In a previous post I presented the infamous grounding objection to Molinism. The problem stems from the fact that for Molinism God’s middle knowledge (MK) is supposed to be
- non-natural, i.e., not part of, entailed by, or grounded in God’s nature
- contingent, i.e., neither necessary nor impossible in the abstract metaphysical sense that I describe in this post
- pre-volitional, i.e., its content is fixed prior to and independently of any contingent divine volitions, including God’s decision to create
And yet there seems to be no plausible source for MK that would allow it to have these properties. I argued that Molinists should just bite the bullet and admit that MK is ungrounded, no matter how counterintuitive it may be to say (as Molinists I believe must ultimately do) that an omnipotent God’s creative options are externally constrained by a bunch of contingent brute facts that are grounded in nothing and come from nowhere. While many Molinists argue that MK doesn’t really need to be grounded, most of them seem to hold that MK nevertheless is grounded, or at least that there remain plausible grounding options. In this post I’m going to examine seven of those options and argue that all of them are abject failures.
What MK grounding requires
In my previous post on the grounding objection, I said that adequate grounds for contingent truths need to be concrete, that is, non-abstract. The problem with abstract grounds for contingent truths is that abstracta like propositions, numbers, etc. are most plausibly regarded as necessary entities. If they exist at all, they necessarily exist. It is hard to see, for example, how the truths of mathematics could be contingent. But if there are necessary grounds for contingent truths, then the truths would be necessary, not contingent. Since MK is, by definition, contingent, it would seem that its grounds, if it has any, must be contingent and therefore not abstract but concrete.
Grounds must also be sufficient for the truths that they ground. The concrete realities that ground contingent truths need to entail the truths in question. In other words, if the ground is given, then the truth in question cannot fail to be true. Suppose, for example, that truth T is grounded in G. If G were insufficient for T’s truth, then it would be possible for G to exist and T be false. But then G wouldn’t be enough to explain why T is true and not false, so it wouldn’t be an adequate ground. We’d need to add something else, say X, to G such that necessarily, if G+X exists then T is true.
Now, since MK consists entirely of “would” and “would not” conditionals like <If C, then O would (indeterministically) occur>, then to ground those truths we’re going to need to some kind of (a) concrete, (b) contingent reality that (c) obtains prevolitionally, that (d) is not entailed by God’s nature, and that (e) entails the truth of exactly one of each would / would not pair of MK conditionals. Anything that fails to meet (a)–(e) is not an adequate ground for a MK conditional. With that in mind, let’s now consider some Molinist grounding proposals (MGPs).
MGP #1: Individual creaturely essences
Our first proposal is that MK is grounded in necessarily existing individual creaturely essences, sometimes called ‘haecceities’. The problem with this suggestion is two-fold: (1) These entities, if they even exist, are abstract, not concrete. So they fail the concreteness test. (2) If it is part of agent S’s essence to do action A in circumstances C, then S cannot be free in C to do otherwise. The circumstances in conjunction with S’s creaturely essence will entail the outcome. This passes the entailment test, but fails the contingency test.
MGP #2: Conditional excluded middle
When applied to MK conditionals, the controversial thesis of conditional excluded middle (CEM) says that either <If S were in C then S would freely do A> or <If S were in C then S would not freely do A>. Bill Craig claims that this principle holds for MK conditionals in which the antecedent (S’s being in C) is “fully specified” for
it seems that if the agent were placed in C and left free with respect to action A, then he must either do A or not do A. For what other alternative is there?
But this is specious reasoning. First, the second disjunct of CEM is not what the Molinist wants. He doesn’t want to say that if S were in C then S would not freely do A because not freely doing A is compatible with being causally determined to do A. What the Molinist wants to say, rather, is that either <If S were in C then S would freely do A> or <If S were in C then S would freely not do A>. The location of “not” in the second conditional matters. In short, CEM says that either would or would not is true but the Molinist needs either would freely or would freely not to be true, and CEM doesn’t license that.
Second, suppose we concede (counterfactually!) that <If S were in C then S would not freely do A> is equivalent to <If S were in C then S would freely not do A>. Now, if S is free in C to do A or not-A, then the most that could follow is that the whole disjunction <Either <If S were in C then S would freely do A> or <If S were in C then S would not freely do A>> is true. It does not follow that one or the other of the disjuncts is determinately true. Indeed, even if S actually and freely does A in circumstance C, it still doesn’t follow that <If S were in C then S would freely do A> because by hypothesis S could just as easily have done otherwise in C. Imagine, if you will, that God repeatedly resets the circumstances and S’s mental state to what it was just before his free choice. Would S freely do A every time? If not, then the fact that S freely does A on some occasions and not others provides no reason for thinking it was ever determinately true that S would do A in C. And if so, that is, if S does A every time in C, then it seems that S’s doing A in C is not free. After all, if S never refrains from doing A despite innumerable opportunities to do so, then the best explanation would seem to be that S’s character and mental state in C are such that refraining from A is not a “live option”, and so not something S could do freely. In short, neither CEM nor S’s actual free choices are enough to establish the determinate truth of one of the disjuncts. And if S’s choices can’t even establish the determinate truth of one of the disjuncts after the fact, then they can’t establish it prevolitionally either.
MGP #3: Divine prevolitional modeling
Keith Wyma has suggested that MK could be grounded in prevolitional “divine modelling”. Basically, the idea is that God prevolitionally imagines a whole bunch of test scenarios in which various creatures are placed in various circumstances where they can make free choices. While running through these scenarios, God takes note of what virtual “choices” each virtual “creature” virtually “makes” and then uses that information to construct His middle knowledge.
I must say, this proposal strikes me as a complete non-starter. Virtual individuals aren’t free creatures. Virtual choices aren’t real choices. Thus, there is no reason to think that what God imagines creatures would freely do bears any necessary correlation with what real free creaturely would actually do if such circumstances were actualized. Even if God virtually runs the same test scenario with S in C dozens and dozens of times and gets the same result (S does A) every time, that’s no reason for thinking that S, if made real would freely do A in C. If anything, it’s a reason for thinking that S wouldn’t be free in that scenario.
MPQ #4: Divine ideas and supercomprehension
I recently pressed Molinist Tim Stratton on the grounding objection, and this is what he proposed:
- Following John Laing, argue that each possible creature is an idea in the mind of God, an idea that “includes in it all the true counterfactuals regarding how the creature it represents would in fact act” (Molinism and supercomprehension: Grounding counterfactual truth [PhD diss., Southern Baptist Theological Seminary, 2000], p. 338). Accordingly, CCFs are objectively there to be known and ultimately grounded in the divine mind. Nonetheless, CCFs “retain their contingent status because although they are grounded in God, they receive their truth values from the creaturely wills which pre-exist in the divine mind as ideas” (ibid.).
- If one believes that option 1 constitutes some form of essence-determinism, one could argue that each individual essence owns but does not include the CCFs appertaining to it (MacGregor, Molinist philosophical and theological ventures, forthcoming).
- Employing perfect being theology, argue that CCFs are grounded not in properties of individual essences but in God’s infallible yet contingent beliefs about individual essences logically prior to God’s creative decision. Such beliefs are timeless divine ideas that constitute the truthmakers of CCFs. I will contend that these beliefs are formed by God’s intuition, a divine cognitive faculty that furnishes knowledge in evidentially underdeterminative cases (MacGregor, ibid.).
My initial thought when Tim presented this to me was something like “WTF??!!” I have a very hard time understanding how Molinists can take ideas like (1)–(3) seriously. None of them seem remotely plausible to me with respect to the grounding objection. Regarding (1), if these creaturely “ideas” in the mind of God include the information that Laing specifies, then this is just MGP #1, an appeal to individual creaturely essences. It fails because it rules out both creaturely freedom and the contingency of MK. Regarding (2), what the heck does it mean for a creaturely essence to “own but not include” the CCFs pertaining to it? I can’t think of any way a mere essence could “own” anything that it doesn’t include or entail, so this seems like a non-starter to me. Regarding (3), where do these “infallible yet contingent beliefs about individual essences” come from? If they are contingent then they can’t come from the essences. Ah, says MacGregor, they come from God’s “intuition”. That’s a seriously lame answer. It basically amounts to answering “How does God come by this contingent information?” with “Well, He just does.” MacGregor and Stratton would be better off if they just admitted that MK is ungrounded rather than come up with vacuous, pseudo-explanatory stuff like this.
MPQ #5: Abstract facts and the disquotation principle
Bill Craig has explicitly promoted this response, and MacGregor seems to advocate the same idea during a debate with Alex Malpass. The so-called “disquotation principle” refers to a widely accepted thesis proposed by philosopher Alfred Tarski. The principle says simply that True(p) iff p. In other words, we can use “truth” talk to switch back-and-forth between an object language statement (“p”) and a meta-linguistic statement (“It is true that p”) about the object language statement. Craig uses this principle in the following way:
- He takes MK for granted. This gives him a set of truths to fill out the left-hand side of the disquotation principle.
- He invokes the disquotation principle to strip away the “truth” predicate and thereby arrive at a set of “facts” corresponding to those truths (i.e., the right-hand side of the disquotation principle).
- He then offers these abstract “facts” as the grounds for God’s MK.
There are several problems with this grounding strategy. First, these “facts” are abstract and so can’t provide the concrete grounds required for the contingent truths of God’s MK. Second, as abstract, these “facts” are presumably necessary, and so can’t ground contingent truths. Third, this whole strategy rests on a misuse of the disquotation principle. The principle merely relates an object language with a metalanguage. As such, it doesn’t have any metaphysical or ontological implications. The supposed “truths” of MK and the “facts” disclosed by disquotation are just two different ways of talking about the same thing. So Craig’s grounding explanation is vacuous. There is no real difference between these abstract “facts” and the MK truths they are supposed to ground. As an explanation, this is on par with Molière’s famous example of a scientist “explaining” that opium has the power to put people to sleep because it has “dormitive power”, i.e., the power to put people to sleep. It just pushes the grounding question back a step. We wanted to know where MK truths come from. Craig responds by pointing to a realm of abstract “facts” that restate MK truths using “fact” language. But now we have the question: where do those contingent, prevolitional, etc. “facts” come from? These supposed facts are just as ontologically problematic as the truths they are supposed to explain.
MGP #6: The conceptualist model of knowledge
Bill Craig proposed in The Only Wise God (Baker, 1987, p. 125) a “conceptualist” model of divine foreknowledge according to which God innately knows which future contingent propositions are true without having to be acquainted with anything extrinsic, as the contrasting “perceptualist” model says. In the following chapter (p. 134), he takes the same line with respect to God’s MK. God, being omniscient, knows all truths. MK conditionals are true. So God knows them, end of story. We don’t need, says Craig, any model of how God can know those truths. He just does.
The main problem with this proposal is that contingent truth is a relational property, not an intrinsic property. Unlike conceptually necessary propositions, which are true in virtue of their intrinsic content, contingently true propositions can only be true in virtue of standing in an external correspondence relation to something else. Mere reflection on a contingent proposition cannot by itself reveal whether the proposition is true. To know that a contingent proposition is true, therefore, does require acquaintance with some external correlate. Craig’s “conceptualist” model is a non-starter.
MGP #7: The actual world
The final proposal I want to look at is not offered by any Molinist I know of as a complete grounding proposal, but only a partial one. Both Alvin Plantinga (The Nature of Necessity) and Edward Weirenga (The Nature of God: An Inquiry into Divine Attributes, pp. 147–148) have suggested that which possible world is actual is relevant to determining which MK conditionals are true on the grounds that “one feature determining the similarity of worlds is whether they share their counterfactuals” (Plantinga, p. 178). To understand this one needs to know that on the standard (Stalnaker–Lewis) semantics for counterfactuals, a counterfactual (A > C) is true if and only if all of the worlds “closest” to the actual world in which the antecedent is true are ones in which the consequent is true. Setting aside questions about what “closeness” between worlds means, what Plantinga and Weirenga are suggesting is that (part of) what makes <If S were in C then S would freely do A> true rather than <If S were in C then S would freely not do A> is that the actual world is such that <If S were in C then S would freely do A> is true. Hence, the antecedent worlds “closest” to the actual world are (all other things equal) ones in which <If S were in C then S would freely do A> is true.
As Molinist Tom Flint notes, many Molinists reject this suggestion, seeing it as “unsatisfying, question-begging, … [or even as] a surrender of the claim that counterfactuals of creaturely freedom are grounded” (Divine Providence: The Molinist Account, p. 136). I believe those assessments are correct for two reasons. First, it makes grounding circular for it says that MK counterfactuals are true, in part, because they are true. Second, it fails to provide pre-volitional grounds for MK. After all, according to Molinism which world is actual isn’t settled until after God’s free decision to create. So information about which world is actual simply isn’t available to do any grounding work in a pre-volitional context.
Above I canvassed several different Molinist replies to the grounding objection and I think I’ve shown that all of them fail to supply anything remotely adequate. In every case the proposed grounds are either not concrete, not contingent, not prevolitional, not sufficient (to entail the truths they are supposed to ground), and/or merely restate and relocate the grounding problem.
Having been a Molinist during my undergraduate and early graduate years, I can readily understand why someone might be attracted to the view and think that it’s benefits outweighs its costs. But even when I was a Molinist, I was conflicted about it. A major factor that eventually pushed me away from Molinism and toward open theism was the cognitive dissonance I experienced in regards to the grounding objection and the nagging question once-posed by William Hasker, “How does God come by his middle knowledge?” Even the most sophisticated Molinist responses to the grounding objection struck me then as incredibly lame. Special pleading on stilts, if you will. As this post makes clear, my assessment of Molinism on this score hasn’t changed.