use it to take away the sovereignty (1) of later parliaments. What the paradox shows is that we cannot ascribe to any continuing institution legal sovereignty in an inclusive sense.
The analogy between omnipotence and sovereignty shows that the paradox of omnipotence can be solved in a similar way. We must distinguish between first order omnipotence (omnipotence (1)), that is unlimited power to act, and second order omnipotence (omnipotence (2)), that is unlimited power to determine what powers to act things shall have. Then we could consistently say that God all the time has omnipotence (1), but if so no beings at any time have powers to act independently of God. Or we could say that God at one time had omnipotence (2), and used it to assign independent powers to act to certain things, so that God thereafter did not have omnipotence (1). But what the paradox shows is that we cannot consistently ascribe to any continuing being omnipotence in an inclusive sense. An alternative solution of this paradox would be simply to deny that God is a continuing being, that any times can be assigned to his actions at all. But on this assumption (which also has difficulties of its own) no meaning can be given to the assertion that God made men with wills so free that he could not control them. The paradox of omnipotence can be avoided by putting God outside time, but the freewill solution of the problem of evil cannot be saved in this way, and equally it remains impossible to hold that an omnipotent God binds himself by causal or logical laws.
Of the proposed solutions of the problem of evil which we have examined, none has stood up to criticism. There may be other solutions which require examination, but this study strongly suggests that there is no valid solution of the problem which does not modify at least one of the constituent propositions in a way which would seriously affect the essential core of the theistic position.
Quite apart from the problem of evil, the paradox of omnipotence has shown that God's omnipotence must in any case be restricted in one way or another, that unqualified omnipotence cannot be ascribed to any being that continues through time. And if God and his actions are not in time, can omnipotence, or power of any sort, be meaningfully ascribed to him ?