The No-Best-World Problem

The difficulties of the last four rungs all concerned particular features of the world: this suffering, that silence. The one taken up here concerns the act of creation itself, and it would remain even if every question about suffering had been answered. It asks what an omnipotent and perfectly good being is supposed to do when choosing what to make, and it turns out that both available answers cost something the theist wants to keep.

Leibniz's assumption

Leibniz's conclusion attracted ridicule almost immediately, and Voltaire's Candide fixed the caricature in place: the philosopher who insists, amid catastrophe, that everything is for the best. The caricature is unfair to what Leibniz actually claimed, which was not that each event is good but that the whole is the best available whole, and that goods and evils are so entangled that removing any of the latter would remove more of the former.

But the interesting problem is not whether Leibniz was right that this is the best. It is whether the framework he was working in makes sense at all. He assumed there is a best world to be chosen. Suppose there is not.

Why there may be no best

The reason for thinking the series is unbounded is disarmingly simple. Take any world you like. Now consider a world just like it except that it also contains one more person living a life of great value, or one more moment of joy, or one additional beautiful thing that costs nothing. That world appears better. And the operation can be repeated without limit, since there is no obvious point at which adding a further good becomes impossible for an omnipotent being.

If that is right, then asking which world is best is like asking which number is greatest. The question has a grammatical form that suggests an answer and no answer to give.

Two ways of resisting the reasoning are worth setting out before the dilemma is drawn, because if either works the problem does not arise.

The first denies that value is additive. Rung 16 introduced the idea of organic unities, on which the value of a whole is not the sum of the values of its parts, and if that is right in general then adding a good to a world does not automatically improve the world. A well-made thing can be spoiled by an addition that is good considered on its own. The trouble is that the argument needs only one operation that reliably improves, not that every addition improves, and the friend of the argument can simply pick a case where the addition is clearly an enhancement rather than an intrusion.

The second denies that worlds have finite value. If God exists in a world, and God's own value is infinite, then every world containing God has infinite value, and no such world is better than any other by the measure being used. This does dissolve the comparison, and something like it has appealed to theists in the tradition who held that the creature adds nothing to the divine perfection. It also proves more than intended. If all worlds containing God are equal in value, then a world of universal torment is no worse than this one, and the argument from evil is answered by a device that no theist should want to use.

Neither escape is obviously successful, and the dilemma is usually allowed to proceed.

The consequence for theism is not that God fails to exist but that a dilemma opens.

The dilemma was pressed with the most care by William Rowe in Can God Be Free? (Oxford University Press, 2004), the last book he published on the philosophy of religion, and it deserves its reputation as one of the cleanest arguments in the field.

The first horn

Suppose there is a best world. Then a perfectly good being with the power to create it and full knowledge of its superiority cannot do otherwise, since doing otherwise would be choosing a worse option in full knowledge, which is what a perfectly good being cannot do.

Three consequences follow, and each conflicts with something most theists affirm.

Creation becomes necessary. If God exists necessarily, has these attributes necessarily, and must create the best, then this world exists necessarily too, and there are no other possible worlds in the relevant sense. Everything that happens happens of necessity, including every evil in it. That is a large metaphysical price, and it undercuts the free will defence, which relies on there being alternatives God could have brought about.

Creation ceases to be a gift. Most theistic traditions treat the existence of the world, and of each creature in it, as something freely given rather than extracted from God by the requirements of goodness. On this horn nothing was given, because nothing else could have happened.

And praise becomes inapt in the way praise is inapt for the unavoidable. We do not thank people for what they could not have avoided doing. Rowe's version of the point is that a being who creates the best of necessity is not thereby praiseworthy for creating it, and if God is not praiseworthy for creating us, a central element of religious practice loses its object.

The theist can accept all this, and some do. The cost is high, and the position is close to the necessitarianism that the tradition has usually resisted.

There is a more moderate way of taking the horn, and it turns on distinguishing two things the word "freely" can mean. On one understanding, acting freely requires that one could have done otherwise. On another, it requires only that one's action flow from one's own nature and settled character rather than from external compulsion. The second understanding is the one on which we call a person of thoroughly good character free when they act well, even if we think it was never really in doubt what they would do, and it has a long history in discussions of God's freedom, where acting from the divine nature has often been taken to be the highest kind of freedom rather than the absence of it.

Praiseworthiness can be handled the same way. Harry Frankfurt's "Alternate Possibilities and Moral Responsibility" in The Journal of Philosophy in 1969 argued that responsibility does not in general require the ability to do otherwise, since an agent who acts on their own reasons is responsible for doing so even if some mechanism would have prevented any other outcome. If that generalises, then the observation that God could not have created a worse world does not by itself show that God is not praiseworthy for creating this one.

What this reply cannot rescue is the modal cost. Even if freedom and praise survive, the claim that this world is the only one that could have existed remains, and it takes the free will defence down with it.

The second horn

Suppose instead that there is no best. Then for any world God creates, there is a better one God could have created.

The difficulty is a comparative claim about God. For any world God makes, one can describe a being who would have made a better one, and that being would seem to be better than God, or at least to have acted better. This threatens perfect being theology directly, since the whole point of that approach is that no possible being is greater than God.

The standard response, and the one that has organised the literature since 1972, was given by Robert Merrihew Adams in "Must God Create the Best?" in The Philosophical Review.

Adams's claim was that the assumption generating the difficulty is false: a perfectly good being is not required to maximise. What goodness requires is that one wrong no one and treat one's creatures well, and neither of these is violated by making a world that is not the best available. Nobody in the created world has been wronged. Each of them has a life worth living. And, on Adams's account, no one who does not exist has a claim to be created, so the merely possible people in the better world have no grievance.

There is a further point in the vicinity, and it is worth separating out because it does a lot of work. The people in a better world would in most cases not be these people. Different creative choices bring different individuals into being, so the person who complains that God should have made a better world is generally complaining that God should have made a world without them in it. That does not look like a complaint they have standing to make.

What satisficing costs

The satisficing reply is not free, and three objections recur.

The threshold looks arbitrary. Any line drawn between worlds good enough to create and worlds not good enough invites the question of why it falls there, and no principled answer has been produced. The theist may respond that arbitrariness is precisely what freedom in creation requires, since a fully determined choice would land back on the first horn, but the arbitrariness has to be located in the choosing rather than in the standard if the standard is to do any work.

The second objection connects this rung to the argument from evil, and it is the reason the topic matters beyond the philosophy of creation. The evidential argument depends on the principle that a good being would prevent a suffering it could prevent without losing a greater good. A satisficing God is not bound by that principle. Such a being may permit an evil that could have been prevented at no cost, provided the resulting world remains above the threshold, and any evil so permitted is gratuitous in exactly the sense the argument uses. So satisficing supplies a route to blocking Rowe's argument that does not go through the sceptical reply at all.

Whether this is a gain for the theist is not obvious. It answers the argument from evil by conceding that God permits gratuitous evils, which many theists find a worse result than the argument itself, and which is difficult to square with the language of divine goodness in the traditions. The theist has traded a problem in the theory of value for a problem in devotion.

The third objection is that satisficing sits badly with how we assess agents in ordinary cases. We do not usually think an agent is beyond criticism for choosing a worse option merely because the chosen option was above some threshold, particularly when the agent is under no constraint of cost or effort. Omnipotence removes every consideration that makes satisficing reasonable for us: limited time, limited resources, the need to stop deliberating and act.

Creating everything worth creating

A third position tries to have both horns' benefits, and it has attracted serious attention over the last two decades.

The move is elegant. The no-best-world problem arises because for any candidate there is a better, and the reply is to stop treating candidates as competitors. If every universe above the threshold is created, then no universe has been passed over, no possible person has been denied existence for the sake of a rival, and the question of why God chose this one rather than a better one does not arise, because God did not choose.

It also changes what the problem of evil is asking. On this proposal our universe is not God's best effort but one of many, and the question of why it is not better becomes the question of why it exists at all, which is answered by its being worth creating.

The costs are substantial.

The first is that the threshold problem returns unchanged. Which universes are worth creating? The multiverse proposal needs a line just as much as satisficing does, and it is the same line.

The second is that the proposal does not obviously help with horrendous evils. A person destroyed in this universe is not consoled by the existence of others, and the objection of rung 16 applies with full force: the reply operates at the level of totals, and the complaint was about individuals.

The third is dialectical, and theists who defend the fine-tuning argument should notice it. That argument's standard reply to the naturalistic multiverse is that positing countless unobserved universes to explain the calibration of this one is extravagant and unmotivated. A theist who then posits countless universes to solve the no-best-world problem has adopted the same extravagance and can no longer object to it on those grounds. The two multiverses are motivated differently, and the theist may argue that a divinely created plurality is better grounded than a brute one, but the easy version of the objection is no longer available.

The fourth is metaphysical. The proposal requires a workable account of what individuates a universe, of whether the collection of all worth-creating universes is itself a single possible world, and of whether the value of that collection is well defined. These are live technical questions rather than obvious defects, but they mean the position is a research programme rather than a settled answer.

The principle has an ancient pedigree in the idea that goodness spreads itself, and the multiverse proposals are recognisably its modern descendants. It also has an old difficulty, which Lovejoy documented: taken strictly, it removes divine freedom altogether, since a God who must realise every possibility has no choices to make. The theist who uses plenitude to solve the no-best-world problem may find that it has solved the problem by conceding the first horn.

Whether the problem is peculiar to theism

It is worth noticing what the difficulty depends on, because a reader who is not a theist may wonder why any of it matters.

The dilemma is generated by three commitments taken together: that God is unlimited in power, that God is perfectly good, and that God's creating was a choice. Drop any one and it dissolves. A limited creator has reasons to stop somewhere. A creator who is not perfectly good is not held to the standard. A world that emerges without being chosen is not answerable to a comparison with alternatives.

That is a general feature of this tier's problems and it is worth stating plainly. The difficulties are not accidents of theology; they are the price of the specific combination of attributes that classical theism assembled, and each of them can be avoided by weakening one of the attributes. Rungs 19 and 21 take up what happens when the target is changed in exactly that way.

What the problem establishes

Three things can be said with reasonable confidence.

The dilemma is genuine and has not been dissolved. Every available response gives something up: necessitarianism about creation, unsurpassability, the well-ordering of value, or divine freedom. Which loss is acceptable is now the substantive question, and there is no consensus.

The satisficing reply is more consequential than it looks, because of its interaction with the argument from evil. A theist who adopts it has an answer to Rowe that does not require the sceptical apparatus of rung 15, and has thereby accepted that some evils are permitted for no compensating good. Very few theists want that combination once it is stated openly, which is one reason the position is discussed more than it is held.

And the problem has clarified what perfect being theology is committed to. If greatness is to be maximised along every dimension, and some dimensions have no maximum, then the method needs an account of what it means to be greatest in an unbounded respect. That question was raised much earlier, at the ontological argument, and it has never been fully answered.

Next

Every rung since the beginning of this tier has asked whether God exists or whether some feature of the world tells against it. The next rung sets that question aside and asks a different one: supposing we do not know whether God exists, would it be better if God did? The answers are less predictable than one might expect, since some atheists hold that a world with God would be better and some theists hold that it would in certain respects be worse.