The Myth of Counterfactual Dependence

At least since Alvin Plantinga’s and David Lewis’s work on modal metaphysics, philosophers have frequently appealed to the notion of counterfactual dependence to analyze concepts like causation and grounding. David Lewis also uses the concept to understand the asymmetry of time:

The way the future is depends counterfactually on the way the present is. If the present were different, the future would be different; … Likewise the present depends counterfactually on the past.

In the philosophy of religion the notion has been widely used to solve problems dealing with God’s relation to creation. Want to reconcile divine responsiveness to prayer with divine timelessness? Just say that what this means is that if we had prayed otherwise, or not prayed at all, then God would have done otherwise. Want to reconcile divine foreknowledge with creaturely freedom? Just say that if we had chosen otherwise than God foreknows, then God would have foreknown otherwise. In both cases, a merely counterfactual relation is invoked to cash out a kind of “dependence” of God on creation.

Despite these uses of counterfactual dependence by venerable philosophers, I maintain that counterfactual dependence is a myth because merely counterfactual dependence is not actually a kind of dependence. Genuine dependence is categorical, not conditional and a fortiori not counterfactual.

In the first place, consider the truism that correlation does not entail causation. That A correlates with B may suggest a causal relationship between them, but it does not and cannot by itself establish such a relationship. In general, a positive correlation between A and B is compatible with (a) A’s causing B, (b) B’s causing A, (c) C’s causing both A and B, and (d) sheer coincidence. The correlation by itself cannot tell us which of (a)–(d) is the case. Of course, when combined with other information, including other documented correlations, it may contribute to a cumulative case that strongly favors one of those options. Correlation can be evidence of causation, but it is not, and cannot, be causation. Nomically governed causal relations may imply correlations, but causation cannot be reduced to correlation. The same goes for any other asymmetric or anti-symmetric dependence relation (e.g., grounding).

Now consider a typical counterfactual (aka subjunctive conditional):

If A were the case then C would be the case.

What does this conditional tell us about the relationship between A (antecedent) and C (consequent)? It tells us that there is some kind of necessary correlation between A and C, i.e., there is no relevantly possible A-scenario that isn’t also a C-scenario. And that’s all it tell us. It doesn’t tell us why this correlation obtains. Is it because A grounds, causes, or in some way necessitates C? Perhaps. Or maybe it’s because C grounds, causes, or in some way necessitates A—and there is no other relevantly possible way A could obtain. That would also satisfy the conditional. Or maybe there is some third thing that grounds, causes, or in some way necessitates both A and C and does so in a way that rules out scenarios in which A obtains and C doesn’t. That would also work. In short, the conditional is compatible with C’s depending on A, A’s depending on C, and with neither depending on the other. The only thing a necessary correlation rules out is sheer coincidence.

Genuine dependence is categorical, not conditional. To say that B depends on A is not merely to say that some kind of conditional relation (e.g., If A then B) holds between them. It is to say something categorical about A and B. For example, to say that A causes B is to say something like this: A obtains or has obtained, B obtains or will obtain, and A brings it about that B. The notion of “bringing about” plausibly entails a conditional like If A then B or If A had not obtained then B would not have obtained, but the central claims are all categorical. That B depends on A tells us something about A, something about B, and something about a relation that actually and not just hypothetically obtains between then.

To wrap up this blog post, I want briefly to consider the two philosophy of religion applications of counterfactual dependence noted above. In each case the counterfactual analysis is compatible with two or more different and incompatible dependencies. Hence the counterfactual analysis fails to establish the actual dependence of one term upon the other.

Example #1: A timeless God’s responsiveness to prayer

If God is essentially timeless, then God cannot change. But then how can God respond to our prayers? As James 4:2 says, “You do not have, because you do not ask” (NKJV). In other words, God responds to our petitionary prayers and sometime does things because we pray that He would not have done had we not prayed. But if I’m free to pray or not, then it seems that whether and how God responds must be conditional upon whether and how we pray. Given the explanatory dependence of God’s response upon our prayers, it seems that God must wait and see how and whether we pray before He responds. But then we run into a problem with divine timelessness, for there can be no before–after sequence in a timeless God. And we also run into a problem with divine impassibility, for the actions of an impassible God cannot be explanatorily dependent on what we do.

Enter “counterfactual dependence”. In this interview with Eleonore Stump, Robert Lawrence Kuhn expresses the worry that a timeless God cannot be responsive to human actions. Stump responds starting at 7:22. She starts out by saying that divine responsiveness requires that God does what He does “because” of what we do, but she then cashes this explanatory relation out in a purely counterfactual way: “If you hadn’t prayed, He wouldn’t have done whatever in response to your prayer.” The problem with this response is that it doesn’t get us to God did this because we prayed. The counterfactual is, in fact, consistent with at least two different scenarios:

1. Prayer moves God: If you hadn’t prayed, God wouldn’t have done X. But you prayed, and God did X because you prayed.
2. God moves prayer: If you hadn’t prayed, God wouldn’t have done X. But God didn’t do X because you prayed. Rather, God eternally decreed and causally ensured both your praying and His doing X. Under no circumstances does He decree X in the absence of your prayer.

Stump’s response to Kuhn’s question assumes that the truth of the counterfactual implies scenario (1), but it doesn’t. Both scenarios are fully compatible with the truth of the counterfactual. Hence, its truth does not show that God is responsive to prayer, and it doesn’t do anything to reconcile divine responsiveness with divine timelessness. Stump has, unwittingly it seems, conflated a real dependence relation (“because”) with a merely counterfactual relation (If you hadn’t prayed, God wouldn’t have done X). She then observes that the counterfactual is consistent with divine timelessness and concludes that real dependence (divine responsiveness) is consistent with divine timelessness. But the problem lies with the first step. The counterfactual construal of divine responsiveness drops any real dependence of God’s actions on human actions out of the picture. That’s why it’s consistent with both God’s action depending on the prayer (scenario 1) and with the prayer and God’s “response” depending on God’s prior decree (scenario 2).

Example #2: Foreknowledge and creaturely freedom

If God is essentially omniscient, then God infallibly knows all truths. It follows that if there is a determinate truth about what I am about to choose, then God knows that. But then how can I be free (in a libertarian sense) to choose otherwise? If that truth, and God’s knowledge of it, is fixed independently of my choice, then it constrains my choice. So the only way that truth and God’s knowledge of it can exist and I still be free to choose otherwise is if that truth and God knowledge of it ontologically depend on my actual free choice. I argue as much in this recent interview with The Analytic Christian starting at around 23:17.

But many Christians from Ockham to Alvin Plantinga have supposed that this dependence relation can be cashed out in purely counterfactual terms: If I choose otherwise (than I am in fact about to do), then God would have always foreknown otherwise. The counterfactual asserts that there is a necessary correlation between God’s knowledge that I choose thusly and my actual choice. But it doesn’t (pace Ockham, Plantinga, et al.) establish that God’s knowledge depends in any substantive way on my choice. To see this, consider that the truth of the counterfactual is compatible with at least two different scenarios:

1. My choice brings about God’s foreknowledge: If I had chosen otherwise, then God would have foreknown otherwise. My actual choice brings about God’s having the foreknowledge He does.
2. God brings about my choice: If I had chosen otherwise, then God would have foreknown otherwise. But it’s not my choice that brings about God’s having that foreknowledge. Rather, God sovereignly decrees and causally ensures that I make the choice I do. God’s foreknowledge therefore depends on God’s causal activity, not on my actual choice.

In short, counterfactual dependence is too weak to reconcile divine foreknowledge and creaturely libertarian freedom (scenario 1) because it is fully compatible with the categorical denial of libertarian freedom implied by scenario 2.