Philosophy of Religion 35: 65-88, 1994. © 1994 KluwerAcademic Publishers. Printedin the Netherlands.

Divine freedom and the choice of a world

EVAN M. FALES

Department of Philosophy, The University of lowa, lowa City, USA

1. A problem about God's freedom In his book Anselmian Explorations, Thomas Morris raises the question how God's freedom of will can be reconciled with His essential righteousness. 1 He provides an interesting and attractive solution to this problem, one which provides for divine freedom with respect to divine acts of supererogation. I shall argue that Morris' solution becomes considerably less attractive, however, if, as I hope to show, there is a maximally good world that God can create. In view of this result, I shall defend a different conception of divine freedom, according to which that freedom is compatible with God's acting as He does necessarily. That way of answering the difficulty posed by Morris will force a confrontation between two alternative ways of formulating the libertarian conception of free will, alternatives which are sometimes taken to be equivalent: first, that freedom involves being able to do otherwise, and second, that freedom involves the capacity to act from reasons one has. 2 Morris regards the first to be fundamental; I take the second to be so. In the case of God, the two conceptions of freedom come apart; so God provides a possibly unique case against which to test the two conceptions. If I am right, and the second conception is the more fundamental one then perhaps an analysis of what it is for God to be free can help us to see why. This second sort of account of God's free will, and of free will in general, has a long history. Versions of it can be found in Plato, Augustine, Anselm, Leibniz, and Hegel, among others. I shall have some things to say in the course of my remarks about the ontological commitments I believe such a conception requires. Morris, on the other hand, thinks that it is an essential and minimal part of any libertarian conception of freedom that an act is free only if the agent could have avoided performing it. Morris calls this condition the Principle of Avoidance (PA), and he formulates it thus:

66 (PA) An agent S performs an act A at a time t freely only if no conditions exist prior to t which render it necessary, or unavoidable, in a broadly logical sense, and by doing so in fact bring it about, that S performs A.3 The 'broadly logical' necessity to which PA refers includes both formal (narrowly logical) necessity and what Kripke calls metaphysical necessity (de re necessity). This principle is, as Morris proceeds to show, incompatible with the view that God's goodness consists - at least in part - in the fulfilment by Him of moral duties, when this is taken together with the view that God is necessarily good, and the view that freedom in the libertarian sense defined by PA is a necessary precondition for moral agency. The incompatibility arises because, if God's goodness is an essential attribute of His, and if goodness requires (at least) the fulfilment of moral duties - such as telling the truth and keeping promises - then it is a de re necessary truth about God that He performs His duties. 4 So there is something about God's nature which, of necessity, prevents Him from failing to act so as to fulfill those duties. This, however, means that PA is violated by God when He acts out of duty - and hence, that these actions of His are not free. If they are not free, however, then God is not, with respect to them, acting as a moral agent at all, and hence not out of moral goodness. Thus one of the foregoing claims must be given up. But theists are standardly committed to a libertarian conception of moral agency, to God's essential goodness, and to the view that moral goodness consists, at least in part, in the performance of duties.

2. Morris' solution Morris' way out of the difficulty is to give up the last of these commitments. He holds that God's moral agency does not consist, even in part, in the performance of duties. To make this solution palatable, Morris makes two points. First, he reminds us of the distinction between following a rule and acting in accordance with a rule. The former involves intending to obey the rule, and acting from or with that intention; the latter does not. I may act in accordance with a rule without following it if I act this way accidentally - perhaps in complete ignorance of the rule - or even necessarily, if (for example) I am mechanically caused so to act, but lack the requisite intention. According to Morris, God acts of necessity in accordance with those moral rules that generate duties (because of His necessary goodness); but God is not bound by any moral duties: he is not following

67 the rules. Thus we can rely upon God to act in these respects as if he were perfectly obedient to the moral law - we can still use the duty model to predict God's actions in suitable circumstances - but He is in these cases acting neither freely nor (therefore) morally. The second point Morris makes to render his solution palatable is to argue that there is, nevertheless, a broad sphere of activity for God with respect to which His actions are genuinely free. For there are many actions God can and does perform supererogatively - acts whose omission would not involve the violation of any rule of duty on God's part. In this sphere, God's goodness manifests itself in activity which is perfectly free and morally perfect. Clearly, Morris' solution is in several ways an attractive one. It preserves the libertarian account of moral agency, and it preserves for God a broad and genuine sphere in which His moral agency can be exercised in a manner deserving our praise. At the same time, it hews to the traditional view that God is not subject to any moral duties, thus preserving His moral sovereignty. Finally, and certainly not least, it preserves the claim that God is necessarily good. Nevertheless, there are several features of Morris' solution that should give a theist pause. Of these, I shall discuss four: (1) Morris' solution entails that God cannot act freely whenever He has a sufficient reason to act. It therefore loosens the conceptual connection between freedom and rational agency in a problematic way; (2) It limits the sphere of God's moral activity to supererogatory actions; (3) With respect to duty-generating rules, it makes God out to be, in moral terms, a kind of automaton, and (4) its attractiveness depends in large measure upon the claim that there are supererogatory actions that God can perform; but that claim is problematic. In contradistinction to Morris, I will argue that a theist is better off solving the problem he poses by rejecting PA, the principle that a free agent can do otherwise. Of course, PA reflects deep libertarian intuitions about free agency. One constraint on a theistically acceptable rejection of PA is that these intuitions - and with them, libertarianism - be preserved. Thus an acceptable alternative to PA will have to enable us to explain the deep plausibility that it has, and to preserve the key intuitions underlying it. Before I proceed to this alternative, I want to motivate it by discussing some of the difficulties Morris' solution faces. Here I shall discuss the second, third, and fourth of the problems just mentioned. The nature of the first problem will emerge in the presentation of the alternative solution. God's moral activity is, according to Morris, restricted to supererogatory actions. Actually, the term 'supererogatory', which Morris uses

68 in this context, is misleading. It suggests that God has duties which He sometimes goes beyond. But on Morris' view, God does not literally have any duties at all. He never (indeed, necessarily never) lies or breaks promises, but He has no duty not to do these things. Morris might more accurately call God's free acts non-necessitated acts, or optional acts. I shall call them optional acts when discussing his solution. Many theists, in any event, would find it disturbing to deny that God, qua moral agent, ever acts out of any sense of duty or obligation toward His creatures. That discomfort can be reinforced by reflecting upon the sense in which, if God merely acts in accordance with duty-generating rules, rather than following them, He becomes a kind of deontic automaton. On this view, God acts necessarily in accordance with every such rule, but nevertheless lacks the intention to follow the rule. He knows, o f course, what the rule is, and that what He does in no way violates it. But He does not act out of a sense of duty - that is, He does not perform the action because of the recognition that it is the right thing to do. This aspect of Morris' solution should make theist uneasy. It might seem to be adequate compensation for this uneasiness that God is not, on Morris' view, bound by moral duties, and so is free from external moral constraints. But one can (and many Christian theists have) accepted the view that God acts from a sense of duty, and have held also that He is free from external moral constraints: for they have held that, in acting dutifully, GOd is acting in accordance with His own nature.

3. The existence of a best possible world Another problem with Morris' solution is that its attractiveness largely evaporates if there are for God no morally optional actions. If that is so, then there is no sphere in which God can exercise His goodness as a moral agent; in short, God will be good, but He will not be morally good except perhaps in the analogical sense that He behaves as if He were a morally perfect agent. Some philosophers - Leibniz perhaps most famously - have held that none of God' s actions are optional. I shall not argue for a view as strong as Leibniz's, but I shall argue for a view that is sufficiently strong to make Morris' solution unattractive. Unlike Leibniz, I do not find it particularly plausible to suppose that God could never be faced with two equally attractive choices. There may be alternatives confronting God with respect to which He knows that there is no balance of reasons favoring any one over the others. In that sort of case, God, unlike Buridan's ass, will manage to make a choice. But though optional, the choice will also be

69 arbitrary. At the same time, the very recognition of optionality in such circumstances is parasitic upon the practice of choosing in accordance with the balance of reasons. But more about this anon. Such special cases do not, I believe, present God with a sphere for free moral action that is adequate to our conception of Him as a perfect moral agent, for that sphere is both limited and characterized by the fact that choices within it are arbitrary. To a much greater degree, the plausibility of Morris' proposal hinges on the question whether there is some best world that God can create. If there is a best possible world, then God necessarily, in virtue of His nature, creates that world. 5 If there are several equally good, maximally good possible worlds, then God has the option to create any one of them. But in neither case will there be room for what Morris means by 'supererogatory' activity on God's part. Such activity is conceivable only if two conditions are fulfilled: (a) there is a worst possible world (or worlds) which is such that any world worse than it would be a world in which (per impossible) God neglects, or acts in violation of, duty-generating rules; and (b) there is no best possible world. Condition (a) is required, I think, because if just any kind of world were morally adequate from God's point of view, we would lose our conceptual grip on what God's moral goodness consists in. Condition (b)is required because, if there is a best possible world, then God cannot in good conscience refrain from creating it. If, however, possible worlds increase in goodness without limit, or are incommensurable with respect to their degree of goodness, then it is logically impossible - hence not obligatory or in any sense necessary - for God to choose a best world to create. In that case the creation of any world, if it is at all better than a worst possible world, will involve 'supererogation' (in Morris' sense) on God's part. Unsurprisingly, Morris is committed to conditions (a) and (b). 6 In this, he is not alone. Aquinas accepted (a) and (b), and in recent years, this view has been accepted as an important part of a solution to the problem of evil by Schlesinger7 and Kretzmann, 8 among others. I shall examine their arguments for condition (b), 9 suggest where these arguments fall short, and then explore what a best possible world might be like. Kretzmann's 'A Particular Problem of Creation' provides a useful and compact treatment of Aquinas' view on (b), with which Kretzmann appears to be in agreement. So I shall begin by following Kretzmann's exposition, then move on to an examination of Schlesinger's argument. According to Aquinas, there are at least four distinct ways in which God could make a better world than any world He does make. (Aquinas argues that God could have made a world better than the one He in fact made, but the argument is clearly supposed to generalize: had God made

70 any other world instead, there are always better worlds yet God could have chosen to create.) In general, there are for Aquinas three features of a world which determine its total goodness: what its parts are, the degree of intrinsic goodness achieved by each of these parts, and the order of the parts (i.e., their relations). Thus, a world can be made better by adding more parts to it, by increasing the goodness of one or more parts, or by improving the order of the parts. 1° Given this, a better world can be made by GOd by: (1) 'the addition of more parts in such a way that many other species would be created and the many grades of goodness there can be would be increased; for between even the highest creature and GOd there is an infinte gap', 12 (2) making every one of the world's parts a better instance of its kind; (3) making any part a better example of its kind, so enabling it better to fulfill its role in the cosmic order; or (4) by making a world consisting of entirely different parts-cum-order. Briefly, then, God can improve a world by adding good things to it, by improving particulars as examples of their kind, or by making an entirely different sort of world. The essential feature, for our purposes, of all of these world-improvement techniques is that there is supposed to be no limit to the value that can be added by means of them. If God had made our world better by having made additional parts, he could have made it still better by adding even more parts; if he had made a better world by making one containing different parts differently ordered, there are infinitely many better worlds still that he could have made in this way. So there is no optimal world that God is morally committed to make. Why does Aquinas suppose this is so? A clue is provided by his claim that 'between even the highest creature and God there is an infinite gap'. Aquinas believes that, if we rank beings on a scale according to their goodness, GOd will uniquely possess an infinite amount of it, while any creature must possess only a finite amount. Thus the goodness-gap between creator and creature is infinitely wide, no matter how good the creature. But since God (being omnipotent) can create creatures possessing any finite amount of goodness, and can create any finite number of creatures, there is no upper limit to the amount of goodness God can install in the universe. It is clear that Aquinas' line of reasoning rests on two premises, viz.: (1) (2)

It is logically impossible for God to create a world as good as, or at most finitely less good than, God Himself, and For any finite degree of goodness, God can make a world with finitely more goodness. 12

71 Supposing that one can sensibly speak of linearly ordering worlds in terms of the amount of goodness they express, (1) and (2) yield the conclusion that there is no f'mite upper bound to the series of worlds of increasing goodness creatible by God. Aquinas' reason for believing (2) is that God is omnipotent; his reasons for believing (1) are bound up with his equation of goodness with being, and his Aristotelian conception of form and matter. Aquinas argues that no creaturely being can be infinite in essence, 13 and that there cannot be an actual infinity of distinct beings. 14 Fortunately, Aquinas' arguments regarding material things need not detain us here. It will suffice to examine his grounds for claiming that no created non-material being (e.g., an angel) could be infinite in essence, and for thinking that such beings could not be inf'mite in number. The former question is taken up in Part I, Q. 7, Art. 2 of the Summa Theologica: If however any created forms are not received into matter, but are selfsubsisting, as some think is the case with angels, these will be infinite, inasmuch as such kinds of forms are not terminated, nor contracted by any matter. But because a created form thus subsisting has being, and yet is not its own being, it follows that its being is received and contracted into a determinate nature. Hence it cannot be absolutely infinite. And Aquinas goes on to say that, quite generally: It is against the nature of a made thing for its essence to be its existence; because subsisting being is not a created being; hence it is against the nature of a made thing to be absolutely inf'mite. Here Aquinas is certainly arguing that created beings cannot be made absolutely infinite. Aquinas' conception of God's inf'mitude invokes the idea that, because God's nature includes all perfections, His being is not limited in any way. Creatures, whose natures are determined by their failing under a genus and a species, have a 'contracted or limited' mode of being which locates them in a hierarchy of being. A creature might be perfect in one respect or another, but it will necessarily lack that unbounded possession of every perfection, that complete absence of limitation, which characterizes the being who stands at the head of this hierarchy. One might read Aquinas as claiming that aseity, rather than being just one perfection among others, is a perfection which supervenes upon possession of all the other perfections. If that is how aseity is to be understood, then nothing could be both a created being and possess every

72 infinite perfection, save only aseity. But why should we suppose that aseity supervenes in this way upon the other perfections? Clearly, Aquinas believes that aseity involves some sort of infmity that creatures, by definition, lack. Even if we grant this, it would of course not follow that a lack of aseity entails a lack of every species of infinite perfection. Indeed, Aquinas allows, in the quoted passage, that angels can be infinite in some respect(s). Finally, we may take note of the reason Aquinas gives for supposing that no creature can be infinite in essence: I answer that, It is one thing to be infinite in essence, and another to be

infinite in magnitude. For granted that a body exists infinite in magnitude, as fire or air, yet this could not be infinite in essence, because its essence would be terminated in a species by its form, and confined to individuality by matter. 15 But again we must ask: why couldn't a created form have an essence that contained infinite perfection in some respect? Why, in particular, couldn't it involve infinite goodness? I presume that, for Aquinas, the answer to this question lies with his equation of goodness with being: a being cannot be infinite qua being unless it is a se; a being which is not infinite qua being cannot be infinite in goodness; a being which is created is not a se; and so a created being cannot be infinitely good. Aquinas' conception of infinite being is of a kind of being that is not limited in any way. Created being is a limited kind of being, in just the respect that a created being depends, for its existence, upon a creator. A creature does not possess the kind of being that is complete in the sense of being entirely self-sufficient. Moreover, if we take self-sufficiency (aseity) to be a good thing, then by definition creatures lack a good-making property. But why must a created being lack infinity in any other respect? Since Aquinas has already granted that creatures can in some respects be made infinite, how does conceding this point about aseity to Aquinas yield the conclusion that there is no best possible world? I do not intend to put these questions forward rhetorically. I shall answer Aquinas' challenge by attempting a description of a world of the requisite sort. But before doing so, I shall attend to two items of business: (1) Aquinas' argument that it is logically impossible for God to create an actual infinity of beings, and (2) Schlesinger's argument that there is no best possible world that God can create. Aquinas' doctrine that there cannot be an infinite multitude of creatures is influenced by Aristotle's arguments for f'mitism in Physics III, arguments it would take me too far afield to examine here; they are in any

73 event arguments primarily directed against an infinity of spatial beings. But Aquinas has a general argument not confined to spatial beings. 16 That argument, however, is clearly fallacious. It is that if God creates a multitude of beings, he must create some number of them. But no number is infinite. So God cannot create an infinity of beings. Aquinas is mistaken, of course, in thinking that there are no infinite numbers. 17 And there is no evident contradiction in the supposition that God has created, say, an infinite number of angels. How do matters stand, then? If Aquinas can show that God is necessarily an infinite being, and that every created being is necessarily finite, he will have shown that between creator and creature there is necessarily an infinite gap in being - hence, in goodness. Then God's omnipotence can be invoked to show that God could increase without limit the goodness of creation. But even if we were to grant that perfect goodness (= perfect being) entails aseity, this line of reasoning encounters two serious obstacles. First, the sense in which God's being is perfect, or infinite, consists for Aquinas in the fact that God's essence, or form, is not received in any 'matter', i.e. substrate [suppositum] distinct from His nature. 18 In that sense, His form is said not to be limited, as the form of a man is limited by the matter of the man. This, combined with the fact that in God alone are existence and essence to be identified, specifies a fundamental distinction between God and His creatures. But it is a mistake to characterize that distinction in terms of the difference between something finite and something infinite, as if creature and creator could in this respect be said to be measured by some extensive quantity of which creatures possess only a f'mite amount, but God a non-finite amount. 19 In short, this conception of God's infinitude does not support the idea that there is a 'gap' between creator and creatures of such a sort that creatures can be linearly ordered with respect to how much they have of what God has in unlimited fashion. Nor, therefore, can any sense be given to the idea that there is no upper limit to the amount of this quality - aseity - that created beings can have. They just don't have it at all. The second obstacle is that, even if creatures could be so ordered, and even if an infinite sequence of possible worlds of increasing goodness could in this way be generated, it has not yet been shown that that sequence has no finite upper bound achievable by God. 2° In short, possible worlds might be arrayed along a scale of increasing goodness in a Cauchy sequence which has as a least upper limit a possible world (or worlds) than which none better is (logically) possible. I know of no argument to show that this is the case. But I also know of no argument which shows that it is not.

74 I turn now to Schlesinger's version of the argument, which is considerably simpler in structure than Aquinas'. Schlesinger considers only a single dimension along which worlds can be ranked according to their degree of goodness. Goodness, for Schlesinger, is measured by the degree of desirableness of the states experienced by God' s creatures. (Schlesinger clearly has in mind experience just in the sense of conscious experience). Schlesinger is at pains to distinguish a creature's degree of satisfaction with its state - that is, how desirable it thinks its state to be - from the objective desirability of that state. And he makes the familiar point that certain states are more desirable even if they involve less contentment: better Socrates unhappy than a fool satisfied. Schlesinger's central claim, then, is that there is no finite upper limit to the degree of desirability of the states into which God could place the creatures He creates. For, even if there were a maximum to the degree of desirability of the states enjoyable by a given sort of creature, there are always creatures God can make which are capable of enjoying states which have a yet higher degree of desirability. Schlesinger assumes implicitly - and without argument - that nevertheless no creature or set of creatures taken collectively can exemplify a state which has an infinite degree of desirability. And he further assumes that there is no finite logical limit to the infinite succession of possible states with increasing desirability. But for the moment I shall ignore these assumptions. Schlesinger's argument for his central claim is, unfortunately, highly compressed. He says: Even the most satisfied man on earth could make up an endless list of requirements the fulfillment of which would increase his happiness. 21 And Schlesinger goes on to imply that there is no limit to the possible beings capable of ever greater heights of happiness. In Religion and Scientific Method, he says simply: '...I take it that, conceptually, there is no limit to the degree to which the desirability of a state may reach'. 22 From this, Schlesinger draws the conclusion that: There being no prima facie case for saying that the possibilities for greater happiness are finite, God's inability to create the greatest state of happiness is seen to be no different from his inability to create the greatest integer. 23 What this amounts to, however, is not so much an argument for there being no limit to possible happiness, as its bald assertion. And surely, this is not the sort of claim which can plausibly be said to be self-evident. But

75 if happiness can increase without limit, why suppose that God could not create an infinite amount of it? Why think that this task would be like the task of creating a greatest integer? To be sure, God can't do the latter; but if He can create all the integers and all the reals, 24 is there any reason to suppose that He cannot create also the infinite cardinals? And, if He can create transfinite cardinals, is there any reason to believe that He cannot create extensive quantities - such as happiness - in inf'mite amounts? Before I reply to these questions, let us consider one final argument for the view that there is no best possible world. This argument has been recently articulated by William Mann. 25 Mann's view is that there are incommensurable goods, and that worlds containing differing goods may therefore themselves be incommensurable. Hence, with respect to the property of goodness, the judgment 'greater than' cannot in general be applied across possible worlds, and the notion of a greatest possible world becomes incoherent. Mann is appealing to the fact that there are .different species of goods e.g., aesthetic goods, moral goods, intellectual goods, and so on. 26 But he seems mistakenly to think that this implies that goods of differing species cannot be compared and ordered along a single scale of degree of goodness. This mistake is, no doubt, abetted by the fact that preference ranking by human beings is characteristically a vague and imprecise enterprise. We are no more able to make precise orderings with respect to degrees of goodness than we are able to assign precise numerical subjective probabilities to our beliefs. But this feature of our value judgments applies just as much to comparative judgments between goods within a species as to cross-species judgments. Nor does this imply that rough cross-specific value judgments are impossible to make, or unintelligible. While I may not know exactly how much aesthetic good represents an equal trade-off for a certain moral good, I certainly have no difficulty making rough estimates of that sort. 27 And God, I think we may safely assume, has no difficulty in making precise judgments of comparative value. So Mann's solution does not appear to be an acceptable one. But to anyone who remains convinced that a world containing even an infinite number of satisfied fools is less desirable than one containing some (perhaps finite) number of unsatisfied Socrateses, let me emphasize that the question before us is whether we can optimize the very highest forms of creaturely perfection.

76

4. The very best possible worlds I propose to explore in this section the possibility of constructing a positive conception of the best possible world. My strategy will be to attempt to answer a series of questions, as follows: (1) (2) (3) (4)

What is the best kind of being (if any) that it is possible for God to create? What is the best possible set of circumstances (if any) for such a being to be in? What is the optimal number of such beings (if any)? Are there any lesser beings which, if added to the universe constructed in answering (1) - (3), would make it better still? And what would be the optimal number and environment for such creatures?

(1) Orthodoxy holds that God, being self-sufficient and complete, creates the world out of love and not to satisfy any need He has. But from our present perspective, it might seem that the best possible world would be one containing God alone. Such a world, after all, contains an inf'mity of perfection and no evil; moreover, finite created beings cannot add any goodness to such a world, since their goodness is finite and does not increase the total amount of goodness. However, we should also note that, even if God could create beings which would add an infinite amount of goodness to the world, this would still not increase the total amount of goodness, if God is infinitely good. Still, if God does create a world, and there is a world (or worlds) He can create that cannot be exceeded in goodness, then it would be inconsistent with His nature not to create such a world. God would maximize the goodness of the created order, even if He would not thereby increase the total goodness in existence. Since God is the paradigm of perfection, I propose to answer question (1) by inquiring how closely God could approximate His own nature in a creaturely nature. Only logic stands as an impediment to what God can do. Hence, He can create any being whose description is not logically incoherent. Such a creature must, for starters, lack aseity and any divine characteristic which entails aseity. I believe it must also be a non-abstract being, but if any theist affirms that God can create abstract beings (such as universals and numbers), that will in no way weaken my case. I shall confine my discussion to concrete beings. I can see no reason why God cannot create a being that is a perfectly free agent, is morally perfect, omniscient, non-material, without temporal beginning or end, which exists in a state of beatitude, and is quasi-omnipotent. 28 Moreover, I can see no

77 reason why such a creature could not possess these attributes essentially. There are other attributes - e.g. simplicity, atemporality, and immutability that could be added to this list, but they are problematic, even as applied to GOd, and I shall leave them out of the discussion. (God knows whether to add them.) Even so, a number of the attributes listed require comment. By quasiomnipotence, I mean the power to do every logically possible thing save only the power to exist a s e . 29 A quasi-omnipotent creature could not make, change, or destroy God; but even God cannot do these things. And, God could not destroy such a creature, if it were contrary to His own nature to do so. 3° It is sometimes supposed that there could not be two autonomous omnipotent beings in the world, for each would limit the power of the other. But, as Morris has shown in his discussion of the Trinity, 31 there is no incompatibility here so long as there is no possible clash of wills. And, as I propose to show in my conclucing section, omniscience and omnibenevolence together will insure a concordance of God's and creature's wills, without infringing autonomy. Could a creature be omniscient? I can't see why not. There being an infinity of propositions, omniscience presupposes an infinitely capacious mind, but theists are committed to the logical possibility of such minds. Similarly - unless it entails aseity - there is no apparent reason why a creature cannot exemplify perfect moral goodness. Finally, there is no reason why such a creature must have a temporal beginning, if it is allowed, as it generally is, that God can efficiently cause a thing to exist at every time whatsoever. -

(2) Let's call the creature possessing all these perfections t~. What is the optimal environment for ix? Since I am proposing a condition of beatitude i.e., a state of intimate communion with God - for {~, it might seem that there is nothing left to be desired. Moreover, if, as I have yet to show, a world containing only God and {x is insured against clashes of wills, then this world would contain no evil. Further, if {x can be said to exemplify infinite goodness, then there is no amount of goodness that the created ~world lacks, and no amount that can be added so as to increase its goodness. Nevertheless, I have two worries. First, it might be that it makes sense to ascribe infinite goodness to God, but not to a creature such as ix. If so, then the world's goodness, being finite, has not been maximized. Second, goodness is essentially diffusive; it involves love for another, ix, of course, loves God. But suppose this love is not a perfect fulfillment of tx's goodness: what if such goodness demands also the opportunity to create? These two worries suggest that a better world than the ix-world might -

78 be a world in which tx creates another being 13, just like itself. Is such a world one in which not everything is caused by God? Clearly not: God remains the ultimate cause of everything. However, this world might seem to be one in which there is a disparity between God's will and tx's will: for it is a world in which tx, but not God, intends to create [3. (If God and tz willed 13% existence, then either God's will would pre-empt ct's, or 13's existence would be a joint venture or else causally overdetermined.) But this is not a genuine problem: God's will can concur with tx's by way of God's willing (in the sense of agreeing to) tz's willing that 13 exist. God and o~, in turn, would agree to the creation, by 13, of another such creature ~,, and so on ad infinitum. I shall call creatures of the kind I have been discussing perfect-creatures. Perhaps God, and each perfect-creature, can make more than one (an infinity?) of perfect-creatures. If so, let the world be one in which God and each perfect-creature fulfills its goodness by creating as many perfectcreatures as possible. Such a world will contain an inf'mite amount of creaturely goodness, for it contains, at every time, infinitely many perfectcreatures. This, then, answers question (3). Now, however, a further problem emerges. Perfect-creatures have identical essences; they all belong to the same natural kind. And, like angels, they are not individuated either by matter or by spatial location. What, then, individuates them? I reply that they must be distinguished, either by differing accidents, or by distinct individual essences. If Kripke is right - as I think he largely is - about essences of origin, then the latter possibility offers a natural solution. It is essential to tz that it was created by God, essential to [3 that it was created by tz, and so on. On this solution (and if perfect-creatures cannot differ in their accidents), it is not logically possible for God, or any perfect-creature, to create more than one perfectcreature. The world, however, would still contain an infinity of perfectcreatures. (4) Could such a world, involving as it would a perfect communion between God and His perfect-creatures, and containing no evil, be one that God could improve? I think not. Clearly, the total amount of good could not be increased; nor could the total amount of evil be diminished. In fact, adding imperfect creatures would mean adding, or running the risk of adding, evil to the world. However, the presence of evil might be thought for three reasons actually to improve the world. First, it might be alleged that the world I have just depicted is a boring world. But I cannot see why such a world would be boring, any more than I can see why a heavenly existence should

79 be boring for the saved. More seriously, it might be argued that such a world makes impossible the existence of certain higher-order goods, goods such as courage and beneficence, which require the existence of evil or potential evil as a logical precondition. I answer that these goods require the existence of evil as a logical precondition for their exercise, but not as a logical precondition of their existing dispositionally. On the other hand, I am not sure that it makes sense to ascribe courage to God. But if not, then we can can be sure that the world is no worse off for that fact. Similarly, I am unsure whether it would make sense to ascribe courage (even dispositionally) to perfect-creatures (who, being necessarily quasi-omnipotent, would necessarily never be in a position to exercise it); but again, the world would be none the worse for that. A third and related objection can be derived from the soul-making theodicies of Hick and Swinburne. According to them, there is intrinsic value in the struggle by means of which our souls attain to moral insight, virtue, and ultimately, salvation. This struggle requires the presence, and overcoming of, evil. In a world of perfect-creatures, there can be no such process. Perfect-creatures are created with fully made souls. My reply, rather too briefly, is to agree that, for creatures such as we, soul-making has intrinsic value. But it has no such value for creatures whose souls are already made, and lack nothing. Such creatures could imagine evil, and in some vicarious way empathetically imagine suffering, just as God can. And in this there would perhaps be a value analogous to that which is created by the process of soul-making. But the intrinsic value of soul-making, however great it might be, comes at a price; and in any case this value, even if infinite, would not increase the total value already existing in a world of perfect-creatures. Finally, one might worry whether the sort of world I have been depicting falls short of optimality (or even contains a large amount of evil) in virtue of some feature of the relations among its creatures (or between them and God). I cannot conceive what such a defect could be. My models in constructing this world have after all been, with respect to individual creaturely natures, the perfect nature of God, and, with respect to the community of being, the community of the souls of the saved in heaven. The conception of a best possible world that I have been sketching is not one whose coherence I can claim to have established. Undoubtedly, it contains difficulties which I have failed to explore. However, I think it has about as much promise as the Anselmian conception of God as a maximally perfect being. If it, or some variant of it (e.g., a world in which imperfect beings after all figure) is a correct conception, then God must, in virtue of His nature, create such a world. Any conception of God's freedom, therefore, according to which God cannot freely choose a world

80 if there is a best possible one, is a conception which is held hostage to the live possibility that there is such a world. It is time therefore to re-examine the question whether what God does necessarily, in virtue of His nature, is a fortiori done without freedom of choice. In re-examining this question, I hope to vindicate both the idea that God freely chooses what He necessarily wills, and the conception of perfect-beings as autonomous creatures who nevertheless necessarily will what God wills, and hence can be made quasi-omnipotent by God. Here I wish to note that Morris' own view of the nature of the Trinity seems to require some such conception of freedom; for otherwise, either some or all the omnipotent persons of the Trinity lack autonomy, or else their universal concurrence is an accident.

5. Freedom and necessity The paradox generated by the problem of free will does not arise solely from the apparent conflict between a libertarian conception of freedom and a causal version of the Principle of Sufficient Reason (viz., that every event has a sufficient efficient cause). It is further exacerbated by the observation made forcefully by Hobart 32 that free will actually requires some kind of agent control - i.e., some kind of determination. This can be seen quite transparently if we consider the character of an action which is purely random, spontaneous, or governed by indeterministic laws. Such an action would not have an explanation (at least no complete explanation), would not be (to the extent of the indeterminacy) under the agent's control, and hence would not qualify as having been chosen by that agent. Choices cannot be just accidental if they are to be free. Let us call this the Determination Condition (DC) on free choice. On the other hand, we associate freedom with the condition that (prior to action) the agent can perform any one among at least two alternative prospective actions, and with the condition (retrospectively) that the agent could have done otherwise. This was, for Morris, a sine qua non of free action; with Morris, I shall call it the Principle of Avoidance (PA). In part, then, the problem of fee will is the problem of 'making room' for free choice in a world of causes; but in part, it is also the problem of showing the notion of free choice to be an internally coherent one, of showing that PA can be reconciled with DC. I believe there is a libertarian way to achieve this reconciliation, and I believe that it is not only the most plausible way of understanding human freedom, but also provides the most adequate conception of divine freedom. The analysis of fee choice which I shall propose derives, in part, from

81 Aristotle's conception of practical reasoning. According to it, acting freely consists in acting as a result of rational deliberation over what act to perform. To choose is to deliberate: it is not to perform some superadded 'act of will'. This conception of freedom makes obvious the connection between being free and having the capacity to engage in rational deliberation. It also makes it clear why PA is so central to the possession of a free will: to have a free will is to have the capacity to act as a result of engaging in efficacious deliberation; but in order for my deliberation to have any point, I must at least believe that I have before me two or more genuinely possible alternative courses of action. Of course, I can take my deliberation seriously even if this belief is false; but in that case, my supposition that I am free to choose is a delusion. So freedom consists in efficacious deliberation - i.e., in the performance of deliberation-guided actions; and efficacious deliberation presupposes genuine alternatives, each possible given antecedent conditions. How is this to be squared with DC? A hint is supplied by the notion that the deliberations of a free agent must be efficacious. If to act freely is to be guided, in acting, by the outcome of one's choosing - i.e., deliberating over how to act - then there must be some non-accidental connection between reasons, choice, and action. Clearly, this relation is not one of logical necessitation. Although the conclusion of a practical syllogism might follow logically from the premises, it does not follow logically from the fact that an agent accepts and employs the premises of a syllogism, that he or she also draws the conclusion. Mistakes are possible, as is the interruption of reasoning by external circumstance - e.g., suffering a heart attack. We naturally conclude that the relation must be a causal one: an agent's accepting and employing certain reasons causes him or her, other things being equal, to draw a certain conclusion and so act. But that way of satisfying DC conflicts with PA. For libertarians there is, I think, only one way out of this dilemma; but fortunately, it is a solution that has a good deal, on independent grounds, to be said for it. This way requires affirming that there is indeed nonaccidental connection between a rational agent's employment of certain premises in a deliberation, and the choice which emerges, but that this connection is neither a causal nor a logical one. I happen, against Hume, to think that causal relations involve a kind of necessary connection, but my present claim is compatible with a regularity theory of causation. It is that reasoning involves a species of non-accidental connection between the fact that an agent accepts one proposition and the fact that he or she then accepts another, that is radically different from the relation between physical causes and their effects, the relation that makes true the laws of nature.

82 That there is such a non-accidental connection between thoughts in a train of reasoning, and that it has a character different from physical causation, are facts open to introspective reflection. When one reasons, what one comes to accept is clearly guided by what one previously accepted. The relation between these is neither arbitrary nor random. But it is not causal. It makes sense to speak of reaching a proper conclusion but not of the occurrence of a proper effect. Reasoning is guided by the semantic content of our thoughts; it involves 'seeing' that one proposition follows from another, and forming our beliefs and action-choices accordingly: it involves an intentional relation to the rules of logic. To use Morris' terminology, it involves following a rule, as opposed to merely acting according to a rule, and also as opposed to merely knowing that rule. Causation involves nothing of the kind. The better our reasoning, the more fully we are rational and free agents; but it makes no sense to speak of a physical entity fulfilling its causal role better or more poorly. How do we know that there exists such a special non-accidental connection between our thoughts when we reason? Bluntly put, we know this because we know that we do reason. If there were no such relation between our thoughts, we would not be reasoners at all, no matter how closely our thoughts mimicked those of a genuine reasoner, whether by accident or because of some merely causal connection. 33 We know we are reasoners because reasoning involves making inferences; that is, grasping logical connections - and to do this is to perform a mental act which is at least sometimes available to introspective reflection: it is, indeed, an utterly common experience. So, I believe, we are aware of a genuine connection, one which accounts for much of the non-random character of our thoughts, which is not causal. To embody connections of that kind lies at the heart of what it is to be a rational agent - and at the heart of what it is to exercise free will by engaging in genuine deliberation. 34 A libertarian, therefore, can hold that DC is satisfied by freely chosen actions, while denying that these actions are caused (more precisely, while denying that there are causes causally sufficient for their occurrence). Now, God is a perfectly rational agent, and He is so out of logical, or metaphysical necessity. Thus His train of thought, when He reasons, embodies perfectly the good-reasons relation (as I shall call it), and is, moreover, not constrained by any causal laws. Since God is also omnipotent and omnibenevolent, there is no distinction in Him between moral and prudential reasoning. Since God necessarily arrives at the correct conclusion in any deliberation, and necessarily does not suffer from weakness of the will, He necessarily does what is morally best (if anything) - and performs each such action with perfect freedom. 35 Thus God's necessarily fulfilling His moral obligations is compatible with -

83 indeed, it is partly constitutive of - His essential perfection as a free and rational being. There are two natural objections to this conception of divine freedom. The first objection holds that the conception simply cannot apply to God because God does not engage in discursive reasoning. God does not have to deliberate; He just knows directly what ought to be done. As we shall shortly see, God's choices do represent a degenerate (in a non-pejorative sense) species of choice. But this objection is misguided. God does act for reasons, and in accord with His reasons. It does not matter whether God must engage in discursive reasoning in order to see that a certain action ought to be performed, or whether, as we believe, He just sees this immediately. In either case, He sees that the action is wanted because the reasons dictate it, and in this way His choosing embodies the crucial goodreasons connection that lies at the heart of the above analysis of freedom. The second objection is more serious. It holds that God cannot be free in the proffered sense because he is essentially rational. This means that it is not logically (or at least not metaphysically) possible for God to do other than He does, given His reasons. If God's actions are logically necessitated by His nature, then He is not free after all. This objection, too, is misguided. It contains the suggestion that we are free only because we are, unlike God, capable of irrationality. It suggests that the exercise of a free choice must fundamentally focus on the question of choosing whether one should be rational or not. It implies that humans are free only because they can choose to be irrational: because God cannot choose an irrational act, he is unfree. But this suggestion is incoherent. No one can freely - that is, rationally - choose to be irrational, although humans, because of their fallibility, can freely choose actions which are in fact irrational. 36 The mistake lies in confusing the necessity of God's being rational with the good-reasons connection between premises and conclusions thereby embodied in all His choices. It is true that, necesarily, if a being is God, He will act rightly. But this de dicto logical truth about divinity does not show the connection between God's having reasons and His acting to be a logically necessary one. Similarly, it is a de dicto logically necessary truth that poison harms, but it doesn't follow that the causal connection between ingesting a particular poison and being harmed is one involving logical necessity. It is a de re metaphysically necessary truth that God is rational. But this only shows it to be metaphysically necessary that God's thoughts instantiate the non-logical necessary connection that obtains between reasons and actions in a rational agent, not that those actions are logically or metaphysically necessitated. Similarly, it is a de re metaphysically necessary truth that an electron has charge -e, and thus that it has certain

84 causal powers. But it is only causally necessary that it therefore repels this other electron. Nevertheless, it is true that, if there is a morally best action that God can perform, He necessarily performs it. Thus it appears that such actions violate PA. Does this not show these actions to be unfree? We can begin to see what is at stake here by considering the case of human actions. We tend to think of actions as paradigmatically free when careful deliberation is a prerequisite to seeing what ought to be done - that is, where the rights of the matter are, initially at least, far from obvious as when one deliberates over whether to join the resistance movement and help free one's country from foreign rule, or to say with one's ill and aged mother. It is in these cases that deliberation seems most necessary, and most obviously serves an important function. But many - indeed legion are the cases in which one acts for reasons, but in which discursive deliberation is quick and easy, even superfluous. These are the cases in which the consequences of alternative acts are clear, and it is clear what consequence is to be preferred. Such cases are surely not cases of unfree action; on the contrary, they are cases in which the good-reasons connections between reasons and action are particularly strong. If I see a small child drowning in three feet of water at the shallow end of a swimming pool (and see no other adults nearby), there is no question over what I should do; and my decision to act will the more compelling, the more alive I am to the strength of my reasons for saving the child. But such clear recognition of what it is that I should do certainly does not render my action undeliberate or unfree. It is a paradigm of free action: choices are not less free because they are more obvious. For God, correct choice is as obvious as it can be. But that means that God is perfectly free. His rationality functions flawlessly, unmarred by ignorance, weak reasoning, or temptation. It turns out, then, that God's freedom, precisely because it is perfect, represents a limiting case of freedom, one in which PA is violated. But that is inevitable. PA is grounded in the fact that the telos of deliberation is (ideally) to choose between options by finding which is best; and this presupposes that there be options, in the sense of causally possible alternatives. But choosing then requires that one be governed by reason. If one already knows what is best, the need for deliberation is short-circuited, but one must still act for reasons, if one is to be said to have chosen so to act. God's choices are not constrained at all by causal necessity: in that sense, he could have done otherwise, more freely than we. But when He does act it is, out of necessity, only for the best of reasons: in that sense He is perfectly rational, and also more perfectly free than we, precisely because He could not do otherwise.

85 If there is a best possible world, and God creates it, then every w o r d directed act He performs, He does in virtue of His essential goodness. There is no room for what Morris calls supererogation on God's part. But this does not have the consequence that God is not free.

Notes 1. Thomas V. Morris, Anselmian Explorations: Essays in Philosophical Theology (Notre Dame, IN: University of Notre Dame Press, 1987), Ch. 2. This chapter first appeared in the American Philosophical Quarterly 21 (1984): 261-268. 2. This second characterization of freedom is commonly associated with compatibilism. In Section 5, I shall show that, given a certain conception of the relation between reasons and action, it can be so understood as to be a robustly libertarian one. 3. Ibid., p. 28. Morris speaks, in this formulation of PA, of 'broadly logical' necessity because God would not in any case be constrained by causal necessities. As it stands, PA properly applies to the deity only if the deity is a temporal being. On this score, Morris holds (Ibid., p. 136) that God is 'not an altogether a-temporal individual'. (I am unsure what 'not altogether' could mean in this context.) In order to apply PA to an atemporal agent, the phrase 'prior to t' must be interpreted to mean 'prior in the causal order' rather than temporally prior. Note that Morris formulates PA in terms of logical necessity because God is not constrained by physical necessity. 4. When Morris explicates the notion of divine duties, the examples he uses are ones involving conditional duties, that is, duties such as promise-keeping and truth-telling that are freely assumable. It is a further question whether God has any unconditional duties, that is, duties which apply to God simply in virtue of his existence or nature. As I shall show, we can leave that question unsettled, for if there is any logically possible action A which it is best for God to perform, then he will be necessitated by his nature to perform it, in violation of PA. Whether the performance of every such action should be regarded a divine duty, is a further matter which need not detain us here. 5. I am including not creating a world at all as one of God's options here: if not creating anything is the best thing God can do in this line of business, then, necessarily, He will refrain from creating. 6. Morris, Ibid., pp. 16-19. 7. George Schlesinger, 'The Problem of Evil and the Problem of Suffering', American Philosophical Quarterly 1 (1964): 244-247; 'Omnipotence and Evil: An Incoherent Problem', Sophia 4 (1965): 21-24; and 'On the Possibility of the Best of All Possible Worlds', Journal of Value Inquiry 4 (1970): 229-232. 8. Norman Kretzmann, 'A Particular Problem of Creation', in Scott MacDonald, ed., Being and Goodness: The Concept of the Good in Metaphysics and Philosophical Theology (Ithaca, NY: Cornell University Press, 1991). Condition (a) seems to me to be correct. 10. Since, given a particular set of parts from which to create the world, God .

86 necessarily maximizes the goodness of their order, improvement purely with respect to order is impossible. See Kretzmann, Ibid., pp. 233-235 for details. 11. Kretzrnann translates this passage from Scriptum super Sententias I, d. 44, q. 1, a. 2. 12. If the possible worlds form a compact linear ordering with respect to amount of goodness, then there could be an infinity of degrees of goodness exemplified by worlds, and no best world, even though there existed a finite upper bound to the amount of goodness a world can display. The analogy here is to a Cauchy sequence, or to an interval on a line that has a finite upper bound but is not closed at that end. Aquinas did not contemplate this possibility, and I shall give reasons for rejecting it in the next section. It can be handily excluded by reading premise (2) as asserting that for some finite amount of goodness n, and any world whose measure of goodness is some finite quantity m, God can make a universe whose goodness is greater than

m+n. 13. Aquinas, Summa Theologica, Fathers of the English Dominican Province, transl., 2nd ed. (London: Burns Oates and Washbourne Ltd., 1920), 1,7,2. I have used the Dominican Friars' translation with one exception suggested by Scott MacDonald: where the Dominicans have 'relatively infinite' I have substituted the clearer 'infinite in some respects'. 14. Ibid., 1,7,4. 15. Ibid., 1,7,3. 16. Ibid., 1,7,4. 17. If anyone should deny that the transfinite cardinals are numbers, I will in turn deny that God's creating angels entails that He creates some number of angels. 18. Ibid., 1,7,1; and 1,3,3 and4. 19. Aquinas commits this error at I,Q.7, Art. 1, where he says that what is not finite is infinite - and then understands finitude as I have indicated. An alternative, and perhaps better, way to understand Aquinas' conception of infinitude in this context is (as previously suggested) as the characteristic of not being limited in any way. Thus, God's goodness is infinite in the sense of not having any limitations. Could a created being have that kind of goodness? I can think of no reason why the will of such a being could not be perfectly good, but it does seem to be a limitation on the goodness of a being if it lacks omniscience and omnipotence. For, lacking these, a being with a perfectly good will lacks the means to insure that this will is translated into correspondingly perfect action. Thus unlimited goodness in the sphere of action seems to require no limitation in knowledge or power. This point has some force: it forces us to consider whether either omniscience or omnipotence requires aseity. I shall take that issue up in the next section. 20. William Shea makes this point in a response to Schlesinger's argument. See Shea, 'God, Evil and Professor Schlesinger', Journal of Value Inquiry 4 (1970): 219-228. 21. Schlesinger, 'The Problem of Evil and the Problem of Suffering', p. 246. 22. Schlesinger, Religion and Scientific Method (Dordrecht/Boston: D. Reidel Publishing Comp., 1977), 62. 23. Schlesinger, 'The Problem of Evil and the Problem of Suffering', p. 246. 24. Unlike some theologians, I have grave doubts about God's ability to create numbers. But He can create things that are numberable, such as spaces and

87 the points in them. 25. William E. Mann: 'The Best of All Possible Worlds', in Scott MacDonald, ed., Op. cit., pp. 250--277. 26. Mann's argument could be construed as appealing to incommensurability between just moral goods. One might, with some plausibility, hold that there is no way of determining a trade-off between, e.g., justice and other moral goods. But the reply I give here applies equally to that case. 27. When, after he has eaten a good meal, we Offer our son a choice of desserts, he invariably begins byzantine negotiations: if we offer a cupcake, how many pieces of licorice would we be prepared to give him in its place? - and so on. He has an all-too-keen appreciation of sweets, and he clearly regards these goods as commensurable. But it is equally clear that, as the various proposals become fine-tuned, his preference-judgments become essentially arbitrary i.e., there are no precise trade-off points. Mann's own cross-specific example is for his purposes peculiarly inapt: he considers a case of alternative careers as an opera singer or as a caregiver for the dying poor in a hospice. Other things being equal, and given talent in both these fields, I, for one, have no difficulty whatever in judging the latter to be better. 28. That is, I find no grounds for supposing that any one of these perfections, or their conjunction in a single being, logically entails or makes metaphysically necessary the aseity of that being. 29. Perhaps there is a conception of power in terms of which only an a se being can be omnipotent. That is, we might suppose that being a se requires the possession of a special power of some kind. If so, no creature can be omnipotent. But it would not follow that creatures cannot be quasi-omnipotent. Moreover, aseity is a bimodal property: either you have it, with the accompanying power, or you don't. So there can be no question here of infinitely many degrees of quasi-omnipotence. There remains, to be sure, the question whether lacking the power of aseity entails lacking any other powers - whether there is any other power such that having the power of aseity constitutes a logically necessary condition of having it. If so, then quasiomnipotent creatures would lack that power. But I find no grounds for thinking that there are such entailments. 30. Creatures of the kind I am envisioning will be necessary beings if they are constituents of any best possible world. However, they are not necessary in the way numbers are. The existence of these creatures is dependent upon the causal power of God: focusing on the causal powers of God, and prescinding His essential goodness, we can say that it is within God's power to create, or refrain from creating, such beings. But I do not think it makes sense to say that it is within God's causal power either to create or to prevent the existence of numbers. 31. Morris, Ibid., pp. 141-144. 32. R.E. Hobart, 'Free Will as Involving Determination and Inconceivable Without It', Mind 43 (1934): 1-27. 33. Our train of thought would then exhibit an order which was in accordance with the rules of logic, but we would not have been following those rules. 34. I have argued for this view at greater length in 'Davidson's Compatibilism', Philosophy and Phenomenological Research 44 (1984): 227-246. 35. If there are two or more alternative acts tied for first place, God can of course arbitrarily pick one. But it would be a mistake to think that this fact captures

88 the essential character of God's freexiom. Choices of this kind are parasitic upon the ability to choose whatever is best, non-arbitrarily and because it is best. And this kind of freedom would exist for God only if there are equally good best choices. 36. One can act irrationally through weakness of the will, through mistakes in reasoning, through being guided by irrational beliefs, and because of incorrect assessments of the relative importance of competing goals. Whether being subject to irrationality must necessarily decrease one's freedom as an agent, is a difficult question; but certainly it will do so to the extent that nonrational choices are caused by reason-independent factors.

Address for correspondence: Prof. Evan M. Fales, Department of Philosophy, The University of Iowa, Iowa City, IA 52242, USA Phone (319) 335 0023; Fax: (319) 335 2535