The Metaphysical Muddles of Molinism – Some Thoughts on MacGregor (Part 3)

By | November 13, 2023

This is my third and final post on Kirk MacGregor’s book, Molinist Philosophical and Theological Ventures (2022). My first post dealt with chapters 1–3. My second post dealt with chapter 4. This post covers chapters 5–6 and part of the concluding chapter.

In Chapter 5 MacGregor defends Molinism’s Biblical credentials against various open theism friendly prooftexts. I find his strategy here clunky and ineffective. In Chapter 6, MacGregor defends Molinism against a recent objection by philosopher Alex Malpass that, in effect, argues that Molinism can’t make sense of branching models of time. This chapter is more effective, though I do take some issues here with both MacGregor and Malpass.

The last two chapters in MacGregor’s book deal with applications of Molinism to sacred music (chapter 7) and universalism (chapter 8), respectively. Since I believe Molinism is incoherent on several levels, I have little interest in constructive applications of the model and so skip over those chapters.

Chapter 5: A Molinist interpretation of alleged open theist prooftexts (pp. 76–90)

MacGregor begins by identifying four categories of Biblical passages that may seem to offer support to open theism:

  1. God is surprised (e.g., Isaiah 5:2–5; Jeremiah 7:31; Ezekiel 22:30–31)
  2. God tests creatures to ascertain something (e.g., Genesis 22:12; Deuteronomy 8:2–21; Judges 3:4)
  3. God “changes his mind” (e.g., Exodus 32:9–14; Isaiah 38:1–5; Jeremiah 18:4–11)
  4. God “repents” or “regrets” his own choices (e.g., Genesis 6:6; 1 Samuel 13:13; 15:10, 35; 1 Chronicles 21:15)

MacGregor declares his intention to argue that all such passages can be understood either metaphorically or literally and that, if metaphorical, then Molinism can disclose “the literal truths to which these passages point” (p. 77). But first he starts us off with some conceptual preliminaries.

Conceptual preliminaries

First, MacGregor contends that true “would probably” conditionals belong to God’s natural knowledge. I concur. To the extent that event probabilities are defined at all, they are defined by the antecedent conditions of the event. So, that event E would probably occur in conditions C is something God could know simply by considering what probabilistically follows given those conditions.

Second, MacGregor contends, following Richard Swinburne, that “every proposition necessarily possesses an intrinsic probability” (p. 78). But this is, on the face of it, a highly implausible claim. Briefly, if some probability is intrinsic to a proposition, then the proposition must have that probability regardless of any other considerations, which is to say that no other considerations (ones not semantically entailed by the proposition) are relevant. For most propositions, however, other considerations are relevant. For example, the probability that a given cat is on a given mat 15 minutes from now depends on many other factors, such as where the cat and the mat are now, whether anything is impeding the cat from getting to the mat in time, whether the mat is in motion relative to the cat, whether the specific cat and mat exist, etc. So that proposition does not have an intrinsic probability. Whatever probability it has is extrinsic, since it depends on other relevant information. The only propositions that can have intrinsic probabilities are those that already include all probabilistically relevant information, and enough such information to define a probability distribution.

Third, MacGregor proposes that “within God’s natural knowledge is either the proposition ‘P would probably freely perform a in C’ (if the intrinsic probability … is greater than ½) or the proposition ‘P would probably not freely perform a in C’ (if the intrinsic probability is … less than or equal to ½)” (p. 79). But the “less than or equal” assignment for would probably not is completely arbitrary. Why not assign the “or equal” case to the would probably side instead? Or better, why not treat the “or equal” case as one in which neither would probably nor would probably not is true? The answer, I think, is that MacGregor is pulling a fast one and hoping no one notices. As a Molinist he needs the would (probably) and would (probably) not cases to exhaust the possibilities. Otherwise, conditional excluded middle fails and the Molinist no longer has any guarantee that either proposition in each would (probably) / would (probably) not pair of conditionals is true.

In any case, with this machinery in hand, MacGregor proceeds to reply to (a–d).

Response to (a) and (b) cases

He responds to (a) and (b) cases in roughly the same way: God is speaking from the perspective of His natural knowledge, not His middle knowledge. When God expresses surprise (a) or tests someone like Abraham (b), He’s telling us that, when viewed from the perspective of His natural knowledge, the way things turned out was probabilistic.

I find this response rather cringe. It seems too clever by half. If God has middle knowledge, why would He present Himself as if He only had natural knowledge? Or, in other words, why would God act as if open theism were true when, assuming Molinism, it isn’t? It seems that God, on MacGregor’s account, merely feigns surprise (a) and pretends to have learned something by creaturely testing (b). At best, that’s unserious, like someone knowledgeable pretending to be ignorant in order to mess with us. At worst, it’s deceptive, like someone trying to fool us into thinking he’s less knowledgeable than he really is.

Response to (c) cases

With respect to case (c), MacGregor cranks the cringe up a notch. He analyzes God’s conversation with Moses in Exodus 32:9–14 wherein God first threatens to destroy the Israelites, who have turned to worshipping the golden calf, but “changed His mind” (v. 14) after Moses implored God not to destroy them. MacGregor models the situation as follows (p. 82, adapted):

Let C = “The Israelites worship the golden calf”
Let W = “God tells Moses of His wrathful intentions toward the Israelites”
Let E = “Moses entreats God to be merciful”
Let D = “God destroys the Israelites”

Finally, MacGregor supposes that God’s natural knowledge contains the following:

    1. Pr((C ∧ W ∧ ~E) ☐→ D) = 0.9
    2. Pr((C ∧ W ∧ E) ☐→ ~D) = 0.9
    3. Pr((C ∧ W) ☐→ E) = 0.1

What are we to make of all this? MacGregor’s idea seems to be that, antecedent to his conversation with Moses, God knew (1)–(3). Thus, he knew (1) that if conditions C and W obtain and Moses does not entreat God for mercy, then there is a 90% chance that God destroys the Israelites. He also knew (2) that if C and W obtain and Moses does entreat God for mercy, then there is a 90% chance that God spares the Israelites. And, finally, He knew (3) that if C and W obtain then there is a 10% chance that Moses entreats God for mercy. Given (1–3), therefore, God expected from the perspective of His natural knowledge that He would destroy the Israelites given C and W. But because Moses did something (E) that was unexpected given God’s natural knowledge, God spared the Israelites thereby metaphorically “changing” His mind.

Persuaded? Me neither. This is the clunkiest take on a Biblical passage I have ever seen. Biblical text aside, MacGregor’s answer to (c) cases faces at least three serious problems.

First, it suffers the same weakness as MacGregor’s answer to (a) and (b) cases: assuming Molinism, there is no clear and plausible rationale for God to interact with Moses from the perspective of His natural knowledge alone. Why the heck should God bracket His middle knowledge like this? To remind Moses (and us) that He can function like an open theist God without it? Lol.

Second, in stating (1–3) MacGregor conflates the probability of a consequent with the probability of a conditional. For God to know that if C obtains then E “would probably” (with probability k) obtain, is not the same as for God to know that the probability is k that if C obtains then E would (definitely) obtain. No. If the relation between antecedent and consequent is merely probabilistic (the “would probably” case), then this falsifies the “would (definitely)” case, in which case the probability of the whole C ☐→ E conditional is zero. So, this is a sloppy logical error on MacGregor’s part. Logically speaking, it’s in the same boat as the dreaded modal fallacy of confusing the necessity of a consequent with the necessity of a conditional.

Third, because MacGregor’s (1) and (2) refer to God’s response, whether D or ~D, they cannot be part of God’s natural knowledge because they are not prevolitional. There can be no natural probability of 0.9 (say) that God destroys the Israelites given conditions C, W, and ~E because it remains up to God whether He destroys them under those conditions. In other words, the conditions don’t define a probability distribution for God. If they did, then the outcome would be up to chance at that point and thus no longer up to God. It would be like God saying to Himself, “If those conditions obtain then there’s a 90% chance I respond with wrath, so let’s roll a 10-sided die and see if the Israelites get lucky.” No one thinks God operates like that.

Response to (d) cases

Finally, with respect to (d), texts that depict God as “repenting” or “regretting,” MacGregor offers a two-pronged response.

First, he suggests a “metaphorical” reading according to which God chose in light of His middle knowledge to actualize a world in which an undesirable result “would” occur despite the fact that, according to His natural knowledge, that undesirable result was improbable. As MacGregor puts it,

This dissimilarity between the “would probably” and “would” counterfactuals in situations God actualizes is captured in the metaphors of divine sorrow, grief, and repentance. (p. 87)

Second, he suggests a “literal” reading according to which God’s middle knowledge provides Him with knowledge of “counterfactuals of divine emotion” wherein “God knows what his emotions would be if each CCF were actual” (p. 87). Why is this part of God’s middle knowledge? Because, says MacGregor, “such knowledge is dependent on God’s knowledge of the CCFs’ truth” (p. 87).

I don’t find either of these suggestions remotely plausible.

The first, like MacGregor’s responses to (a), (b), and (c) cases, has God responding as if He only had natural knowledge and thus as if open theism were true, even though (assuming Molinism) it’s not. As before, this is completely artificial and implausible. Why would God bracket His middle knowledge like this? Does MacGregor really expect us to believe that the point of the divine regret metaphor is to teach us that God is functionally an open theist with respect to His natural knowledge even though God has middle knowledge and so doesn’t literally regret anything? On Molinism God specifically decrees everything that comes to pass. So, why would God ever regret, repent, or become sad or angry about anything? Whatever happens, God gets exactly what He wanted when He selected this feasible world as the one to actualize. I can see how a Molinist God might feel chagrin over not having had better middle knowledge options, but I can make no sense of divine regret etc. with respect to what God has actually ordained.

The second response is equally puzzling. First, I don’t see how “counterfactuals of divine emotion” could possibly be part of God’s middle knowledge. MacGregor’s argument on this point is poor. What matters for these merely hypothetical emotions is not the actual truth or falsity of CCFs, but their possible truth or falsity. So, it must be part of God’s natural knowledge. Second, it’s really weird to think of God (or anyone) having an emotional response to the truth-value of a proposition. When I hear good or bad news, my emotional response is not to a proposition or its truth-value but to the situation described by the proposition. MacGregor has a bizarrely abstract way of thinking about emotions.

In any case, to wrap up this chapter, MacGregor concludes that “a Molinist exegesis of alleged open theist prooftexts reveal[s] them to be largely if not exclusively temporal metaphors for timeless relations of logical order” (p. 89, underline added). Now, I may have missed something, but I don’t recall seeing any exegesis in MacGregor’s chapter. The Molinist readings of (a–d) cases he proposes aren’t drawn out from the text. Rather, he’s taking Molinism for granted and trying to fit it into the text (eisegesis) as best he can. Ironically, this is exactly the sort of thing he says we shouldn’t do in Chapter 1—“philosophical constructs cannot legitimately be superimposed on Scripture” (p. 1).

Chapter 6: The logical consistency of Molinism on branching time models (pp. 91–106)

This chapter is a response to a recent criticism leveled against Molinism by philosopher Alex Malpass. As MacGregor presents it, Malpass uses a branching time model to argue that Molinism is “logically inconsistent” in three ways:

  1. Given consecutive temporal moments T1, T2, and T3 and a branching temporal array (representing causal indeterminism) in which multiple branches stem out from every temporal moment, there can be no truth-value at T3 to a proposition concerning what would have happened if a non-actual event had occurred at T2.
  2. “Molinism makes propositions about possible worlds containing compound tenses fail.”
  3. Branching time models force Molinists to posit “two different types of actuality,” which is absurd. (p. 91)

None of these charges is perspicuous, so we’ll have to try to unpack them as we go. Before diving in, MacGregor tells us that he does “not regard branching time models as literally conveying the structure of time” because “they incoherently combine the A-Theory and the B-Theory of time, among other problems” (p. 92). This is a weird charge because many branching time models are strictly A-theoretical (i.e., there is an absolute past, present, and future) and do not even attempt to overlay a B-theoretical (i.e., past, present, and future are merely relational, not absolute) conception on top of that.

Malpass’s first objection

Imagine that at T1 we are deliberating whether to flip a coin and that our choice is indeterministic. At T2 we make our choice. There are two possibilities: Flip (F) and No Flip (~F). Let’s assume that the flip is also indeterministic. If we flip at T2 and the coin lands at T3, then there are two possibilities: Heads (H) and Tails, i.e., not heads (~H). Finally, let’s stipulate that at T2 we choose to not flip (~F), in which case at T3 we are still holding an unflipped coin.

Now, let’s consider the propositions, “If we had flipped the coin at T2 then it would have landed heads” and “If we had flipped the coin at T2 then it would have landed tails.” Malpass argues that neither of these propositions is true at T3 because (a) both are equally close to the actual world (in which no flip occurs) and (b) it would therefore be completely arbitrary if either of them were true. Malpass concludes based on the standard Stalnaker–Lewis (SL) semantics for counterfactuals that both propositions are neither true nor false from the perspective of the actual world at T3. (One can also non-arbitrarily suppose that both are false, but I’ll set that complication aside and simply say that neither proposition is true at T3. That covers both possibilities.) Malpass generalizes this argument to conclude that there are no true would counterfactuals of indeterminism (p. 93).

I believe Malpass is correct in his assessment. There are no true would counterfactuals of indeterminism and so there are no true Molinist CCFs. MacGregor objects, however, that SL semantics is merely an “epistemological device” expressing the fact that we don’t know how the coin would have landed had it been flipped, and not that there is no fact of the matter as to how it would have landed (p. 94). At one level this retort works, but it doesn’t refute Malpass. The coin-flipping example was intended to capture all causally relevant aspects of the situation. What MacGregor is suggesting is that it leaves out a whole realm of causally irrelevant “counterfacts” (Patrick Todd’s term, not MacGregor’s) that somehow non-arbitrarily specify which would counterfactual (whether F > H or F > ~H) is true. The problem is that, aside from the assumption of Molinism, we have no reason to think this realm of counterfacts exists. In other words, while MacGregor can defensively retort that Malpass (and appeals to SL semantics generally) beg the question against Molinism, he hasn’t given us a positive reason to think Malpass is wrong. As best, we’re left with a stand-off.

MacGregor next charges Malpass with making the implausible claim that the mere passage of time “forces indeterminate events to either occur or not occur” (p. 94) or that time has “causal power over events” (ibid.). This is nothing but a straw man. Nothing in Malpass’s argument requires such assumptions. The problem, from Malpass’s perspective, is not that time somehow “forces” or “causes” events, but rather that there is seemingly nothing that resolves the indeterminism in favor of either F > H or F > ~H.

MacGregor then proposes to “demonstrate, contra Malpass, that counterfactuals of indeterminacy possess truth value” (p. 95), or rather and more precisely, that some would counterfactuals in which the consequent is only indeterministically related to the antecedent are true. MacGregor’s “demonstration,” if we can call it that, consists in an appeal to the controversial principle of conditional excluded middle (CEM). Now, I’ve already written a separate post on CEM and why it’s implausble, but there is also a clear non sequitur in MacGregor’s argument. He illicitly moves from the fact that, given that the coin is flipped, it must be that one of the two possible outcomes (H or ~H) results to the claim that it “must be true” that one of the two possible outcomes would result. In other words, he’s conflating the obviously true principle “Necessarily, F > (H ∨ ~H)” with the controversial CEM principle that “Necessarily, (F > H) ∨ (F > ~H).” The latter simply doesn’t follow from the former, at least not unless we make additional question-begging assumptions.

I conclude that MacGregor fails to rebut Malpass’s first objection. His only salient point is the strictly defensive one that Malpass’s reliance on SL semantics is implicitly question-begging against Molinism. As soon as MacGregor tries to go on offense and show that Malpass is wrong, however, he completely whiffs it.

Malpass’s second objection

Malpass’s second objection exploits Molinism’s commitment to the thesis that, in every possible indeterministic scenario, there is a definite way things would go. He wants to show that this commitment is incompatible with branching time and a principle that Patrick Todd has called “retro-closure” (RC). According to RC, if some event does happen then it was true at all previous times that it was going to happen. To see how this objection unfolds, let’s consider the coin-flipping example from above. There are, if you will, three possible worlds represented in this scenario:

W1: F ∧ H
W2: F ∧ ~H
W3: ~F

Let’s assume that W3 is the actual world, in which the coin is not flipped (~F). According to Molinism, there is nevertheless a definite truth about what would have happened if the coin had been flipped. Let’s assume that it would have landed heads as in W1 (F ∧ H). According to RC, Malpass reasons, if the coin had been tossed and landed heads, then it would have been true at T1 that the coin was going to land heads. But that can’t be correct because we’re assuming that W3 is that actual world, and so it was not true at T1 that the coin was going to land heads. Since Malpass takes RC to be a “tautology” (that is, a trivially necessary truth), he concludes that Molinism is false. It is not the case for every possible indeterministic scenario that there is a definite way things would go.

This is a clever but ultimately ineffective argument. If I were responding to it as an open futurist, I would reject both RC and Molinism. On open futurism, retro-closure is far from a tautology. In fact, it’s just plain false. But MacGregor can’t press this objection because he is committed to the idea that there is a unique actual future (UAF). Thus, for any event E on that UAF, if E occurs at time T, then it was previously true that E at all times prior to T that E was going to occur at T.

While MacGregor doesn’t put it this way, his response to Malpass amounts to denying that RC can be applied counterfactually. In other words, you’ve got to fix the world before you apply RC. So, if W3 is the actual world, then if ~F obtains at T2, then it was always the case in W3 that ~F was going to obtain in T2. If, however, W1 had been the actual world such that F ∧ H obtains at T3, then it was always the case in W1 that F ∧ H was going to obtain at T3. It is still true in W1 that if W3 had been actual then it would have been true at T1 that F ∧ H were going to obtain at T3, but that F ∧ H obtain at T3 in W3 does not entail that it was true at T1 in W1 that F ∧ H are going to obtain at T3.

In my judgment this a perfectly fair response to Malpass given the dialectical context. Even if we grant (contrary to fact) that RC is a tautology, Malpass’s trans-world application of it is not tautological. At most, it’s an intra-world tautology.

Malpass’s third objection

MacGregor states this objection as follows: “if each node in the [branching time tree] possesses its own actual future, this forces us into the metaphysical absurdity of postulating two different types of actuality” (p. 101). How so, exactly? He doesn’t explain things that clearly, but does offer an illustration. Malpass has no sister. So, what sense does it make to say “if Malpass had a sister [which one?], then she would actually choose ice cream rather than toast”? As near as I can make out, the objection is that, if there is a specific way things would go at any indeterministic decision point, then that way is the “actual” future relative to that point. But some of those points are purely hypothetical. They never actually come to pass, just as the coin in Malpass’s scenario is never actually flipped. And so some things are relatively “actual” without being categorically “actual.” And, this, apparently, is supposed to be an absurdity.

I confess that I don’t see the problem. At best this seems to me to be a variation on the unsuccessful second objection. Nevertheless, MacGregor’s response is less than clear. Rather than simply distinguishing (as I did) between relative and categorical “actuality” and noting that the first is conditional and the other isn’t, in which case there is no conflict, he goes off into problematic territory.

First, he departs from established Molinist terminology. Several decades ago, Tom Flint, perhaps the most respected living Molinist, defined a “galaxy” as the collection of all possible worlds (i.e., complete histories) compatible with the content of God’s middle knowledge. Since middle knowledge is contingent, it can vary from possible world to possible world. A “Flint galaxy” is therefore a group of possible worlds that happen to share the same middle knowledge content. MacGregor, however, takes a “galaxy” to be a collection of possible worlds that share the same “trunk” on an indeterministic branching time tree. This is very different from Flint’s usage. In a Flint galaxy, worlds aren’t essentially connected to a common history or “trunk.” The worlds may, in fact, have completely different histories, or no history for that matter, because God may simply decide not to create. In a “MacGregor galaxy,” in contrast, all worlds have the same initial conditions up to the first indeterministic decision node. So, when MacGregor says that “God knows the one feasible world emerging from each possible galaxy” (p. 103), he’s saying something that would almost certainly confuse scholars used to the established terminology.

Second, after presenting several branching diagrams illustrating his notion of a “galaxy” and relating it to God’s natural, middle, and free knowledge, MacGregor forgets all about Malpass’s objection and starts talking about “transworld damnation” (p. 105)! For those who don’t know, transworld damnation is the conjecture that any individual who winds up damned in the actual world would have wound up damned in any other feasible world in which they exist. This is an incredibly implausible concept. For if we suppose (as Molinists do) that any salvific decision must be indeterministic, then any given individual would have to be astoundingly unlucky if their creaturely essence were such as to freely reject salvation in every feasible salvific decision scenario in which they could be placed. (I say “unlucky” because on Molinism the truth values of a creature’s CCFs have to be fixed independently of whether that creature exists. So the truth values are not up to person in question.)

Concluding remarks

MacGregor’s chapter 5 is, I think, a weak response to alleged open theism Biblical prooftexts. His response to Malpass is chapter 6 is pretty decent overall when it remains defensively focused on the objections. When he departs from that focus, however, MacGregor gets himself into trouble.

On pp. 136–137 of his conclusion chapter, MacGregor summarizes what he takes to be some of the main takeaways from chapters 5 and 6. I am not at all impressed by what he says here.

He there says that, contrary to the standard (Stalnaker–Lewis) semantics, “might-counterfactuals are seen not to be the contradictories of would-counterfactuals … but rather as statements of intrinsic probability” (p. 136, underline added). Um, no. First, the claim is not that might-counterfactuals are the contradictories of would-counterfactuals but that might not counterfactuals are the contradictory of would-counterfactuals. Second, nothing MacGregor says anywhere in his book makes a strong argument that the standard semantics is wrong and there’s certainly nothing strong enough that the semantics can be “seen” to be wrong. As I explained above, his best argument is a non sequitur because it conflates the obviously correct F > (H ∨ ~H) with the controversial CEM claim that (F > H) ∨ (F > ~H).

He also says that “branching time models give the lie to the open theist allegation that there is a logical distinction between the locutions ‘x will not’ and not [x will]” (p. 136, underline added). Again, no. First, this is an open futurist allegation, not an open theist allegation per se. Second, MacGregor is wrong that the non-equivalence of will not and not will depends on Stalnaker–Lewis semantics (which MacGregor erroneously takes himself to have refuted). While there is certainly a structural parallel between the would–might duality entailed by those semantics and the non-equivalence of will not and not will, the latter is not at all dependent on the former. See my comments on Chapter 3 of Patrick Todd’s 2021 book defending open futurism for explanation.

Finally, he says that “in branching time models, there simply are no ‘might’ branches on the trees” (p. 137). Emphatic no. There are several different branching time models. Some are open futurist, some (like MacGregor’s) are Ockhamist, and some are open-closurist. On any branching model that countenances indeterminism besides Ockhamism, there are “might” branches, things that might happen and that neither will nor will not happen. When MacGregor says categorically that “there simply are no ‘might’ branches” on branching time models he displays his ignorance of the literature.

Leave a Reply

Your email address will not be published. Required fields are marked *