Van Inwagen vs. Van Inwagen on Freedom and Randomness

Does God get do-overs?

Apr 29, 2024

On the issue of free will, one of the challenges faced by those who unequivocally affirm human freedom is that of distinguishing freedom from randomness. If our acts are necessitated, then we have no moral responsibility, say the advocates of freedom, to which their opponents reply that if our acts are not necessitated, then they are merely random, which also negates moral responsibility. Those who affirm free will and deny determinism must show, then, that their opponents have made necessity and randomness into a false dichotomy: freedom is neither.

One way the advocates of freedom can do this is to point out that the concept of randomness or chance does not properly apply to particular acts considered singly; randomness is the absence of a pattern, and patterns can only apply to several particulars considered collectively. Peter Van Inwagen argued along these lines in An Essay on Free Will (1983, pp. 128-129), but has since changed his mind. In “Free Will Remains a Mystery” (2002), he explicitly repudiates his earlier argument and argues that human acts are indeed random. He maintains his belief in freedom, however, and he states that he fervently hopes that something is wrong with his argument for the randomness of human acts (175).

Old Van Inwagen’s fervent hope is well-founded, but his argument is not. I shall defend Young Van Inwagen’s position: free acts are not random or chance events. Old Van Inwagen’s argument is based on a thought experiment involving the notion of God “replaying” the universe. The thought experiment begins with the supposition that an agent, Alice, chooses to tell the truth on a certain occasion on which she could have lied. She chooses freely; her choice is undetermined. Van Inwagen continues, “Now suppose that immediately after Alice told the truth, God caused the universe to revert to precisely its state one minute before Alice told the truth [. . .] and then let things “go forward again.” What would have happened the second time? [. . .] One can only say that she might have lied and she might have told the truth.” (171-172)

Van Inwagen goes on to consider what would happen if God “replayed” the universe in this way a thousand times. Since Alice is undetermined, Van Inwagen supposes that she would tell the truth in some cases and lie in others, and that a ratio of lies to truth-tellings would emerge, and that it is reasonable to expect that this ratio might approximate 50/50 after a sufficiently large number of replays. Van Inwagen concludes, “If we have watched 726 replays, we shall be faced with the inescapable conclusion that what happens on the 727th replay will be simply due to chance [. . .] Now, what holds for the 727th replay holds for all of them, including the one that was not strictly a replay, the initial sequence of events. [. . .] Therefore, an undetermined action is simply a matter of chance [. . .]” (173)

Van Inwagen characterizes this argument as “plausible” and “intuitive”. I say it is neither, and it misapplies probability theory to free choices. Let me begin with the point about probability. Probabilities cannot be assigned to free choices. If I have a jar of 100 marbles, some of which are black, and the rest of which are red, and while blindfolded I reach in and grab one, the probability that I will grab a black one is defined by the ratio of black marbles to red marbles. The probability has nothing to do with how many times I replace the marble I’ve grabbed and repeat the experiment; the ratio is what it is, and the same goes for the probability. But a free choice has nothing that is analogous to the number of marbles. There is nothing there, and that nothingness is metaphysical and not just epistemic. The outcome of the free choice is analogous to the color of the drawn marble, but there is nothing in the agent or the alternatives or anything else that corresponds to the ratio upon which probabilities can be calculated. The probability that an agent will freely choose one course of action over another is undefined.

Thus, using Van Inwagen’s example of Alice’s telling the truth or a lie, we have no basis for assigning a probability to either outcome. Supposing that God were to reset the universe in the manner described an arbitrary number of times, we have no reason to suppose that the percentage of times Alice told the truth would come closer and closer to 50 as the number of trials increased. If it did in fact turn out that way, we would have no reason to suppose that a further arbitrary number of trials would not yield a significantly higher or lower percentage. For that matter, supposing that God were to reset the universe an infinite number of times, probability theory not only gives us no reason to suppose that Alice would not make the same choice an infinite number of times, or never again make the same choice she made the first time; probability theory gives us no reason even to suppose that Alice’s making the same choice an infinite number of times is less likely than any other outcome.

Probability theory may be irrelevant for a more important reason. In attempting to apply probability theory to free choices, Old Van Inwagen is in effect denying the uniqueness of free choices. But every free choice is unique in a relevant sense. It may be that all free choices are type-identical in some sense, but no two free choice-tokens involve exactly the same agent in exactly the same situation. Young Van Inwagen’s argument against randomness stands.

Old Van Inwagen’s counterexample attempts to get around this. But what is meant by “revert” and “replay” in this context? We must be careful with our metaphors here. Van Inwagen carefully (and correctly) avoids supposing that God can change the past, strictly speaking. God returns the universe to a given physical state after Alice makes her choice; God does not (and cannot) go back in time and obliterate the morally relevant fact that Alice made her choice in the first place. So God can reset the universe, but Alice does not get a “do-over” in the fullest sense—and therefore neither does God. She has made her choice, and every time God resets the universe, Alice will act the same way. The options were open to her the first time; they are not open to her the second time. Her choice has become part of the fabric of the universe. Were God to reset the universe, God would then have something like Suarez’s “middle knowledge” of what Alice would do in the case of the particular choice in question. Were it not so, free choices would have no effective moral relevance. In accordance with his own perfect benevolence, God would simply reset the universe until Satan chose not to rebel, Adam and Eve chose not to eat the forbidden fruit, and Alice chose to tell the truth (or whatever outcome would best suit God’s Providential plan).

Who is to say that an account in which Alice chooses the same way every time is less “plausible” or “intuitive” than any other account? Use of these terms when describing a hypothetical situation that could not possibly have any bearing on human knowledge or action even if it were truly the case is even more irresponsible than the usual careless way analytic philosophers throw these words around. This careless appeal to plausibility and intuition, together with the misuse of probability theory, make Old Van Inwagen’s argument little more than an appeal to ignorance. My intuition tells me that once Alice has made her choice, she would make the same choice every time God reset the universe.

Here is another way to look at it: the appeal to reset universes may be compared to the way certain compatibilists have appealed to possible worlds. Compatibilists who are realists about possible worlds have argued that “could have done otherwise” means that an agent’s “counterpart” on some literally existing possible world actually did do otherwise (cf. Lewis 1973, pp. 39-43, 87). I do not understand why anybody would find this convincing. Arguments about what Alice’s “counterpart” would do in some other world are irrelevant; the moral significance of Alice’s acts depends on what Alice can do or could have done or will be able to do in this world. But supposing that the universe can be reset raises similar problems of identity. What makes Alice the same person in each reset universe once we suppose that she does not make the same choices? For that matter, what makes it the same universe? How can we distinguish between (a) resetting the universe and (b) obliterating the old universe and replacing it with a new universe whose initial conditions at the moment of creation are structurally the same as that of the old universe at some point in its past, but whose subsequent conditions differ?

Finally, if Old Van Inwagen were to insist that it is in fact the same universe, with the same Alice, and that Alice is in fact repeatedly facing the same choice (i.e., the same choice-token, and not just a type-identical choice), then I would insist that a fortiori, the truly intuitive conclusion is that Alice would freely choose the same way every time that choice-token is repeated, or to put it more accurately, Alice would repeatedly do what she decided to do once and for all the first time. But were God to reset the universe, although Alice’s choice would be the same, other events taking place at the same time as Alice’s choice might not be the same, namely those that involve genuinely random processes.

Old Van Inwagen feared that his thought experiment of the reset universe had proven that our choices were just manifestations of chance, but his notion of resetting the universe has instead proven the opposite, for it has led us to a difference between chance and freedom. Were God to reset the universe to some point immediately prior to some agent’s free choice, the free choice would turn out the same, but the chance occurrences might turn out differently.

So, yes, free will remains a mystery, but not because it is indistinguishable from randomness. Young Van Inwagen’s argument that free choices are not random remains tenable.

Logos and Liberty

Discussion about this post