If Molinism is true, what can you do?

Abstract

Suppose Molinism is true and God placed Adam in the garden because God knew Adam would freely eat of the fruit. Suppose further that, had it not been true that Adam would freely eat of the fruit, were he placed in the garden, God would have placed someone else there instead. When Adam freely eats of the fruit, is he free to do otherwise? This paper argues that there is a strong case for both a positive and a negative answer. Assuming such cases are possible under Molinism, we are left with a puzzling question: if Molinism is true, what can you do?

Introduction

The Molinist, or Middle Knowledge, view of divine providence consists of (at least) the following three claims: (i) that for any possible (created) agent, S, and any possible circumstance, C, that S might be in, there is a fact as to what S would freely do were she in C; (ii) that such facts, often called “counterfactuals of creaturely freedom” (or “CCFs”), are contingently (i.e. not necessarily) true but their truth is independent of God’s will; and (iii) that prior to creation, God uses his knowledge of such CCFs—his so-called “middle knowledge”—in determining which world to actualize.¹ For instance, suppose God has settled on actualizing a world where the first free creatures find themselves in a paradisical garden. According to Molinism, God knows how each creature would freely behave, were they placed in the garden, and can use this knowledge in deciding whom to create. Suppose it is very important to God, for inscrutable reasons, that the individual placed in the garden freely eat the forbidden fruit. Looking over the CCFs, God sees that only Adam and Ben would freely do so, but for ancillary reasons, God ranks Adam just ahead of Ben. Importantly, if it hadn’t been true that Adam would freely eat the fruit, God would have ranked Ben ahead of Adam. But as things actually are, God actualizes a world where Adam freely eats the fruit in the garden and Ben doesn’t exist.²

Here’s the central question of this paper: when Adam freely eats the fruit, is he able to do otherwise? My contention is that there is a strong case for both a positive and a negative answer. On the one hand, if Adam had not eaten the fruit, God wouldn’t have placed Adam in the garden to begin with—God would have placed Ben in the garden instead and Adam would be the non-existent one. Or, at the very least, for Adam to do otherwise, facts about what either Ben or God would do would need to be different. So, Adam must eat the fruit. On the other hand, given that it is only contingently true that, were Adam placed in the garden, he would freely eat the fruit, that means it is possible for Adam to be placed in the garden and not freely eat of the fruit. Given some plausible further stipulations, that implies that he is able to avoid eating the fruit. Assuming this kind of case is possible, we therefore have a novel argument against the Molinist view of divine providence. Or at the very least, we have a difficult question: if Molinism is true, what can you do?

In what follows, I will first articulate these rationales more carefully. I will then conclude by considering an objection which brings out a more fundamental question for the Molinist position.

Adam must eat the fruit: self-defeating actions and counterfactual power

I offer two arguments for the claim that Adam must eat the fruit. The first, put simply, is this: Adam’s doing otherwise than eating the fruit constitutes a “self-defeating” action. But no agent is able to perform a self-defeating action. Therefore, Adam is not able to do otherwise. The second, put simply, is this: if Adam is able to do otherwise, then Adam has counterfactual power over facts concerning what other agents would do in various circumstances. But Adam has no such power. Therefore, Adam is not able to do otherwise. Let’s take each argument in turn.

Adam’s doing otherwise is self-defeating

What exactly are “self-defeating” actions? The most famous example comes from the (in)famous Grandfather Paradox: suppose Tim has a time machine and travels back to a time when Grandfather was a young man, long before Grandfather met Grandmother. Tim approaches Grandfather on the street, gun in his hand, hatred in his heart…³

Can Tim kill Grandfather? It would seem not. After all, if he were to kill Grandfather, Grandfather would never meet Grandmother, which means Father would never be born, which means Tim would never be born, which means Tim would never travel to the past and kill Grandfather to begin with. Contradiction. Tim’s killing Grandfather would appear to “defeat” the performance of that very action. More generally, we might say that a “self-defeating” action is such that, were the agent to perform it, she wouldn’t perform it. Presumably, no agent can perform such an action.⁴

For the same reason, it would seem as if Adam cannot do otherwise than eat the fruit. After all, if he were to do otherwise, then the relevant CCF—that if Adam were placed in the garden, he would (freely) eat the fruit—wouldn’t have been true, which means God wouldn’t have placed Adam in the garden (or created him at all), which means Adam wouldn’t have done otherwise than eat the fruit. Contradiction. Adam’s doing otherwise appears to be as self-defeating as Tim’s killing Grandfather.⁵

Admittedly, this rough gloss on a “self-defeating” action is inadequate. Return to Tim’s case. According to the standard semantics, a counterfactual such as “If Tim were to kill Grandfather (at time t), then Tim wouldn’t kill Grandfather (at time t)” is true just in case all the closest worlds where Tim kills Grandfather (at t) are worlds where Tim doesn’t kill Grandfather (at t). Obviously, there are no (possible) worlds like that. Moreover, there would seem to be worlds where Tim successfully kills Grandfather without any contradiction. Consider a world where Tim kills Grandfather but then Grandfather is later resurrected from the dead, goes on to meet Grandmother and so on. That may be an incredibly distant world, but it’s closer than non-existent (or contradictory) worlds. So, according to the standard semantics, it’s false that “If Tim were to kill Grandfather, Tim wouldn’t kill Grandfather.” Similar comments apply to Adam’s doing otherwise than eating the fruit. We need a more plausible understanding of a “self-defeating action.”

Here’s just one way of refining this notion. Start by focusing on two features of Tim’s case. First, it seems obvious, relative to an ordinary context, that if Grandfather had not survived his youth—either because Grandfather was murdered, or because he fell ill, etc.—Tim wouldn’t have existed, and for the reasons just mentioned: relative to an ordinary context, at least, if Grandfather hadn’t survived his youth, Father wouldn’t have been born, and so on. Second, it’s not just that there is a counterfactual relation between Grandfather’s survival and Tim’s existence, but there is an explanatory relation as well: Tim exists partly because Grandfather survived his youth.

So, Tim’s existence is both (partly) explained by and counterfactually dependent on Grandfather’s survival. To introduce some terminology, let’s say that Tim’s existence is explanatorily dependent on Grandfather’s survival, where fact F is explanatorily dependent on fact G just in case (i) had G not obtained, F would not have obtained (relative to an ordinary context) and (ii) this counterfactual relation obtains because G at least partly explains F, directly or ancestrally.⁶

With this in mind, we can understand the claim that “no one can perform self-defeating actions” as the claim that “if an agent’s existence is explanatorily dependent on fact F, then the agent cannot perform any action that would require F to not obtain.” More formally:

No Self-defeating Actions (NSA): For any agent, S, action X, time t, and fact F, S cannot perform X at t if the following is true: (i) S’s (actual) existence is explanatorily dependent on F and (ii) if S were to perform X at t, F would not obtain.⁷

Just as NSA implies Tim cannot kill Grandfather, so it implies that Adam cannot do otherwise than eat the fruit. With regard to condition (i), it seems quite plausible that Adam’s actual existence is explanatorily dependent on the relevant CCF, namely, that if Adam were placed in the garden, he would eat of the fruit. After all, Adam was placed in the garden partly because God knew this CCF, and, given our stipulations, if the CCF hadn’t obtained, God would’ve placed Ben in the garden instead. With regard to condition (ii), if Adam were to do otherwise than eat of the fruit while in the garden, the relevant CCF wouldn’t have obtained, at least given standard assumptions.⁸ Therefore, Adam is no more able to do otherwise than Tim is able to kill Grandfather.

While NSA offers a promising way of articulating the claim that no agent can perform a self-defeating action, it is important to note that NSA isn’t necessary for the argument. For instance, after reflecting on Grandfather-style cases, Joshua Spencer (2013) offers the following principle:

JS: Agent S is able to refrain from doing action A only if, had S not done A, S would have done something else instead.⁹

Spencer points out that there are two importantly different ways that S can fail to do A: by doing something else instead or by not doing anything at all. Given the setup of Adam’s case, if it hadn’t been true that Adam eats the fruit, Adam wouldn’t have done anything at all—Ben would be the existent one doing all of the eating. Therefore, JS is another principle motivated by Grandfather-style cases which also implies that Adam is not able to refrain from eating the fruit.

So, while I’m inclined to explain Tim’s inability to kill Grandfather in terms of NSA, NSA is not absolutely necessary for the argument. All that is needed is that Adam’s doing otherwise than eating the fruit is analogous to Tim’s killing Grandfather. Given that Tim cannot kill Grandfather—which is a widely, though not universally, endorsed claim—we thus have good reason to think Adam cannot do otherwise, regardless of whether NSA is the best way of articulating exactly why that is.

It must be admitted, however, that there is always the possibility that a relevant difference between Tim’s and Adam’s cases will be discovered.¹⁰ Moreover, the Molinist may simply “run the argument in reverse” and insist that, just as Adam is able to do otherwise, so Tim is able to kill Grandfather (though of course neither agent will do so). With this in mind, I offer a second argument for the claim that Adam must eat the fruit, one that stems from commitments arguably more “internal” to Molinism.

Adam’s doing otherwise requires too much counterfactual power

If Adam were to do otherwise than eat the fruit, some parts of the world would’ve been different and others would’ve remained the same. If Adam hadn’t eaten the fruit, he presumably wouldn’t have been banished from the garden then and there—that part of the world would’ve been different. In contrast, Adam’s action would seem to make no difference whatsoever as to when the Earth was formed, or the distant past more generally.

Let’s introduce some terminology for ease of expression. Those facts which are counterfactually independent of Adam’s action are commonly said to be “resilient” to Adam’s action, and those that aren’t are “non-resilient.” More formally, we can put the distinction like this. Suppose F is a fact and that S performs action X at time t. Fact F is resilient to agent S’s performing X at t if the following is true: F still would’ve been true, had S done otherwise than X at t. Finally, S has “counterfactual power” over F if F is non-resilient to S’s performing X at t and S is able to do otherwise than X at t—that is, S is able to do otherwise and F would (or at least might) not have obtained had S done so.¹¹

Now consider the following question: Are CCFs resilient or non-resilient to creaturely actions? A standard view is that some are and some aren’t. For instance, consider the central counterfactual we’ve been concerned with so far:

CCF_Adam: If Adam were placed in the garden, he would freely eat of the fruit.

It is immensely plausible that CCF_Adam is non-resilient to Adam’s doing otherwise.¹² That is, if Adam had done otherwise than eat of the fruit, CCF_Adam might not (indeed wouldn’t) have been true. The details need not detain us here, but the basic idea is simple enough: If p counterfactually implies q, then, if p is true, q had better be true, too.¹³ So, if Adam is able to do otherwise, then Adam has counterfactual power over CCF_Adam.

However, some CCFs appear resilient to Adam’s doing otherwise. Consider another CCF we have been concerned with:

CCF_Ben: If Ben were placed in the garden, he would freely eat of the fruit.

It is immensely plausible that CCF_Benis resilient to Adam’s doing otherwise, at least if we assume that Adam is able to do otherwise. It is widely held by Molinists that not even God has counterfactual power over CCFs, such as CCF_Ben.¹⁴ If God lacks counterfactual power over CCF_Ben, then so does Adam, presumably. (Only Ben might have such power over CCF_Ben.) So, if Adam is able to do otherwise, it must be that CCF_Ben is resilient to his doing so. And this makes intuitive sense. How could facts about what Adam actually does in the garden make any difference to facts about what Ben would do in the garden? It’s hard to see how CCFs about one creature could depend, even counterfactually, on the actions of another creature.

Finally, consider facts about what God would do given the truth of certain CCFs, often called “counterfactuals of divine freedom” (or “CDFs”). For instance, we have been implicitly appealing to the following CDFs so far:

CDF_Adam: If the CCFs were such that only Adam and Ben would freely eat of the fruit when placed in the garden, God would (freely) place Adam rather than Ben in the garden.

CDF_Ben: If the CCFs were such that only Ben would freely eat of the fruit when placed in the garden, God would (freely) place Ben rather than Adam in the garden.

In some ways, the discussion surrounding CDFs is even trickier than the one surrounding CCFs, but all that is relevant for our purposes is this: it is immensely plausible that Adam lacks counterfactual power over CDF_Adam and CDF_Ben. Just as Adam would seem to lack counterfactual power over facts concerning what other creatures would do in a given circumstance, so he would seem to lack counterfactual power over facts concerning what God would do in a given circumstance.¹⁵ So, if Adam is able to do otherwise, CDF_Adam and CDF_Ben are resilient to Adam’s actions.

All these points together provide another argument for the claim that Adam is unable to do otherwise than eat the fruit. Suppose, for reductio, that Adam is able to do otherwise. If he had done otherwise, then CCF_Adam would’ve been false. But what would lead God to actualize a world where Adam is in the garden, especially if God knew that CCF_Adam was false? The are only two options: either it wasn’t feasible for God to actualize a world where the first free creature in the garden freely eats the fruit or it was feasible. The first option requires that no other creature, including Ben, would have freely eaten the fruit either. That means that had Adam done otherwise, CCF_Ben would have been false—that CCF_Ben is non-resilient to Adam’s doing otherwise. Under the second option, although CCF_Ben might have still been true, God must not have been so keen on having the first free creature in the garden freely eat the fruit. That means that, if Adam had done otherwise, CDF_Ben would have been false—that CDF_Ben is non-resilient to Adam’s doing otherwise. Given that these are the only two options, if Adam is able to do otherwise, it follows that he either has counterfactual power over some CCFs involving other creatures, like CCF_Ben, or he has counterfactual power over some CDFs, like CDF_Ben. Since both options are problematic, we should conclude that Adam is not able to do otherwise.

Brief summary

I have offered two arguments for the claim that Adam is not able to do otherwise than eat the fruit. The first appeals to the parallels between Adam’s case and other “self-defeating” cases, such as the Grandfather paradox; the second appeals to plausible claims about the limits of Adam’s counterfactual power. But it should be noted that these arguments do not generalize to every agent in every circumstance. I am only claiming that agents who find themselves in situations such as Adam’s are not able to do otherwise. The crucial feature of such cases is that, had the relevant CCF not been true, God wouldn’t have created the agent to begin with. There’s no reason to think that feature holds for every individual in every circumstance.

This distinguishes the challenge here from more traditional versions of what is sometimes called the “freedom problem” for Molinism (Hasker, 2017, p. 93). Not only do such versions typically invoke relatively controversial principles about freedom,¹⁶ but they also typically aim to show that such principles are violated in every case under Molinism, not just some.¹⁷ In contrast, the arguments given so far invoke less controversial principles and only establish a relatively modest conclusion.

In light of this, the Molinist need not feel seriously threatened at this stage of the argument. If she accepts that Adam is not able to do otherwise, she can simply reject the claim that freedom requires the ability to do otherwise. Indeed, several authors have suggested that Molinists ought to do just that, with William Lane Craig—a prominent Molinist—offering the clearest and most relevant concession:

I’m persuaded that so long as an agent’s choice is not causally determined, it doesn’t matter if he can actually make a choice contrary to how he does choose. Suppose that God has decided to create you in a set of circumstances because He knew that in those circumstances you would make an undetermined choice to do A. Suppose further that had God instead known that if you were in those circumstances you would have made an undetermined choice to do not-A, then God would not have created you in those circumstances (maybe it would have loused up His providential plan!). In that case you do not have the ability in those circumstances to make the choice of not-A, but nevertheless your choice of A is, I think, clearly free, for it is causally unconstrained… (2007)¹⁸

Although Craig does not articulate the argument in terms of self-defeating actions or counterfactual power, he clearly concedes the upshot of the foregoing arguments. So, at this point, we have good reason, even by the lights of some Molinists, to admit that Adam is not able to avoid eating the fruit.

Adam needn’t eat the fruit: circumstances, contingency, possibility, and ability

And yet, we also have good reason to admit that Adam is able to avoid eating the fruit. The argument proceeds in four steps. The first is to note what the “circumstances” in CCFs are meant to pick out.

While there is no standard account of what constitutes an agent’s “circumstance,” there is one point that is widely agreed upon: it is extremely inclusive. This is because of a formal feature regarding counterfactuals, namely, that “strengthening of the antecedent” is not truth-preserving. Consider the following counterfactual:

• 1. If Charlie were at home right now, she would be watching television.

Even if (1) is true, the following counterfactual appears false:

• 2. If Charlie were at home right now but the power was out, she would be watching television.

And although (2) is false, the following may very well be true:

• 3. If Charlie were at home right now, the power was out, but she had a backup generator, she would be watching television.

More generally, from the fact that a counterfactual of the form ‘p > q’ is true, it doesn’t follow that ‘(p & r) > q’ is true.¹⁹

Now consider CCF_Adam again: that if Adam were in the garden, he would freely eat of the fruit. If the circumstance specified by “the garden” isn’t inclusive enough, then it may also be true that “If Adam were in the garden but the snake wasn’t, he wouldn’t freely eat of the fruit.” If so, then CCF_Adam won’t be providentially useful enough for God: another agent could, for instance, interrupt God’s plan by kicking the snake out of the garden.²⁰

In light of this, it is commonly understood that the “circumstances” in CCFs, such as “the garden,” are simply a shorthand for a “maximally specified description of certain circumstances” (Cohen, 2015, p. 234). Different authors fill in the details in various ways,²¹ but for our purposes, we only need the following principle:

“All Ability-relevant Facts” (AAF): If the obtaining of a fact, F, is relevant to what an agent, S, is able to do in circumstance, C, then C specifies (or at least entails) whether F obtains.²²

Given how inclusive the circumstances are supposed to be in CCFs, this principle looks quite plausible, but I also offer a more general argument. Assume, for reductio, that F is relevant to what agent S is able to do in circumstance C, but that C doesn’t specify (or entail) whether F obtains. For simplicity, let’s suppose that F is relevant in that F’s obtaining would prevent S from performing action X, although this supposition isn’t necessary.²³ Given our assumption for reductio, it is possible for the following claims to be jointly true: (i) S would freely perform X, were S in C; (ii) that S is in C; and (iii) that F obtains. But if (i)-(iii) are jointly true, then F will prevent S from performing X and, a fortiori, prevent S from freely performing X. Thus, God’s knowledge of what S would freely do, were S in C, won’t be providentially useful enough for Molinism.

That is the first step of the argument: to note that the circumstance of “the garden” must include any fact that is relevant to what Adam (or any other agent) is able to do in that circumstance. The next step is to note that, since CCF_Adam is only contingently true, it is metaphysically possible for Adam to be in the garden and yet not freely eat the fruit. Using the standard “possible world” semantics for precision, since CCF_Adam is contingently true, that means there is a metaphysically possible world, w₁, which is accessible to the actual world, w_a, where CCF_Adam is false.²⁴ In order for it to be false at w₁, there must be a world, w₂, which is among all of the closest worlds (to w₁) at which the antecedent is true and the consequent is false. So, there is a world, w₂, at which it is true that Adam is in the garden but it is false that Adam freely eats of the fruit.

While the standard semantics certainly help to make this second step easier to see, I don’t think it’s absolutely necessary for the argument. For instance, consider a counterfactual that is necessarily true rather than only contingently so, such as: “If Adam were placed in the garden (at time t), Adam would exist (at t).” To deny such a counterfactual would be incoherent precisely because it’s incoherent for the antecedent to be true but the consequent false. Thus, to say that CCF_Adam is not necessarily true is to assert that it’s perfectly coherent that Adam be placed in the garden but not freely eat the fruit—that there’s nothing in the antecedent which necessitates the consequent. The standard semantics understands this “coherence” as there being a possible world, which allows us to add some precision. But even if one rejects talk of possible worlds, one should still accept the basic idea of this second step: the fact that CCF_Adam is only contingently true means that it’s perfectly “coherent,” “intelligible,” “possible,” etc., for Adam to be in the garden but not freely eat of the fruit. Nonetheless, I’ll continue to use the standard semantics in what follows with ‘w₂’ designating the relevant possibility.

Now, what is Adam doing in w₂ if he is in the garden but not freely eating the fruit? There are only two relevant options: (i) Adam is not freely doing anything, and (ii) Adam is freely doing something other than eating the fruit. The plausibility of each option depends on further stipulations of the case, particularly the details about the garden and Adam. For instance, if we specify that there is a demon in the garden, one who is willing to make Adam eat the fruit, should Adam show any hesitation, then option (i) will presumably be the most plausible option. But this brings us to the third step of the argument: that given not-too-far-fetched stipulations, option (ii) can be made the most plausible; that it is metaphysically possible for Adam to be in the garden and freely do otherwise than eat the fruit.

What are these not-too-far-fetched stipulations? Obviously, let’s suppose that there aren’t any demons like the one just mentioned—there are no (created) agents who are in a position to make Adam do anything. More generally, let’s suppose that there aren’t any obvious obstacles that would prevent Adam from doing otherwise. Moreover, let’s suppose that, although God’s ideal situation is one where the agent in the garden freely eats of the fruit, should it turn out that no agent would do so, God would prefer the agent in the garden to freely do something else rather than unfreely do anything. (Perhaps God has even elected to place the first agent in a paradisical garden rather than a desert, say, because the garden is so conducive to free choice.) Finally, let’s suppose that God favors placing Adam in the garden over Ben partly because Adam’s intrinsic properties are more conducive to Adam freely doing otherwise—that Adam, unlike Ben, likes vegetables just as much as he likes fruit, or that Adam’s psychology disposes him to think more carefully about these kinds of decisions, or whatever one deems relevant to the ability to do otherwise.

With these stipulations, what would stop Adam from freely doing otherwise? Adam would seem to have all of the relevant “intrinsic” and “extrinsic” properties. Indeed, if AAF is correct, every fact relevant to Adam’s freedom must be specified and held fixed by “the garden.” Why think that these facts rule out Adam’s freely doing otherwise? Or to put the point another way: if one denies this third step of the argument, then one is committed to the claim that, necessarily, if Adam is in the garden but not freely eating the fruit, he is doing something unfreely rather than doing something else freely. Given our stipulations about Adam and “the garden,” this claim seems ad hoc. So, I take this third step of the argument—that it is metaphysically possible for Adam to be in the garden and yet freely do otherwise—to have sufficient motivation.

This brings us to the last step of the argument: that this metaphysical possibility implies that Adam is able to do otherwise. Admittedly, it is widely noted that one cannot infer that an agent is able to perform an action from the mere metaphysical possibility of them doing so.²⁵ However, the cases which show that this inference is invalid are importantly different than the case here.

One kind of case which shows that mere possibility is insufficient for ability involves possibilities where agents find themselves in drastically different circumstances or with drastically different intrinsic properties.²⁶ Think of the out-of-shape sports fan, sitting on his couch, criticizing a striker for failing to score a difficult goal, and saying “I could have made that.” In terms of mere metaphysical possibility, the fan is right: there is a possible world where he scores the goal. But plainly, he lacks the ability to do so in the relevant sense since any world where he makes that goal is one where, first, he is not on the couch but instead on the field and, second, he is in much better shape.

Such cases don’t apply here. In worlds like w₂, Adam finds himself in the exact same circumstances—the antecedent of the CCF demands it. Likewise, Adam’s intrinsic properties needn’t be any different either. Recall that we stipulated that God chose Adam over Ben in part because Adam’s intrinsic properties are so conducive to the ability to do otherwise. And again, given a principle like AAF, everything relevant to Adam’s freedom must be the same. So, this kind of case isn’t worrisome.

Another kind of case involves possibilities where the agent’s performance of an action requires a massive amount of luck.²⁷ Imagine an average golfer facing an absurdly long and difficult putt. Arguably, she doesn’t have the ability to make it, given that any success would be incredibly lucky. But it’s certainly possible, even holding fixed her circumstances and intrinsic properties.

Such cases don’t apply here either. Think of that world, w₂, where Adam is in the garden and successfully does otherwise than eat the fruit. Was his success massively lucky? I don’t see any reason to think so. Unlike the golfer, there seems to be a reliable connection between his trying to perform the relevant action and successfully performing it; unlike the golfer, there’s no reason to think that worlds like w₂ constitute an insufficiently small portion of relevant worlds; unlike the golfer, it seems wholly appropriate to attribute or explain his success in terms of his abilities, etc.

So, the standard cases which show the illegitimacy of inferring ability from mere possibility don’t apply here. I am not in a position to prove that worlds like w₂ establish that Adam is able to do otherwise, in part because there are still large debates over when, exactly, an agent has that ability. But given that the circumstances must specify every fact relevant to Adam’s abilities, that they must be exactly the same in w₂, and that they have been stipulated to be highly conducive to the ability to do otherwise, it’s hard to see why w₂ wouldn’t be sufficient to establish the claim that Adam is able to do otherwise.

To recapitulate the four steps of the argument in this section: first, the circumstance of “the garden” must specify every fact that is relevant to Adam’s abilities; second, the fact that CCF_Adam is only contingently true implies that it is metaphysically possible for Adam to be in the garden and not freely eat of the fruit; third, that given not-too-far-fetched stipulations about “the garden,” it is metaphysically possible for Adam to be in the garden and freely do otherwise than eat the fruit; and fourth, that this metaphysical possibility implies that Adam is able to do otherwise than eat of the fruit while in the garden.

Before moving on, allow me two brief points. The first is to reiterate that, while I’ve invoked “possible worlds” in this argument (and others), it’s solely for precision’s sake. Notice that the summary of the argument I gave just above does not invoke talk of “possible worlds” whatsoever. So, I don’t think the argument can be avoided simply by abandoning such talk. The second is that this argument does not generalize to any agent in any circumstance—in particular, there’s no reason to think that the not-too-far-fetched stipulations about “the garden” will hold of every agent in every circumstance. Hence, even if successful, this argument does not commit Molinism to the claim that an agent acts freely only if she is able to do otherwise. It only establishes the modest claim that Adam is able to do otherwise in his particular circumstances. But if the arguments of the previous section are correct, even this modest claim is enough to spell trouble for the Molinist, for we now have compelling reasons to think that Adam is and is not able to do otherwise. In conclusion, let’s consider one last question: what can the Molinist say?

Conclusion: which facts are relevant to freedom?

Let’s suppose that the Molinist is moved by the arguments so far. How might she respond? One obvious answer is to deny the possibility of cases like Adam’s, but this strikes me as quite costly. There are two relevant features of Adam’s case: (i) that if the relevant CCF hadn’t been true, God wouldn’t have created the relevant agent in the first place, and (ii) that the circumstance the agent finds herself in are conducive to the ability to do otherwise. To deny the possibility of cases with these two features would severely limit the providential control that Molinism is supposed to afford. If anything, I suspect most Molinists will admit that there are plenty of actual cases that exhibit these two features.

Another obvious response is to simply “bite the bullet” and admit that agents can sometimes perform self-defeating actions and sometimes have counterfactual power over CCFs involving other agents or even CDFs. This also strikes me as costly, for obvious reasons, and some prominent Molinists seem to agree, as already noted.

There is a less obvious response that is worth discussing, especially since it raises a more fundamental question for the Molinist view. Return to the case of Tim and Grandfather. While it is widely accepted that Tim cannot kill Grandfather, everyone admits that there seems to be some sense in which Tim can do so. Indeed, what makes the Grandfather Paradox a “paradox” is that, at least initially, we seem to have equally good reason to think both that Tim can kill Grandfather and that he cannot.

The standard solution is to say that the sense (or restriction under) which Tim can kill Grandfather is distinct from the sense (or restriction under) which Tim cannot kill Grandfather. He can kill Grandfather in the sense that Tim has a gun, the necessary skill, the opportunity, and so on; he can’t in the sense that his very existence depends on Grandfather’s survival. In terms of possible worlds, although there are plenty of worlds where Tim successfully kills Grandfather, such worlds are fairly distant, involving natural laws that permit Grandfather to be later resurrected or some such thing. If we restrict our attention to only nearby worlds, then it is correct to say that “Tim can’t kill Grandfather”; if we don’t restrict our attention so, then it is incorrect to say so. Contradiction avoided; paradox solved.²⁸

This suggests a possible response for the Molinist: Adam can do otherwise in the sense that his circumstances allow for it, no one is coercing him, and so on; he can’t do otherwise in the sense that his very existence depends on his eating the fruit. And though there may be possible worlds where Adam does otherwise than eat the fruit while in the garden, these worlds are somewhat distant, requiring that certain (actually true) CCFs or CDFs be false. If we restrict our attention to only nearby worlds, then it is correct to say “Adam can’t do otherwise”; if we don’t, then it is incorrect to say so. If this kind of response can solve the Grandfather Paradox, why can’t it solve the puzzle here?²⁹

The issue with this response appears when we ask a further question: which facts or worlds are relevant to Adam’s freedom? Even if there are possible worlds where Tim kills Grandfather—e.g., worlds where Grandfather is later resurrected—such worlds are too “far away” to be relevant to what Tim can do in the “ordinary” sense, the sense at stake in debates over freedom. (Similarly for the out-of-shape sports fan mentioned above: even if there are worlds where he scores a goal, they are too distant to be relevant to what he is able to do in the sense relevant to freedom.) The question for the Molinist is: are worlds where Adam does otherwise than eat of the fruit while in the garden too distant or not?

In trying to answer this question, the same dilemma resurfaces. On the one hand, it looks as if the Molinist must claim that facts about (actual) CCFs or CDFs are not relevant to the “ordinary” or “freedom-relevant” sense of abilities, and so worlds where such facts are different are not too far away. Here’s why. Recall that CCF_Adam—the counterfactual that if Adam were placed in the garden, he would freely eat the fruit—is supposed to be contingently true. That means the facts picked out by “the garden” cannot include any fact which necessitates that Adam freely eats the fruit. Thus, “the garden” cannot include CCF_Adam itself, nor can it include facts about what other agents would freely do in “the garden,” like CCF_Ben, nor can it include facts about certain CDFs, like CDF_Adam.³⁰ But if AAF is correct, then every fact relevant to Adam’s abilities must be included in “the garden.” Taken together, this implies that these CCFs and CDFs are irrelevant to Adam’s abilities. Thus, worlds with differences in these facts need not be too far away to be relevant to Adam’s abilities. And this seems to be exactly the right result since Molinists have always claimed that the existence of CCFs and God’s knowledge of them are no threat to freedom.³¹

On the other hand, given the strong analogy between Adam’s doing otherwise and Tim’s killing Grandfather, or the plausible limits of Adam’s counterfactual power, such worlds must be too far away. According to the arguments of the second section, since Adam’s existence is explanatorily dependent on CCF_Adam, NSA implies that any world where CCF_Adam doesn’t obtain is too far away; and since Adam lacks counterfactual power over CCF_Ben and CDF_Adam, any world where those don’t obtain is too far away.

This brings us to our more fundamental question for Molinism: are facts about (actual) CCFs and CDFs ever relevant to an agent’s freedom? Traditionally, Molinists seem to answer negatively, but cases like Adam’s suggest that, at least sometimes, such facts can be relevant. If so, then the Molinist owes us a deeper story of when such facts are relevant and how this fits with the broader Molinist theory of providence.³² Without such a story, there’s no saying what we can do if Molinism is true.