Introduction
Molinism claims that: (i) there are counterfactuals of creaturely freedom (CCFs), and (ii) God knows all true CCFs prior to creating anything. In this article, we will focus on (i). A CCF is a non-trivially true, contingent proposition of the form C □→ A, where C is a maximally specific description of the contingent circumstances of a choice, and A is a report of a creaturely free action in those circumstances.Footnote 1 The non-triviality clause rules out cases where even many non-Molinist libertarians would say the conditional is true. The simplest such cases are where C and A are actually true.Footnote 2 (We adopt the common convention of calling ‘C □→ A’ a counterfactual even when C is not contrary to fact.)
It is a standard and much-discussed objection to Molinism that when C □→ A is non-trivially true, it is ungrounded.Footnote 3 David Manley (in conversation) raised a very interesting question that brackets the grounding question: do CCFs have explanations? A negative answer would increase the costs of Molinism. First of all, if a Principle of Sufficient Reason (PSR) that implies that all contingent truths have explanations is true,Footnote 4 and if non-trivially true CCFs are contingent and do not have explanations, then there are no non-trivially true CCFs, and Molinism is false. Second, even if no such PSR holds, it counts against a theory that it posits additional explanatorily brute (i.e. unexplained) truths, and typical Molinists posit infinitely many such truths.
In this article, we shall give a conditional answer to Manley's question about the explainability of CCFs by arguing for the following thesis:
(1) e(C □→ A) if and only if C □→ eA,
where (i) C □→ A is a CCF, (ii) C is false, and (iii) e(S) says that the proposition S has an explanation. This conditional is sufficiently strong to show that libertarians cannot employ the PSR to argue against Molinism. For the libertarian either does or does not hold that free actions require explanations. If free actions require explanations, then by (1), so do CCFs. But if free actions don't require explanations, then the libertarian must either claim that (i) in actual fact all free actions do have explanations, or that (ii) some free actions lack an explanation. Option (i) is extraordinarily unlikely: it's unlikely that all free actions just happen to have an explanation if free actions don't require an explanation. And option (ii) contradicts the PSR (assuming facts about free actions are contingent). A libertarian who thinks free actions are sometimes counterexamples to the PSR could, perhaps, still run the argument that Molinism multiplies the number of explanatorily brute facts. But the lack of an explanation of a CCF would then just be a function of the fact that its consequent would lack explanation, and so Molinism wouldn't imply any additional explanatory cost beyond what this libertarianism is already willing to pay.
The compatibilist, on the other hand, who thinks that if libertarianism were true, then free actions would violate the PSR, could deploy a PSR-based argument against the libertarian Molinism under discussion. But the direct target of such an argument is really libertarianism, not Molinism per se. It seems, therefore, that if (1) is true, then pursuing PSR-based arguments against Molinism leads to a dead end. The more promising line of inquiry will continue to focus on the grounding question.
We will begin our defence of (1) by offering a useful principle about counterfactuals and explanation. We will then go on to argue for each direction of (1), with and without the principle.
A principle about counterfactuals and explanation
Start with this counterfactual:
(2) Were I to eat many potato chips, I would be thirsty.
This is neatly explained by:
(3) Were I to eat many potato chips, I would be consuming a lot of salt.
This explanation would seem to hold as long as it's true that:
(4) Were I to eat many potato chips, my consuming a lot of salt would explain my being thirsty.
The explanatory connections here fall out of the following general principle:
(5) If p is possible and p □→ (q E r), then (p → q) E (p → r)Footnote 5
where q E r says that q explains r. To interpret (5), we make a couple of assumptions about counterfactuals. First, we assume the counterfactual axiom, Weakening:
(6) If p is possible, p □→ q and q entails r, then p □→ r.
Second, we assume that explanation is factive: p E q entails both p & q. Thus, both (2) and (3) follow from (4) and (6). We think it is also plausible that the fact that (3) explains (2) also follows from (4). What makes us confident that (3) explains (2) is precisely the conditional (4).
The principle gives a nice, general way to generate explanations of counterfactual facts, and we have not found any convincing counterexample. But in case there are doubts, we may cautiously treat (5) as defeasible, such that there is a presumption of truth in any given case.
From counterfactual explanations to explanations of counterfactuals
We will now argue that if C □→ eA, then e(C → A). One argument uses (5): if p is possible and p □→ (q E r), then (p □→ q) E (p □→ r). We are given that were C to hold, then A would have an explanation. But just as the typical Molinist thinks that once C is a sufficiently specific description of the circumstances of the choice, then there will be a fact of the matter about how the agent would act, so too, the typical Molinist thinks that there will be a fact of the matter about what, if anything, would in those circumstances be an explanation of the action. So if C □→ eA, then (according to the typical Molinist) there will indeed be an explanation in circumstances C, and so likewise there will be a fact of the matter about what that explanation would be. Therefore, there is a proposition p such that C □→ p E A. Then by (5), we get that (C □→ p) E (C □→ A). Hence, C □→ A has an explanation.
Here is a second argument. Consider, first, the kinds of explanation that a libertarian could give of non-derivatively free actions. There are two main proposals. One is that free actions are explained by the fact that an exercise of agent causation occurred. The other is that free actions are explained by some sort of a description of the reasons on which the agent deliberated. One could combine these. In all these cases, if the free action occurs in circumstances described by C, the maximally specific description C will include the explanans: so, for example, if the agent exercises agent causation, then that she exercises agent causation will be reported by C (though it will not be reported in which direction she exercises it); and if the agent acts on reasons, those reasons will be reported by C as well. In either case, then, we can read off from C what would explain the agent's action in C. Hence we can read off from C a claim p reporting the exercise of agent causation and/or deliberation on such-and-such reasons that would explain A in C, assuming C □→ A. But C □→ A follows from C □→ eA by Weakening and the factivity of explanation. So were it to be the case that C, then p would explain A: C □→ p E A. That's the first step.
Now as before, we could complete the argument by using (5) to infer that C □→ A has an explanation. But perhaps we are not fully convinced by (5). Nonetheless, the claim that (C □→ p) E (C □→ A) can still be defended, given that C □→ p E A. Suppose, for instance, p is a description of the reasons on the basis of which the agent chooses. Then we may claim that the fact that the agent would act in the specified way is explained by the fact that she would in C have reasons described by p. This claim is surely no less plausible than the claim that the reasons described by p explain the action described by A in a world where C holds. Of course, some will deny that reasons explain free actions. But our goal, recall, is to argue for the conditional claim that if free actions can be explained (such as by reasons), then so can CCFs. We suggest here that it is no more problematic to explain counterfactual action by counterfactual reasons than it is to explain actual action by actual reasons.
Someone might be worried that the fact that the agent would have these reasons in C does not guarantee that A would hold in C. But this worry is compelling only to the extent that the non-necessitation of action by reasons would be a worry about the explanation of free action in the categorical case. And, again, arguments against the possibility of explaining free actions are irrelevant to our arguments for the conditional claim.
As a final consideration, there is also this option: C ‘s entailing (or, relevantly entailing) p, where p is an explanation of A that would hold were C to hold, is what explains C □→ A. This option is available because the counterfactual explanation p for A is built out of ingredients in C. Some readers may find the deterministic explanation featured here more plausible.
All these arguments support the inference from the right side to the left side of (1).
From explanations of counterfactuals to counterfactual explanations
Suppose now that e(C □→ A), and assume that C is false.Footnote 6 Then there is an explanation of C □→ A. The only plausible candidate explanation of C □→ A is p □→ A (or perhaps that p entails or relevantly entails A), where p is a report of reasons and/or exercise of agent causation. But this a plausible candidate for an explanation of C □→ A only in so far as C □→ p E A holds: if C □→ p E A doesn't hold, then p has no chance of playing a role in explaining C □→ A. So, suppose C □→ p E A holds. Then since p E A entails eA, it follows from Weakening that C □→ eA, as desired. We have thus argued from the left side to the right side of (1).
Conclusions
Thus, CCFs have explanations if and only if free actions can be explained. Our argument moves from the counterfactual explanation of free actions to the explanation of counterfactuals about free actions. Notably, a similar move in the case of grounding is not available: even if free actions are grounded (say in agents' possessing action tropes or in agents' exemplifying action universals), it doesn't follow that CCFs are grounded (see, for example, Merricks Reference Merricks2007).
A Molinist might hope that our method for explaining CCFs would also help with the grounding problem.Footnote 7 After all, sometimes an explanation is also a ground. That in such-and-such a constitutional system the Queen gave royal assent to a bill explains why the bill became law, and it grounds the bill's having become law. But explanations do not always furnish grounds. In particular, it is normally thought that when the explanans fails to entail the explanandum, we do not have grounding.Footnote 8 For instance, Jones went broke because he put all his money into lottery tickets. This is a perfectly good explanation, but it doesn't seem to be grounding, since his putting all his money into lottery tickets does not entail his going broke. Similarly, our proposed explanations for CCFs fail to entail the CCF. One proposed explanation is that were C to hold, the agent would exercise agent-causal freedom, while the other describes the agent's reasons in C. But these are precisely the sorts of explanations that the libertarian will insist are non-entailing in the case of non-derivatively free actions. Thus, unless one at least gives up the entailment requirement on grounding,Footnote 9 the grounding problem remains.
