The Essential Divine Perfection Objection to the Free Will Defense
Abstract: The Free Will Defense (FWD) holds that the value of significant free will is so great that God is justified in creating significantly free creatures even if there is a risk or certainty that these creatures will sin. A difficulty for the FWD, developed carefully by Quentin Smith, is that God is unable to do evil, and yet surely lacks no genuinely valuable kind of freedom. Smith argues that the kind of freedom that God has can be had by creatures, without a risk of creatures doing evil. I shall show that Smith’s argument fails—the case of God is disanalogous to the case of creatures precisely because creatures are creatures.
The deductive argument from evil holds that there is a logical incompatibility between the existence of an all-knowing, all-powerful and all-good God and the presence of evil. Various flavors of the Free Will Defense (FWD) argue that if significant freedom of will requires the possibility of acting wrongly and significant freedom of will has great value, then it is logically possible that God is fully justified in creating significantly free beings that sin. Significant freedom here is the freedom to choose between a permissible and an impermissible action.
A powerful objection against this argument is that traditional theism does not attribute to God the possibility of choosing an impermissible action—God is essentially morally perfect, and can no more do wrong than make a square circle. But God is not lacking in anything of value. Thus, either significant freedom is not of value, and the FWD is undercut, or significant freedom does not require the possibility of acting wrongly, and the FWD is again undercut. This is the “essential divine perfection” objection to the FWD: God can be essentially perfect, both morally and in every other way, without being able to do wrong, so why can’t he create creatures that are like that?
The best developed version of the essential divine perfection objection has been given by Quentin Smith. I shall argue that Smith’s argument fails, and that its failure makes it unlikely that any argument along the lines of the above sketch works. Thus, the Free Will Defense survives the essential divine perfection objection. I will first sketch Smith’s argument, and then respond to it. I will then consider two objections to my response, and end with some general considerations.
Henceforth, page number references will refer to Smith’s work.
Smith begins by distinguishing three kinds of freedom that he accuses the proponents of the FWD of conflating:
A person is externally free with respect to an action A if and only if nothing other than (external to) herself determines either that she perform A or refrain from performing A. […]
A person is internally free with respect to an action A if and only if it is false that his past physical and psychological states, in conjunction with causal laws, determine either that he perform A or refrain from performing A. […]
A person is logically free with respect to an action A if and only if there is some possible world in which he performs A and there is another possible world in which he does not perform A. A person is logically free with respect to a wholly good life (a life in which every morally relevant action performed by the person is a good action) if and only if there is some possible world in which he lives this life and another possible world in which he does not. (p. 149, italics added)
Now the question is what sort or sorts of freedom are of “metaphysical worth”. God does not have logical freedom, and this follows from his “maximal greatness, which includes the property of being wholly good in each world in which he exists” (p. 151). Consequently, “[i]t is false that it is intrinsically better to be logically free with respect to a morally good life than logically determined” (p. 151).
God has internal and external freedom with respect to a wholly good life, but not logical freedom with respect to it. This shows that internal plus external freedom does not entail logical freedom. Smith asks whether there is a relevant difference between God and creatures such that the entailment would go through in the case of creatures but not in the case of God, and cannot find any. What is of value in the case of God is having internal and external freedom with respect to a wholly good life, and Smith thinks a creature could also have internal and external freedom while lacking logical freedom. Nor would such a creature’s freedom be any less valuable than that of a creature who had all three kinds of freedom, since the case of God shows that internal and external freedom is not less valuable than having all three sorts of freedom. And if God exists, then he should have created only creatures that have internal and external freedom but are logically determined to lead a wholly good life, in which case there would have been no evil. But there is evil, and hence, Smith concludes, God does not exist.
I shall offer a simple argument, however, that external freedom with respect to a wholly good life plus creaturehood entails logical freedom with respect to a wholly good life. If so, then God could not create persons logically determined to lead a wholly good life but who were nonetheless externally free in respect of a wholly good life, and so Smith’s objection to the FWD and the atheological argument it supports both fail.
It will turn out that there is one place in my argument that can be questioned. To support that place, I shall argue that the definition of internal freedom should be amended in a certain plausible way, and once amended in this way, one will have an even better argument that creaturehood plus internal and external freedom in respect of a wholly good life entails logical freedom in respect of a wholly good life.
I shall show that if a creature lacks logical freedom with respect to a wholly good life, then she either lacks internal or external freedom (or both) with respect to a wholly good life, though under one set of controversial assumptions, I shall have to modify Smith’s definition of internal freedom. Consequently, being a creature and having internal and external freedom in respect of a wholly good life entails having logical freedom in respect of a wholly good life.
The initial form of my argument is very simple. If Patricia is a creature who lacks logical freedom with respect to a wholly good life, then by Smith’s definition either it is a necessary truth that if Patricia exists, Patricia leads a wholly good life, or it is a necessary truth that if Patricia exists, Patricia does not lead a wholly good life. For concreteness, take the first case: that Patricia exists entails that Patricia leads a wholly good life. (The other case can be handled in exactly the same way, and the case I am considering explicitly is the one that is more relevant to the problem of evil.) Then, that God creates Patricia entails that Patricia exists. Therefore, that God creates Patricia entails that Patricia leads a wholly good life. But surely that means that Patricia is determined to lead a wholly good life by something external to her, namely by God’s creating her. Hence, she is not externally free with respect to leading a wholly good life.
Observe that this argument shows clearly the difference between the cases of God and Patricia: it is because Patricia is created that she lacks external freedom. God is not created, and so he does not lack external freedom.
Notice, too, that as yet internal freedom has not been needed for the argument. But there is a possible complication to the argument. For it may be that God does not directly create Patricia, but that Patricia is caused to exist by her parents, and it is only some more remote cause, say the Big Bang, or Adam and Eve, or a unicellular proto-organism, that is directly caused by God.
As long as the causal chain between God and Patricia is deterministic, we can still unproblematically say that God creates Patricia, and the argument above continues to work. But what if some link in the chain is indeterministic? Well, as long as there is a later stage in the chain at which determinism reasserts itself, we will be able to identify a cause C in the chain that deterministically causes Patricia to exist, and external freedom will be violated for the same reason as before—C will determine it to be the case that Patricia leads a wholly good life.
One last case remains, and this last case will turn out to be the most troublesome. It may be that the causal chain is indeterministic all the way up to Patricia—no matter how close to Patricia in the chain one goes, it is not determined that Patricia will come to exist, and even the conjunction of the descriptions of all the items in the chain will not determine it to be the case that Patricia comes to exist. Patricia’s coming-to-exist is something stochastic. Maybe there is, right at the end of the causal chain, some indeterministic process such that either Patricia comes to be or Fred comes to be (or maybe the other option is that nobody comes to be). Moreover, this stochastic process is not under God’s control, since it were, then Patricia’s existence would be determined by God’s controlling the process.
What I have shown so far is that if Patricia is a creature without logical freedom in respect of leading a wholly good life, then either Patricia lacks external freedom or else she is the immediate outcome of an indeterministic process that God is not controlling. The latter is an odd hypothesis. It is one that puts into question whether we can genuinely call Patricia a “creature” in the full sense.
Suppose that we agree that it is a conceptual truth about divine sovereignty that everything except possibly creaturely free choices is under divine control. Then the only way we could have the above situation is if Patricia’s coming into existence is directly caused by a creaturely free choice. But if so, the choice that was in fact made by the creature or creatures determines Patricia to exist, and hence entails that Patricia leads a wholly good life, again contrary to Patricia’s external freedom. So on a plausible assumption about divine sovereignty, we have plugged the gap in the response to Smith.
Suppose now that we think that divine sovereignty does not require that every non-agential process be under divine control, and that Patricia comes to exist from a non-agential process P not under divine control. We can still say this: P’s causing Patricia’s existence is causally prior to Patricia’s life. Moreover, that P caused Patricia’s existence entails that Patricia’s life was wholly good. Hence, once again, we seem to have identified something external to Patricia, namely P’s causing Patricia to exist, that determines Patricia’s life to be wholly good. If this argument works, once again we have shown that Patricia lacks external freedom.
But perhaps P’s causing Patricia to exist is not really an event wholly external to Patricia, and perhaps the definition of external freedom was concerned with things wholly external to the agent. Maybe the event P’s causing Patricia to exist consists merely in P’s occurrence plus Patricia herself. I suspect this is mistaken—Patricia does not seem to me to be a part of event of P’s causing Patricia to exist. At most, P’s causing Patricia to exist consists in P’s occurrence plus the Patricia’s existence (or the concrete state of affairs of her existing). But insofar as Patricia’s existence might be thought to be a part of her on some metaphysical views—namely Thomistic ones—perhaps this would still make P’s causing Patricia to exist be external to Patricia.
Suppose we persist in considering the hypothesis. I will now shift to considering internal freedom. Suppose first that Patricia makes the first morally significant choice—i.e., the first choice between a permissible and an impermissible action—of her life at some time after her coming into existence. Let S be a complete description of Patricia’s psychological and physical states at the moment of her coming into existence. Then, that Patricia satisfies S entails that she will lead a wholly good life, since that she satisfies S entails that she exists, and that she exists entails that she leads a wholly good life. Hence, it seems we have internal determination here, and thus Patricia lacks internal freedom.
Maybe, though, internal determination would require not that Patricia’s satisfying S should entail that Patricia will lead a wholly good life, but that anybody’s satisfying S should entail that that person will lead a wholly good life. Even if so, it might, for instance, be that psychological and physical states are tropes or accidents, and that nobody but Patricia could thus be in the same state as she is in, since nobody but she could have numerically the same trope or accident (Helga could have a trope or accident just like Patricia’s, but it would be numerically different). And if so, then having these-and-these states will entail acting rightly, since only Patricia can have these-and-these states, and once again internal freedom would be violated.
But suppose that states are not tropes or accidents, but are more like descriptions. And suppose that internal determination would indeed require that anybody’s satisfying S should entail that that person will lead a wholly good life. This leads to an interesting question. If Patricia is a person that is logically determined to lead a wholly good life, and Helga is another person who has the same initial psychological and physical state as Patricia, does it follow that Helga also logically determined to lead a wholly good life? If yes, then Patricia indeed lacks internal freedom, because anybody’s starting in state S entails that that person will lead a wholly good life. But among psychological and physical states there are dispositional states that ground whatever facts there may be about how the person would behave under different circumstances. It seems that if Patricia is logically determined to lead a wholly good life, this is reflected in her dispositional psychological and physical states in such a way that anybody else who has the same psychological and physical states will share that logical determination—she has the dispositional state of being such that in any set of possible circumstances she will act well, after all. It would be really odd if logical determination to lead a wholly good life were not supervenient on the psychological and physical state of the person. Whence the source of the logical determination, then? Does the haecceity of the person in some mysterious way predetermine how the person must act? (This kind of view would seem to share the disadvantages of Molinism without Molinism’s advantages.) We shouldn’t take this route—we should just say that Patricia lacks internal freedom.
But all this presupposes that the first morally significant choice was not made at the first moment of Patricia’s existence. For beings that think and choose discursively, this can be taken for granted. But, for aught that we know, it is possible for there to be beings that do not need to think before they choose. They can just choose in the first moment of their existence. Aquinas thinks angels are like that. (Cf. Smith’s discussion of Swinburne on p. 152.)
Where are we at, then? We have Patricia, a person that comes into existence via a non-agential stochastic process not under God’s control, who makes a morally significant choice in the first moment of her existence, and whose existence is not determined by anything external to her. To get to here we had to make a controversial assumption about divine sovereignty—viz., that it is possible to have a non-agential stochastic process that God is not controlling—and a controversial metaphysical assumption that that process’s causing Patricia to exist could fail to be external to Patricia. Moreover, Patricia is a non-discursive finite chooser, since she is able to choose in the very first moment of her existence. A atheological argument that depends on all those assumptions working out will not be very powerful. But I have one last move left.
Smith’s definition of internal freedom talks of the action as not determined by past psychological and physical states. Now we have two possibilities. Either causal influence requires the elapse of time or it does not. If causal influence requires an elapse of time between cause and effect, our argument appears complete. For it seems to be a conceptual truth that the choice of a finite agent—or at least a choice subject to moral evaluation—is influenced by the agent’s beliefs (and on some, more controversial views, also desires). But if Patricia makes a choice in the first moment of her existence, which is the one remaining case to be considered, and her choice is influenced by her beliefs, then the causal influence here cannot involve any elapse of time—she couldn’t have had beliefs before she existed, after all! So on the view on which causal influence requires an elapse of time, the remaining problematic case turns out to be impossible.
Suppose now that causal influence does not require the elapse of time. (Theists who think that God is outside of time are going to have to agree.) In that case, Smith’s definition of internal freedom is simply mistaken. If my action is causally determined by a psychological compulsion simultaneous with my choosing, that action should no more count as internally free as when the compulsion preceeds my choosing. (Of course, we could take Smith’s definition as stipulative, in which case it couldn’t be questioned, but then the objection would be that what is valuable is not specifically internal freedom in Smith’s sense, but internal freedom in the sense I will give below.) Whether time elapses between the determining psychological and physical states and the choice seems irrelevant. What is relevant, of course, is that the determining states not include the specific choice made.
I propose a simple revision to Smith’s definition: Replace pastness with causal priority. Thus we will say that a person is internally free with respect to an action A if and only if it is false that those of his physical and psychological states that are causally prior to his choice, in conjunction with causal laws, determine either that he perform A or refrain from performing A.
And now the argument can be completed. For let S be the complete description of those of Patricia’s physical and psychological states that are causally prior to her first morally significant action, which, recall, occurs at the first moment of her existence. There will be such states, since morally significant choices are causally conditioned by beliefs. Patricia’s satisfying S will entail Patricia’s making the right choice now, and indeed her making the right choice in every morally significant choice situation in the future. As before, questions arise. Are states trope-like, so that nobody but Patricia could have numerically the same states, in which case lack of internal freedom follows? Are states description-like, so that the question is whether it would be true for Helga that her satisfying S would entail her making the right choice? And again, the same considerations apply. Psychological states include dispositional properties, and hence Helga would seem to have the dispositional property of acting rightly in every possible situation. And if having the property of leading a wholly good life essentially does not supervene on the state of the being, then the property does become mysterious indeed. And what confidence can we have that a being could have such a non-supervenient modal property, even if all the preceding problems could be solved?
It is worth seeing how this argument goes through in the one case that Smith describes of what a creature that has internal and external freedom but lacks logical freedom would be like. Smith follows Swinburne’s idea that if God is omniscient and not causally influenced, then God will necessarily choose rightly. Smith argues that omniscience is not required—all that’s needed is to know what is morally required in every circumstances one would be in. Thus, Smith imagines a mind x that satisfies the following two conditions:
“[(i)] for each possible world W in which x exists, and for each circumstance in which x is faced with a moral choice, x knows all the factual and moral truths he needs to know to make a correct choice.
“[(ii)] This mind x is neither causally determined nor causally influenced by any external or internal factors.” (p. 153)
Smith then argues that on Swinburnean assumptions, this mind couldn’t do wrong.
Notice, first, an immediate problem with this setup. If the mind is not influenced by any internal factors, then it is not influenced by its knowledge of the factual and moral truths in point (i). So if (ii) holds, (i) seems irrelevant. But of course (i) is relevant. The morally upright person makes choices for reasons that she believes to be morally significant. Her beliefs certainly causally influence her action—otherwise why would she choose to heal rather than to hurt, say? Thus, (ii) is incompatible with moral uprightness.
Perhaps I am being uncharitable here. Maybe (ii) means to say that x is not influenced by external or internal factors other than the knowledge in (i). Fine. But then it seems that this creature’s actions are determined by her state, specifically her state of knowledge and her lack of internal temptation, and so the creature lacks internal freedom.
And so Smith’s example fails.
I. Entailment versus determination. At several points in the argument in the previous section, I made use of the following move: C is causally prior to E, and C’s occurrence entails E’s occurrence, so C determines E. For instance, God’s creating Patricia is causally prior to Patricia’s wholly good life and entails that wholly good life, so God’s creating Patricia determines Patricia to lead a wholly good life.
This move might be questioned. One might insist that there is more to C’s deterministically causing E than just C’s causing it and the occurrence of C entailing E. But, first, I do not need to claim that deterministically causing just was identical with causing plus entailment. All I need is that causation plus entailment logically entails determination—but it might be that there is something more to determination.
Second, at least in the above case, it is hard to deny determination. Presumably, that Patricia is logically necessitated to act rightly follows from her nature. She has a certain nature, and God has created Patricia as having that nature. But surely then God has determined her to act rightly. Suppose that there is a possible particle whose nature is to be repelled from all other particles of the same type. Then by creating the particle in that nature, God has determined it to be repelled from all other particles of the same type.
There is a variant of this objection to my argument, found in a remark Pruss makes in response to Gale’s argument that Molinism is incompatible with creaturely freedom. Gale had claimed that if God knows everything that a person would do if created, then by creating the person, God has brought about all of the person’s actions, and thus canceled the person’s freedom. Pruss counters, however, by talking of Anscombe and double effect, saying that “a person counts as bringing about a foreseen effect if and only if that person intends that effect.” Thus, God brings about the person’s actions only if he creates the person in order that she perform these actions. One might likewise say that God’s intentional creation of Patricia only counts as determining her actions when God creates Patricia in order that she perform those actions.
However, if Smith were right that creaturely external and internal freedom without logical freedom is possible, and if he is right that a perfectly good God would create creatures logically determined to live wholly good lives, then surely such a God would intend that these creatures lead wholly good lives. For on Smith’s view, God’s goodness would require God to create creatures that lead such lives. God, being morally perfect, would surely have to act out of the requirements of his goodness, and not merely in accordance with these requirements, and hence God would have to intend to create creatures that lead such lives. But then God would indeed be bringing it about that the creatures live wholly good lives; he would be determining it in the intentional sense.
II. Does God have internal freedom? To get out of one particularly troublesome case—that of a person who makes a choice at the first moment of her existence, and who came into existence via a non-agential stochastic process—I had to modify the definition of internal freedom. But it is not clear that God’s activity satisfies this modified definition. For it seems that God’s beliefs and psychological states cause him to act rightly. Hence, it seems that on the modified definition, God lacks internal freedom. But if God lacks internal freedom, then just as Smith argued that logical freedom was not valuable, so too internal freedom would not be valuable. And Smith could then conclude that only external freedom is valuable, and a creature could have external freedom while having neither logical nor internal freedom, which would also undercut the FWD.
First, however, note internal freedom was brought into the argument only on the assumption that some earlier arguments concerning external freedom failed. If these arguments were sound, then lack of logical freedom in a creature entails lack of external freedom, and so the proposed modification of Smith’s argument would also fail.
Second, I can make the following move. Modify the notion of external freedom as follows. A person is externally free with respect to an action A if and only if every C that causes either that she perform A or that she refrain from performing A satisfies at least one of the following two conditions: (i) C is not external to the person, or (ii) some locus of indeterminism in the causal chain from C to A is not external to the person. In other words, a person who does A is not externally free provided that there is a cause C of the person’s doing A, and this cause is external to the person, and all loci of indeterminism for this chain are outside of A.
Here, I say that D is a locus of indeterminism for a causal chain from C to E provided that D is causally relevant to E’s resulting from C, and D’s causal influence on the chain from C to E is indeterministic. For instance, suppose that a neurologist is successfully using electrodes to induce an action in Jones, but some of Jones’ neurons are indeterministically countering that attempt in a way that could have been efficacious. In this case, the neurons’ activity is causally relevant to the chain from C to E—the activity might have countered the action-induction. Moreover, the locus of indeterminism is in Jones—the neurons responsible for the indeterministic activity are in Jones. Therefore, Jones does not count as externally unfree by either Smith’s original definition or by my modified definition. Jones’ action is caused by an external cause, but the causation is indeterministic, and at least one locus of indeterminism is inside Jones.
Incidentally, I suspect we would hesitate to call Jones genuinely free in a case like this, and this may show that there is something wrong with equating metaphysically valuable freedom with internal and external freedom. But that is a problem for Smith, not for me. All I need in the present argument is that external freedom is necessary for metaphysically valuable freedom, which is very plausible, at least to those who have libertarian intuitions.
The difference between my definition of external freedom and Smith’s is that my definition recognizes that external freedom can be removed by indeterministic causation as long as the locus of indeterminism is outside the person. It shouldn’t matter vis-à-vis the agent’s freedom whether some process leading to an effect in her is deterministic or not, unless the indeterminism in question is something she is in some way responsible for, and she is only going to be responsible for it if the locus of indeterminism is in her (some say that not even then, but that is a different question).
Now, once we assume the modified version of the definition of external freedom, we see that Patricia is not externally free even if she comes into existence due to an indeterministic process. For the process is causally prior to her existence, and her existence is causally prior to her life being wholly good (she can only act because she exists). The chain leading to her actions being morally good is indeterministic, but the locus of indeterminism is outside her—for the indeterminism consists solely in whether the process actually produces Patricia (or someone else or nobody at all), and the locus of that indeterminism is surely outside her. (It is not like Patricia struggles for existence prior to existing, in the way that Leibnizian concepts of monads are said to clamor for existence in the mind of God.) Thus Patricia is not externally free in respect of leading a wholly good life if she is not logically free in respect of it.
Hence, even if we do not require internal freedom for having metaphysically valuable freedom, as long as we require external freedom and allow my modified definition, my response to Smith works.
A third move I can make to the objection that on my modified definition of internal freedom God is not internally free is to invoke the doctrine of divine simplicity. If the doctrine of divine simplicity holds, there are no causal relations within God, and in particular his actions are not causally influenced by his beliefs—rather, his will is identical with his understanding. One difficulty with this move is that as a response to Smith, it would require defending the claim that the only possible being who can be like that. One route to such an argument would be to argue that lack of internal causal relations is only possible in a being that is pure actuality in the Thomistic sense, and only God can be pure actuality.
Quentin Smith’s version of the essential divine perfection objection to the FWD fails. It fails because, unless one makes some controversial assumptions, external freedom plus creaturehood entails logical freedom. These assumptions included the claim that C’s causing x to exist is not external to x, and that either God lacks sovereignty or sovereignty is compatible with non-agential indeterministic causation not under God’s control. Significant further work was needed to wrap up the argument in the case where these controversial assumptions were accepted.
It seems likely that the basic argument made against Smith’s version of the essential divine perfection objection will apply to any version of that objection. For the crucial part of the argument was that if God creates us to lead wholly morally good lives, and upon being created we cannot but lead wholly morally good lives, then God’s act of creation cancels our freedom. This is very plausible.
 For the two best-known versions, see Alvin Plantinga, The Nature of Necessity, Oxford (1974), Chapter 9 and Robert M. Adams, “Middle Knowledge and the Problem of Evil”, American Philosophical Quarterly 14 (1977), 109–117. For one version agnostic of middle knowledge considerations, see Alexander R. Pruss, “A New Free-Will Defense”, Religious Studies 39 (2003), 211–223.
 Quentin Smith, Ethical and Religious Thought in Analytic Philosophy of Language, Yale (1997), pp. 148–157.
 It is important for this situation that things not be as follows: An indeterministic event occurs, and if it goes one way, a deterministic process leading to Patricia’s existence occurs, and if it goes the other way, a deterministic process leading to Fred’s existence occurs. For in that case we could just look at the deterministic process leading to Patricia’s existence and notice that it negates external freedom.
 Cf. Robert C. Koons, “A New Look at the Cosmological Argument”, American Philosophical Quarterly 34 (1997), 193–212, Section 8.11.
 Richard Swinburne, The Coherence of Theism, Clarendon (1977), p. 146.
 Pruss, op. cit., Section 5.
 Richard M. Gale, On the Nature and Existence of God, Cambridge (1996), pp. 152–168.
 Cf. Pruss, op. cit., Section 5.
 I am grateful to Tedla G. Woldeyohannes for pointing out that divine simplicity might be relevant to Smith’s argument.
 I am grateful to Vlastimil Vohánka for comments on a draft of this paper.