SOPHIA (2012) 51:341–349 DOI 10.1007/s11841-011-0287-8

On the Argument from Divine Arbitrariness Peter Forrest

Published online: 21 June 2012 # Springer Science+Business Media B.V. 2011

Abstract William Rowe in his Can God be Free? (2004) argues that God, if there is a God, necessarily chooses the best. Combined with the premise that there is no best act of creation, this provides an a priori argument for atheism. Rowe assumes that necessarily God is a ‘morally unsurpassable’ being, and it is for that reason that God chooses the best. In this article I drop that assumption and I consider a successor to Rowe’s argument, the Argument from Arbitrariness, based on the premise that God does not act arbitrarily. My chief conclusion will be that this argument fails because, for all we know, there can be non-arbitrary divine choices even if there is no best act of creation. Keywords God . Perfection . Choice . Arbitrariness . Buridan's ass . Rowe . Argument from evil

William Rowe in his Can God be Free? (2004) argues that God, if there is a God, necessarily chooses the best. Combined with the premise that there is no best act of creation, this provides an a priori argument for atheism. Rowe assumes that necessarily God is a ‘morally unsurpassable’ being, and it is for that reason that God chooses the best. In this article I drop that assumption and consider a successor to Rowe’s argument, the Argument from Arbitrariness, based on the premise that God does not act arbitrarily. This argument comes in two versions. Here is the first. The Bold Argument from Arbitrariness: If there is a God, God does not act arbitrarily. Therefore, if there is a God, God performs the best possible act. God does not perform the best possible act. So there is no God. I presented a version of this paper at the April 2010 Metaphysics and Philosophy of Religion workshop at San Antonio. I would like to thank the organizer, Mike Almeida, and all who participated in the discussion. I would also like to thank the referees who pointed out several serious errors in an earlier draft. P. Forrest (*) School of Humanities, UNE, Armidale, NSW 2351, Australia e-mail: [email protected]

342

P. Forrest

I claim that, for all we know, this argument is not deductively valid because there could be acts that are neither arbitrary nor best. Such acts are possible only when there is no best, and they must satisfy the principle that I call No Vicious Perfection.1 This principle, elaborated below, states that an agent should ignore those excellences (of acts) that cannot be maximized, and choose accordingly. 2 Here is another version, one that is, I believe, deductively valid, but whose third premise lacks warrant: The Modest Argument from Arbitrariness: If there is a God, God does not act arbitrarily. Non-arbitrary acts are constrained by No Vicious Perfection. God does not perform an act thus constrained. So there is no God. Both arguments share a premise that may well be rejected but I find plausible, namely that if there is a God, God does not – I would say cannot – act arbitrarily (although God might fail to act without a reason for failing to act – see below.) As indicated, my chief conclusion will be that these arguments fail because, for all we know, there can be non-arbitrary divine choices even if there is no best act of creation. Whether or not this shows that God can act in a non-arbitrary way depends on non-contingent truths about what is valuable, and crucially, about hard to discover truths about the ordering of possible situations as more or less valuable. My conclusion, therefore, is the modest one that for all we know there are more cases of non-arbitrary action than permitted by Rowe’s requirement that God perform the best. My reply to the Argument from Arbitrariness is not intended to explain how God is free. If God is the kind of being that has reasons for acting, then these reasons constrain God as much or as little as if God must create the best.

Why Bother with the Argument from Arbitrariness? Rowe argues that if God performs act X when there is some better act Y that God could perform, then God is not ‘unsurpassingly good,’ because it is possible there is an agent that acts better than God (Rowe 2004: 97). For that reason critics have concentrated on this moral requirement (e.g. Howard-Snyder and Howard-Snyder 1994, Hasker 2005).3 But Rowe’s is not the only or even the most serious problem concerning how God acts. For, on the one hand, even if Rowe’s argument is I call it this because not ignoring these excellences would be a case of vicious perfectionism, ‘making the best the enemy of the good.’ 2 We could use the term ‘perfection,’ but perfection seems to imply the highest degree of some quality such that the higher the degree the better, so I use the term ‘excellence’ instead. By an excellence I mean a property that contributes unconditionally to the value of an act or an agent, as the case may be. By an unconditional contribution I mean one that does not presuppose some defect. I am assuming that God has many excellences, but we may reconcile this to a weak doctrine of divine simplicity by saying that all these excellences are consequences of the simple divine essence. 3 Howard-Snyder and Howard-Snyder were responding to Rowe (1993), not, of course, to (Rowe 2004). 1

On the Argument from Divine Arbitrariness

343

persuasive we might react by saying that God lacks the impossible excellence of being unsurpassingly good (namely, acting at least as well as any agent could act), but has the possible excellence of being as good as possible for a being of that kind. Rowe would then have made a valuable contribution to our understanding of God but not provided a serious difficulty for theism. On the other hand, there might, for all we know, be two or more equally good (and equally reasonable) possible divine acts that are better than any others. In that Buridan’s Ass situation, an arbitrary choice between them would be compatible with God’s ‘unsurpassing goodness,’ but nonetheless an arbitrary act. Hence, a problem remains, provided the No Arbitrary Choice principle holds, namely that God does not perform an arbitrary choice. Here I stipulate that to choose arbitrarily means performing one act rather than one or more other acts, without having a reason for the performance of that act rather than the others. No Arbitrary Choice implies the following Best Act principle: God will perform the unique best overall possible act if there is one. Here, and throughout the article, I shall, with one exception, assume that inaction is deemed a possible act. So if all actions in the strict sense lead to an outcome worse than doing nothing then God will not act. The exception concerns the situation in which God has no reason to act and no reason not to act. In that case I take action but not inaction to be arbitrary. So God does not act. In this article, I assume No Arbitrary Choice holds and examine the consequences. I defend the position that even if there is no unique best overall possible act, there might, for all we know, be one that God can choose without being arbitrary. This establishes the deductive invalidity of the Bold Argument and undermines the third premise of the Modest Argument.

Situations that Suggest an Arbitrary Choice I have already mentioned the Buridan’s ass situation in which two possible acts are judged by the chooser to be equally good.4 How do humans choose in such cases? Usually it just occurs to the person to choose one way. It would then be irrational to waste time deliberating further. Instead, we sometimes toss a coin. In that case there is a prior arbitrary choice of what heads stands for (right hand bale or left hand one). Both of those procedures result in an arbitrary choice, and if God does not choose arbitrarily then neither of them is available to God. There is, however, a third method: ignore the characteristics that are irrelevant to the decision and decide on the basis of the remaining ones. I start with an example that I owe to Keith Lehrer. It concerns the choice between two wines. Suppose I get really annoyed whenever someone comments on the color of red wine on the grounds that it is usually irrelevant where the wine is on the scale from red to purple provided it is clear. One evening I have a choice between two wines indistinguishable as regards taste, smell 4 I say ‘judged’ because I am not here restricting attention to the divine case. I assume that God judges infallibly and hence in the divine case the phrase ‘judged by the chooser to be’ may be deleted.

344

P. Forrest

and clarity. One is of a distinctly purplish hue, the other redder. I choose by ignoring the characteristics that are of no help deciding and consider whether I would prefer to drink something red or something purple, even though I do not think the color important. Having a preference for purple, I choose that wine. Notice that in this situation the choice is like that of Buridan’s ass in being a choice between equally good outcomes, but unlike Buridan’s ass in that there is a reason for performing one of the two acts. I have used this example to illustrate the idea that there can be a choice between equally good acts, but nonetheless a reason for performing one of them: in this case color is just a ‘tie-breaker.’ It might be objected that this shows the color really is an excellence, although not a very important one, so I concede that this and similar Buridan’s ass cases only serve to illustrate the idea, not to show that it is correct. I now consider the more complicated cases of two or more optimal acts. By an optimal act I mean one than which none is better. So two or more equally good acts better than all others are optimal. But there can also be multiple optimal acts in cases in which the ordering is not linear.5 Non-linearity occurs if there are acts x and y that are non-comparable, that is, x is neither less good than, nor better than, nor equally good as y.6 If such a situation arises with the evaluation of acts, then an agent may choose to make additional comparisons so as to restore linearity. Often, however, the choice of how to linearize the order is itself arbitrary. Hence, additional comparisons cannot always remove the threat of arbitrariness.7 A human example of non-linearity is the choice between family and vocation. Politicians – especially those whose constituencies are a long way from the Capital – tend to neglect their spouse and children, if they have them. That is a choice they make, and if their motivation is not primarily a lust for power, then we may be sympathetic enough to say that in some respects it is better to choose family and in others better to choose vocation. In such cases there is no overall better act, but the acts are not equal either. For if they were equal, then adding some further slight consideration would resolve the dilemma by showing one of the choices to be better (e.g., a new air service that reduces the time to fly home from five hours ten minutes to five hours). Sometimes the dilemma posed by two non-comparable acts is resolved by making a third choice better than either of the two. Maybe the politician is martyred for a cause worthy enough to leave (by death) both vocation and family. If not, then there are two or more optimal acts that are not of equal value. In such cases we may decide between them without asserting that a trivial extra consideration makes one of the outcomes better. Instead the trivial consideration operates in a different fashion, by making one of the two choices more reasonable. We can decide by ignoring the 5

Technically, the equivalence classes of choices form a partially ordered class that is not totally ordered. Here two choices are equivalent if they are equally good, like the bales of hay in the Buridan’s ass example. The partial ordering ≤ is defined by [x] ≤ [y] just in case y is either better than x or x and y are equally good, where [u] denotes the equivalence class of u. 6 In this context, I do not restrict the comparison to good acts: one bad act is better than another if it is less bad. 7 There is sometimes an objective procedure for linearization of the non-linear ordering. It occurs if the derived relation of being neither better nor less good is itself an equivalence relation. In that case we may linearize by treating that derived relation as precisely the equal-value relation. This procedure will in general fail because that derived relation is not always transitive.

On the Argument from Divine Arbitrariness

345

excellences that cannot be compared and concentrating on those that can be. This may seem flippant, but it is often the only non-arbitrary way to decide. Thus, the politician might use a dislike for the Capital’s cold winters as a reason to prefer family to vocation. Reluctantly, I concede that dislike of cold is subjective. Nonetheless, given the subjective preference (for Darwin over Canberra, or Miami over Washington, D.C.), this is a trivial but nonetheless non-arbitrary way of choosing. It does not imply, however, that the chosen outcome of family plus warm winters is a better situation than vocation plus cold winters. Likewise, the divine choice between optimal acts of creation might be resolved by ignoring all the excellences that provide no basis for a choice and concentrating on those that do, even if they are of less significance. For example, suppose, contrary to my belief, that God can create only one universe and so has to choose between the various laws of nature that would result in a universe suited to the evolution of rational animals who may become blessed.8 Suppose that each of these is an optimal act but they are non-comparable. Then God might decide by selecting the simplest laws on aesthetic grounds, even though these aesthetic considerations might be far less significant than others to do with the existence of rational animals. My repeated use of the phrase ‘as far as we know’ indicates the highly speculative nature of my enquiry. This is not because of problems with understanding the divine nature. The difficulty is with the evaluation of possible acts that strike us as noncomparable. I concede that maybe there is an objective ordering that we humans do not know. Because my overall purpose is to expand the range of metaphysical hypotheses on which God does not act arbitrarily, the speculative nature of the enquiry is, however, acceptable. Another source of arbitrariness is the one that Rowe has in mind, namely a sub-class W of possible divine acts that is linearly ordered but lacks an upper bound. (An upper bound is a possible divine act, not necessarily in W, that is as good as or better than any member of W.) For example, there might be a choice between a sequence of possible acts wn, n=1, 2, etc., such that wn+1 is better than wn and such that no act is better than all the wn. We cannot, however, always restrict the situation to the countable case. Indeed, we might even be considering a proper class of possible divine acts. I now provide another example to do with blessed beings to illustrate the principle of No Vicious Perfection when it comes to the situation that Rowe envisages. Suppose, as is plausible, that there is value in the cardinal number of blessed beings, value in variety, but disvalue in too much repetition of (near) indiscernibles. For simplicity, ignore other excellences, so we just have these three: the number of the blessed, variety, and lack of repetition. In this example I suppose God can create at least one instance of various types of universe that result in blessed beings.9 How 8

For the purposes of theodicy it is good to make as few controversial assumptions as possible. Hence, it is appropriate to consider both the case in which God creates just one, maybe infinite, universe and that in which God can create many of them. The chief difference is that in the latter case even the most fundamental laws of nature could vary. 9 Possible universes may be classified into types in finer or coarser ways. Here I am supposing there is some appropriate classification. I am also supposing that the class of all possible types of universe is a set rather than a proper class. (Compare David Lewis’ concession that there is a limit on the cardinal number of possible worlds Lewis 1986: 102–4.)

346

P. Forrest

many instances of each type should there be? Other things being equal, the more blessed beings the better. God cannot, however, maximize the number of instances because, I say, God cannot in a single act create a proper class of universes – only a set of them.10 And for any set there is one of greater cardinality. I shall suppose that the disvalue of repetition is outweighed by the value of there being more blessed beings. Hence, if, for ease of exposition, we consider only these three excellences, then there is no best act of creation, because all things considered more is better. Nonetheless, I claim, God has reason to perform the act of creation that provides the best combination of variety with lack of repetition (this might well be the act that results in precisely one instance of each type of universe that contains blessed beings), for God would ignore the excellence that cannot be maximized.

No Vicious Perfection The principle of No Vicious Perfection states that, in the absence of a unique overall optimal possible act, the agent should ignore some excellences and so perform the unique optimal act with respect to the remainder, if there is one. This principle presupposes the thesis that that there are distinct excellences. That we do distinguish excellences is common sense. For in many cases we judge that one situation is better than another all things considered, although not as good in some respects. For example, we often trade off the quality of the result with the amount of effort needed. An excellence is that in virtue of which some possible acts are better than others in the respect in question. I do not, note, assume that there is a numerical measure of the excellence. (If there is then we may assign a certain number of units – call them 'platos– to each possible excellence in such a way that the more 'platos the better the excellence.) The objectivity of the excellence should not be as controversial as it no doubt is, for a survey of disputed topics in the theory of values supports the thesis that the subjectivity so frequently remarked upon in our culture has in fact to do with the linearization of non-linear comparisons. For example, although they might defend the legality of abortion on demand, most ‘pro choice’ advocates are prepared to concede that some reasons for abortion, notably a preference for boys over girls or vice versa, are inadequate. Conversely, ‘pro life’ advocates grant that often a woman has a good reason for an abortion, even when they say that other considerations 10

Clearly there are some issues that need further investigation concerning how, if at all, proper classhood puts limits on divine knowledge and power. In this connection I note that O’Connor mentions the idea of God creating a proper class of universes (2008: 119). What I here reject is the thesis that God could perform a single creative act resulting in a proper class of universes. This is because I assume that that God can survey a class of outcomes of possible acts of creation in order to choose between them. Proper classes cannot be members of classes. So the possible act of creation cannot itself have the structure of proper class of universes, as it would if it were the sum of members of a class of universes. (This last step in the inference follows because ‘universe’ is a count noun not a mass noun, so any mereological sum of them has a natural structure as a class.) I am neither endorsing nor rejecting the thesis that God could perform a proper class of acts, for I am here considering a single divine choice about one act of creation. I note though that the strong doctrine of divine simplicity endorsed by classical theists implies that God performs only a single act and hence could not create a proper class of universes.

On the Argument from Divine Arbitrariness

347

trump that reason. I note that the supposed subjectivity here is a matter of how the different considerations are weighed up. I have yet to say which excellences are irrelevant. If there is no unique overall best act, No Vicious Perfection tells us to ask which is best if we ignore one excellence. The answer, if any, might depend on the excellence we ignore. Maybe there is just one excellence ignoring which would result in a best choice X. In that case the non-arbitrary choice is X. Maybe there are several excellences ignoring just one of which results in a best choice, but the best choice is the same, namely X, no matter which one of these excellences is ignored. In that case, once again the nonarbitrary choice is X. Or maybe ignoring one excellence gives us a best choice Y and ignoring another gives us a different choice Z. In that case we have to consider what happens if we ignore two excellences and repeat the above procedure by considering the results of ignoring any pair of excellences. The other situation that could arise is that in which there is no one excellence ignoring which we would obtain a best choice. In that case we likewise have to consider the result of ignoring two excellences. If in all these cases the best-ignoring-one-excellence is the same, then that is the reasonable act to perform. If there is no case where there is a best or if the best-ignoring-one-excellence is the not the same in all cases, then we repeat the process, but this time ignoring two excellences, and so on. Given the plausible thesis that the number of excellences is finite, this process will either result in a nonarbitrary choice or a situation in which there are no excellences left. The ‘no excellences left’ situation is perplexing because in it God has no reason to act in the strict sense of act, but also no reason not to act. In that case No Arbitrary Choice applies only to action and so God fails to act.

Objections and Replies I consider first an objection to No Arbitrary Choice and then second to No Vicious Perfection. We humans have freedom and so act in mysterious ways. Hence, it may be objected, there is no multiplication of mysteries if we assume that God acts in mysterious ways. To be sure, the details of God’s act are mysterious to us, but that is not the issue here anymore than it was for Leibniz when invoking the Principle of Sufficient Reason. (What is mysterious to us if there is a best is not that God performs the best but the description of the best.) What the objector is considering is the mystery of not performing the best, if there is a best or, more generally, the reasonable act if there is one. I have two responses to the objection. The first is based on the speculation that human beings act freely by choosing either to act reasonably or to act in accordance with non-reasonable impulses. God is not like that, not being subject to non-reasonable impulses. Hence, our excellence in being able to choose freely presupposes the defect of not being free in a different sense, namely free from non-reasonable impulses. Therefore, God lacks that excellence along with any other excellences that presuppose defects. It follows that the mystery of a divine act not in accordance with No Vicious Perfection would be a different mystery from that of human freedom. My other response is that even if we human beings are sometimes free to choose between equally reasonable acts there is an additional mystery in the divine case. For

348

P. Forrest

we may suppose that there will be an infinity of possible divine acts not excluded on other grounds. In the human case, where the choice is finite, if the acts are equally reasonable we should expect there to be equal probability for each act, but there is no unique natural way of assigning probabilities to the acts in an infinite class unless that class has some further structure. For example, the instruction to choose a positive integer less than 11 will, in the absence of other motives, result in a probability 1/10 for the choice of the number 1. But what about the instruction to choose a positive integer with no stated upper bound? Some further rule is required as to how to set about choosing, such as, ‘Toss a coin until you get Heads. The number of tosses is the positive integer you have chosen;’ or, more abstractly, ‘Let the probability of integer N be (½)N.’ This is a fine rule, but its application requires that the choices be only countably many and themselves linearly ordered as first, second, third, etc. The only application I can think of is the case in which God decides how many instances of a creation-worthy type of universe to create on the implausible assumption that God has reasons for creating only finitely many. I conclude that there are additional problems with divine arbitrary choices that do not arise with human ones. The first objection to No Vicious Perfection is that it implies that God might choose the less good over the better. This would occur if the most significant excellence has no maximum and so gets ignored. So whatever choice is made could be bettered. This would, I grant, be counter-intuitive if God was choosing between two possibilities only, one better and the other less good, but that is not God’s choice. The second objection is that that God would not ignore the excellences that offer no guidance but be confounded by them, and so not act at all. I disagree, but there is a fallback position, namely that God, realizing that not to create is less good than creating, would limit the divine power prior to creation in such a way that the acts of creation that can be performed would not differ with respect to irrelevant excellence. Strictly speaking, this would be a case of God performing a unique best act, namely one of self-limitation followed by a unique best act of creation. However, the consequences of these two acts are the same as a single act as described by No Vicious Perfection.11

Application to the Argument from Evil What, if any, difference does it make to the Argument from Evil if there is no best act of creation, but instead God performs the unique non-arbitrary act? Let us consider a version of the Argument from Evil based on the initially implausible premise that God would minimize suffering rather than maximize the balance of joy over suffering. An additional, but plausible, premise is that God could create in ways that allow some joy but no suffering at all, even though the resulting universe would perhaps be inferior to ours all things considered. The third premise is obvious: our universe contains suffering. No Vicious Perfection could be used to argue that the 11 The possibility of such divine self-limitation is heterodoxical and requires us to deny that God is essentially omnipotent (Forrest 2007).

On the Argument from Divine Arbitrariness

349

first premise is not so implausible. For there is an important question concerning joy and suffering. Is there, as I believe, one excellence, or two? If we combine joy and suffering on a single scale, then it has no maximum and, as George Schlesinger has argued, can be ignored (Schlesinger 1977: Chaps. 9, 10). In that case God’s act would largely be dictated by other excellences, including aesthetic criteria. If, however, there are two excellences, degrees of joy and degrees of absence of suffering, then the first cannot be maximized, but the second can be maximized by having no suffering. This would render considerations of joy largely irrelevant and put the emphasis on suffering, as in the above version of the Argument from Evil.12 I grant that pains and pleasures are not on the same scale, for we can have both simultaneously. But in my experience joy and suffering are not like that. The very fact that they exclude each other supports the thesis that they are to be measured as positive and negative quantities on the same scale and hence that Schlesinger is right.

Conclusions My chief conclusion is that even if there is no best act of creation there might well be a unique non-arbitrary act for God to perform, thus increasing the support for theism. Applying this to the Argument from Evil, I note that if joy and the absence of suffering are distinct excellences then there is a serious version of the Argument from Evil based on the prediction that God would eliminate suffering. This would greatly decrease the support for theism. I consider it more plausible, however, that joy and suffering are measured on the same scale.

References Forrest, P. (2007). Developmental theism: From pure will to unbounded love. Oxford University Press. Hasker, W. (2005). Can God be free?: Rowe’s dilemma for theology. Religious Studies, 41, 453–462. Howard-Snyder, F., & Howard-Snyder, D. (1994). How an unsurpassable being can create a surpassable world. Faith and Philosophy, 11, 260–268. Lewis, D. (1986). On the plurality of worlds. Blackwell. O’Connor, T. (2008). Theism and ultimate explanation: The necessary shape of contingency. Blackwell. Rowe, W. (1993).‘The problem of divine perfection and freedom.’ In E. Stump (Ed.), Reasoned faith (pp. 223–232). Cornell University Press. Rowe, W. (2004). Can God be free? Oxford University Press. Schlesinger, G. (1977). Religion and scientific method. Reidel.

12

Not quite irrelevant, however. I submit that for each individual there is a derived three-point scale, consisting of a positive, negative, or zero balance of joy over suffering, which God will act to maximize. To do so requires an afterlife.