The probabilistic structure of F-conditionals

I may not understand the problem, or your constraints on a solution.

It may be just a plausible claim about the nature of rational agency that the probability of an agent's performing some act increases when the strength of the reasons for performing that act increases; that the probabilities will be roughly on a par, if the strengths of the respective sets reasons is roughly on a par; and so forth. These are just platitudes about the relationship between action and the strengths of reasons that any account must satisfy. So these necessities explain why the space of F-conditionals is as it is.

Maybe there are some questions here about how such probabilities work out as they do given libertarian free will. But the distinctive Molinist additions do not seem to do any work.

It is plausible on the grounds of the nature of rational agency that: P(Curley would accept a $10,000 bribe in circumstances C) is greater than P(Curley would accept a $5000 bribe in circumstances C). That's pretty easy to explain, as you point out. But the conditional probability claim in (1) is substantially different.

Alexander,

Molinists typically think (so I’ve been told) that the F-conditionals are brute. (I think I had a post on this a while back.) The reason being is that there are problems with grounding the F-conditionals in the powers and dispositions of persons, viz., there aren’t any persons prior to creation when God takes these F-conditionals into account. But I assume you’re not objecting to the bruteness of the F-conditionals. Rather given the F-conditionals what explains probabilities about F-conditionals? I was first inclined to say that these probabilities are grounded in the powers and dispositions of the relevant individuals, but that won’t work because of the bruteness issue. Another option is that these probabilities reflect the truth of other brute counterfactuals. So (1) is grounded in (1*) Were it the case that Curley would accept a $5000 bribe in C then Curley would accept a $10,000 bribe in C. This sort of approach handles 1, 2, & 4. The independence issue might just fall out of the space of F-conditionals together with the standard probability calculus.

Hi Alex,

I take it that these propositions, more explicitly, look like this, Pr(B []-> A), where the condition B is in the antecedent of the subjunctive. So we are talking about the probablity of a subjunctive conditional. Is that right?
If that's so, why would Molinism be committed to such propositions? Why couldn't the Molinist get by with just "would" and "might" counterfactuals? They don't seem to need more for middle knowledge claims. I guess I'm not seeing the Molinistic need for more subtle formulations.

Mike:

The reason the Molinists should accept these claims is because they seem to be true. Doesn't it seem to be the case that if someone were such that she would take a bribe, she would probably also have taken a larger one?

Ted:

Would these probabilistic claims be brute and contingent or brute and necessary? I think brute and necessary. Otherwise we're adding one more contingent level on top of the F-conditional level, the level of the probabilities that govern the F-conditionals (hmm... but how can they govern them, if the F-conditionals are brute?). Besides, their plausibility seems be such that they should be necessary.

So, brute and necessary. Yes, one can say that. So we would have a structure of necessarily intricately interrelated probabilistic facts about F-conditionals, though brute. I don't know if brute facts can have such intricate interrelations. The interrelations do suggest that some probabilistic facts are likely to be prior to others in the order of explanation. So only some of them are going to be brute.

Alexander - you don't think the probabilities are epistemic, and I'm inclined to agree, but I think that a Molinist might be bound to disagree.

I suggest that the objective probability of any true statement is 1, and the objective probability of any false statement is 0. Call this assumption (A).

The Molinist thinks that God knows whether 'Curley would accept a bribe of $5000 in C' is true or false. If it is true, then, by (A) its probability is 1, if false its probability is 0.

If something is true, but I don't know that it is true, then I'm left with epistemic probability. Since I lack God's middle knowledge, the best I can do is come up with odds that Curley will accept the bribe in C.

Before a sporting event, one can figure out the probability that some team will win. After the event, they either won or did not. But if the event is over, but I don't yet know the result, it makes sense for me to estimate the probability that they did win. With regard to middle knowledge, God is like someone who has observed the game, I am like someone who doesn't yet know the result.

A further note concerning objective probability.

I can think of reasons to say that an event took place, and yet even in retrospect, it was improbable. "The race is not always to the swift", but it usually is. If the slower runner wins the race because the faster runner has a freak accident, we can say that it was an improbable victory.

If Molinism is true, then it is possible that in circumstances C, Curley will accept any bribe greater than $5000, except for a bribe of $5470. That might just be a brute fact.

Curley's wife prays to God to protect her husband's virtue. I am about to pay Curley a bribe of $5469, but God arranges that I should see a dollar on the floor, and decide to add it to my bribe to increase the chance of success.

From God's point of view, the probability of Curley's accepting a bribe of $5470 was zero. But God also knew that this number was, so to speak, a freak number, because a cent either way would have been the difference between sin and virtue for Curley: God can say that, thanks to divine grace, Curley had a lucky break.

If I happen to know that Curley has accepted bribes of $5000 and greater in the past, I would judge the probability of his accepting $5470 to be very high, and I would have good reason to do so, but I would be mistaken.

This reinforces the point that, if Molinism is correct, we are dealing with epistemic rather than objective improbability. 'Unlikely victories' or 'lucky breaks', after the event, are cases where the objective probability is 1, but a survey of all the facts indicates that before the event, evidence available to (non-divine) observers would suggest a very low probability.

The reason the Molinists should accept these claims is because they seem to be true. Doesn't it seem to be the case that if someone were such that she would take a bribe, she would probably also have taken a larger one?

But that is a reason for anyone to accept them, not Molinists in particular. I'm actually not sure whether they sound true. I'm not sure, for instance, that my intuitions distinguish well between,

1. Pr(B []-> A) > 0

2. B M-> A

That is, the chances that A would occur were B to occur is greater than 0 and A might be the case were B the case. When I assent to the plausibility of (1), it is not clear to me that I don't really have (2) in mind.

Mike:

Yes, I agree that everyone should accept these statements (cf. the last, parenthetical, paragraph of my post).

Ben:

Objective probabilities of true statements will be one if one is conditioning on these statements. Otherwise, not necessarily. The objective probability that an indeterministic coin flip that occurred today would be heads, conditioned on how things were just before the flip, is 1/2.

But even if these probabilities are epistemic, there is still the question of what grounds them.

Alexander, as I understand Molinism, F-conditionals are brute truths. They are capable of grounding other truths, but are not themselves grounded. The whole point of this theory is that, since human behaviour is free, it is not entirely predictable. At the same time, it can be admitted that there are tendencies to behaviour that most human beings have.

Curley has a tendency towards taking bribes. His will is weak and the temptation is, for him, strong. But sometimes, his weak will wins out against a strong temptation, in a way that defies human prediction, and that's just a brute fact. Only God can know when those occasions will arise.

My estimate of the probability that Curley will accept a bribe is based on past evidence of his taking bribes. I judge that since he has accepted bribes of $500, he is bound to accept bribes of $600, but less likely to accept bribes of $300, because the bigger the bribe is, the bigger the temptation. This judgement supposes that Curley acts like a machine, which he does - most of the time. What I don't know, but God does, is the ungrounded truth that if Curley is offered a bribe of $600, he will reject it.

So, the judgements about the probability of F-conditionals are grounded in a model of human nature that is deterministic. This is not a 100% accurate model of human nature, but it is accurate enough that these probability judgements are useful to us when planning to offer bribes. We will expect, if we are Molinists, that this model will occasionally lead to false expectations on our part, but we have no way of predicting when these errors will arise.

I did read about some research in neurology that would be consistent with the Molinist view (I tried to track down where I read this, but can't find it). According to this research, one can predict that we are about to make a decision before we become consciously aware that we are going to do so. However, when we become consciously aware, we have a brief opportunity to reject the decision - decisions arise from the unconscious, and can go directly to action, but the conscious mind can exercise a right of veto, although this happens only rarely.

The Molinist could say that it is this rare event that is predictable only to God and not to us: our judgements of probability of F-conditionals cannot take this factor into account.

Dear Ben:

This is a very helpful suggestion. But I am still not fully convinced. Here's one worry. Most of the time the model is right, you say. Why? Why should brute facts fits any model?

I guess my general worry is this: If a set of facts are irreducibly brute, how can there be probabilistic relationships between them or probabilistic "constraints" on them?

Maybe a part of the answer is as follows. We want to know what grounds the value of P(C*->A), where *-> is a subjunctive conditional. Answer: P(C*->A) reduces to P(A|C). (Cf. Edgington on subjunctives.) And there is no difficulty in grounding P(A|C): it is grounded in the laws (including laws of mental operation) that are a part of C (this only works if laws are a part of C; but a Molinist only assumes C*->A has truth value for maximally determinate C, so that's fine).

So now we have a story about what explains some of the probabilistic facts about F-conditionals.

It doesn't help with all of them, though. On the contrary, there is no way of making sense of (1)-(4) on the theory that P(C*->A) reduces to P(A|C). We now have a semantics for statements like P(C*->A)=p, but none for statements like P(C*->A and D*->B). So I am still stymied.

Alexander - I think that the Molinist’s best argument would be to say that F-conditionals such as (4) are not in fact true, but to explain why they should be intuitive. As I’ve said, I’m not in fact a Molinist, but rather than adding reminders, I’ll write the rest of this post as if I were a Molinist presenting my theory.

Clearly, much human behaviour is predictable, even when we are making an effort to be unpredictable. This is demonstrated by poker: a skilful poker player learns to read the personality of the other player, and form conditionals such as ‘She wouldn’t go all in unless she had a flush…’ with some success. But we are free, and that means that we have the ability to perform actions that go against our personal tendencies. Such actions cannot be the majority of our actions, because in that case, we would not have observable personal tendencies at all, and clearly we do have such tendencies. Of course, we are free whether we consent to our tendencies or resist them, but free acts of resistance to tendencies are much rarer than free acts of consent.

If I play poker against the devil, I’m likely to lose. Even if he cannot see my cards, he has the best possible knowledge of my tendencies. Perhaps he knows of tendencies that I don’t know about myself - without realizing it, I might have a tendency to bluff if one of my cards is a seven, because subconsciously it makes me feel lucky, or I tend to fold more frequently if I ate cheese the day before. What the devil cannot predict, but God can, is that sometimes I will resist my natural tendencies.

Through careful observation, we can hope to approximate the type of knowledge that the devil has (angels have this knowledge too, so in developing such knowledge we aren’t necessarily developing evil tendencies). In developing such knowledge, we are using our natural faculties as they were designed to be used, so whatever maxims are helpful in developing such knowledge will have an intuitive appeal. Any piece of information about the causal history leading to my opponent’s current hand is potentially useful for helping me to understand their natural tendencies as a card-player. But any piece of information outside of that causal history cannot play any such role. So, if we only take natural tendencies into account, a proposition such as

(4) is true: P(Curley would accept a $5000 bribe in C | Curley would accept a $5000 bribe in C*) = 1 if C and C* differ only in respect of something not in the causal history of Curley's bribe decision.

But in fact, (4) may not be true. Perhaps Curley would accept a bribe in C, but resist in C*, since his resistance is an unpredictable (to us) free action, not the result of anything in his causal history. If our natural faculties are functioning correctly, and we respond to grace, we will become Molinists and accept this theory, and we will realize that (4) is not really true. However, we will also realize that (short of divine revelation), (4) is as close as we will ever come to the truth: information outside of the causal history of the decision could never give us a reason for supposing Curley will or will not accept the bribe.

Since God knows all counterfactuals of creaturely freedom, there are probability judgements available to God that are not available to us. God knows the frequency with which F-conditionals that satisfy a particular description will turn out to be true. So God might in fact know that 99.9994% of the time, conditionals of form (4) are correct. All we now, and all we need to know, is that it is useful to rely on (4) when trying to figure out someone’s tendencies. In practice, when we make a false prediction, we will often be unable to tell whether it was because of some fact of causal history that brought to light a hitherto dormant personal tendency, or a free act of resistance to a personal tendency.

Alexander:

Like Ben, I'm not sure that (4) will generally hold true for the Molinist. That is, it's not clear that the Molinist can or will agree that P(S would choose X in C | S would choose X in C*) = 1 if C and C* differ only in respect of something not in the relevant causal history (even if it seems true). The worries arise primarily from the already-noted "bruteness" of F-conditionals on the Molinist picture, and the fact that they concern indeterministic scenarios. I take it that they are standardly taken to be contingent brute truths (mostly to assuage libertarian concerns), and so even if the probabilities that concern them are somehow fixed by the space of possible worlds (so as to be necessary), the probability of something like (4) will be less than 1 - i.e. there is at least one possible world in which it is true that S would choose X in C but false that S would choose X in C*.

Moreover, this issue is directly pertinent to an interesting recent paper (in progress) by Dean Zimmerman. A major step in his novel anti-Molinist argument depends crucially on the assumption that, for the Molinist, F-conditionals are brutely and contingently true, insofar as they are not grounded in categorical facts about the past in conjunction with the laws of nature nor those concerning the space of possible worlds. As a consequence of not being so grounded, Zimmerman insists that by strengthening the antecedent of a F-conditional with causally irrelevant categorical details about the distant past (such as is suggested in your (4)), its overall truth-value may well change - according to the Molinist (but not, of course, according to the standard semantical story). So the truth of a F-conditional with the one complete past C as antecedent implies nothing about the truth of another concerning the same free choice with another complete past C* differing from C only in some minor, remote and casually irrelevant way.

The upshot is that Molinist principles entail that by trivially adjusting the antecedent circumstances C—for example, by adding one more atom to the contents of the remote universe or changing the temperature of a distant star by .001°C—God could reliably bring it about that free creatures act and choose in exact accordance with his will. In such a scenario, God would enjoy “voodoo-like” complete control exercised by his manipulation of distant, insignificant details of the universe. And this, Zimmerman maintains, emasculates the freedom of those creatures—i.e. transworld manipulability (and the “voodoo”-worlds in which it obtains) is incompatible with genuine creaturely freedom, contra the Molinist suggestion.

Furthermore, he denies that there is any plausible "cutoff" boundary that the Molinist can offer. For instance, if the Molinist proposes the boundary be fixed by the past lightcone of the spacetime point in question, this will conflict with repeatability of the categorical circumstances just prior to the given indeterministic event (i.e. free choice).

So it seems to be a (at least potentially) big deal as to whether or not (4) is true, let alone necessarily true. Would you mind going into a little more detail as to why you think such a probability should be taken as necessary and why that won't end up offending the libertarian sympathies of the traditional Molinist (of which I am not one)?

This is all helpful, Ben. Thanks for taking the time.

In a paper I once used a stronger claim that was like (4), and Tom Flint objected to it. I suspect he would have objected to (4) as well. So I think you're right that the Molinist would reject (4).

Still, two issues remain. First, there is actually a plausible principle that yields (4):

Difference Principle: Suppose C and E are actual events, and C* is a possible but non-actual event. Assume that the following holds: Were C* to have occurred instead of C, then E might not have occurred. Then the difference between C and C* is causally relevant to E.

My second worry is this. Let's take the claim that most of the time we follow our natural tendencies. This seems to be a probabilistic claim that P(x would do A in C|x has a natural tendency to do A in C) is high or at least that P(x does A in C|x has a natural tendency to do A in C) is high when C actually obtains. What grounds this probability claim?

Maybe it's a brute fact, so there are other (presumably infeasible) worlds where it's not true? But if it's a brute fact, what reason do we have to put confidence in its future continuation? (Suppose that it was a brute fact that every hour hitherto, without cause and for no lawful reason, a brick jumped into existence ex nihilo atop the Great Pyramid. If the fact was brute, induction would not justifiedly apply, I think.)

Alexander, I have a Molinist response to your most recent suggestion. I don’t know how to answer the points raised by Daniel.

The Molinist model I’m considering is that free-will involves resisting or consenting to natural tendencies. Let me add to this that we are not going to resist unless we have a motive to resist, such as the realization that the exercise of our natural tendency conflicts with our long term goal.

I think this is in keeping with strong libertarian. The strong libertarian isn’t suggesting that, even though I know smoking is bad for my health, and that the initial experience of smoking would be unpleasant, and that it would be an expensive habit, I am still likely to take up smoking for no reason. Having no reason whatsoever to smoke, I simply don’t buy cigarettes: free but predictable.

On the other hand, I like to eat chocolate. I realize that this is bad for my teeth. So my habit conflicts with desire. When placed in temptation, in a position where I know the long-term desire conflicts with natural habit, my decision is an unpredictable brute fact. Since there are two possibilities, we could describe it as a fifty-fifty chance, but this is simply a way of saying we have no reason for preferring one prediction to the other.

However, this conflicts with my statement that we should expect people to consent to a natural desire more frequently than to reject such a desire. Let me suggest a couple of factors that would explain this: first, repeated actions may change natural habits, and secondly, resisting natural habits may sap energy.

I am tempted by chocolate. I resist. If this happens repeatedly, my desire for chocolate may be reduced. My natural tendency is now in accord with my long-term desire, and so I will not eat chocolate, nor will I be tempted to, unless something happens to give me a new long-term desire that involves eating chocolate, in which case I will have to change my natural tendency again. So, if I repeatedly resist a natural desire, the net effect is likely to be that I no longer have a conflict between goals and desires, and so no need to resist. When I then decline chocolate in accordance with my desire, my action is free because it is a desire that I chose to have.

I’m also saying that resisting the natural habit reduces one’s energy. This means that if I resist natural desires too often, I need to sleep longer. If I’m deprived of this extra sleep, I am likely to die sooner. So too much resistance of natural desires results in less conscious activity (either because of more sleep or a shorter life) and thus less decision-making. Of course, the only people who will suffer as a result are those who repeatedly resist a natural desire (to eat chocolate), thus generate a new tendency (to resist chocolate) then break this tendency, and so on. Such people may well exist, but will, on average, engage in fewer conscious decisions because of the energy-sapping life-style. So, the sum total of human conscious decisions will involve more decisions where people accept natural tendencies.

One would have a reason, in certain rare situations, to keep on developing then breaking natural tendencies. A man played cards with the devil all night. The man won, but he lost thirty years of life in the process.

Daniel, you write,

The upshot is that Molinist principles entail that by trivially adjusting the antecedent circumstances C—for example, by adding one more atom to the contents of the remote universe or changing the temperature of a distant star by .001°C—God could reliably bring it about that free creatures act and choose in exact accordance with his will. In such a scenario, God would enjoy “voodoo-like” complete control exercised by his manipulation of distant, insignificant details of the universe. And this, Zimmerman maintains, emasculates the freedom of those creatures

I don't offhand see the manipulation problem. Suppose it is true that C are some specified circumstances and C []-> B. Suppose further that C & A []-> ~B, God knows this, and A is the addition of some atom to C. So long as it is possible that the very same circumstances C hold along with A and you do B, I don't see where anyone loses their freedom. That is, so long as we still meet the liberatarian conditions for freedom, where is the worry about manipulation? Indeed, I don't right off see why this would be called manipulation, if masnipulation is supposed to conflict with freedom in this situation. So, how does the problem arise?

Dean Zimmerman's article, by the way, is here: http://fas-philosophy.rutgers.edu/zimmerman/Anit-MolinistArgument(January07).pdf

I am unconvinced by it. In fact, it seems to me that the main point of Molinism is to make such manipulation possible, thereby rendering sovereignty, efficacious grace and libertarian free will compatible. To claim that the manipulation is freedom-canceling seems to me to beg the question against the Molinist.

The new thing in the paper is the claim that small changes in antecedents can change the truth value of the conditionals. This allows the manipulation to be more subtle. But I don't think the subtlety of the manipulation affects the issues.

Mike:

Good question. My grossly condensed summary of Zimmerman's argument (which, in total, exceeds 50 pages) doesn't provide a ready answer and I now realize it is potentially misleading in the following way: Zimmerman contends only that Molinist principles entail that there is at least one possible* world in which every possible choice of every possible creature not only could, but would be controlled by God’s manipulating distant, irrelevant facts. In such a world, every possible creature would be subject to transworld manipulability (TM), according to which, for every subjunctive conditional of freedom (SCF) of the canonical form

SCF1: If S were in circumstance C, S would freely choose A,

that happened to be true, there would be an arbitrarily close circumstance C* that God could bring about such that, if C* were to obtain, S would freely choose ~A. That is, were SCF1 true in such a world, the following conditional would also be true:

SCF2: If S were in circumstance C*, S would freely choose ~A.

Moreover, if SCF1 happened to be false, there would be an arbitrarily close circumstance C** that God could bring about such that, if C** were to obtain, S would freely choose A. That is, were SCF1 false, a conditional of the following form would be true:

SCF3: If S were in circumstance C**, S would freely choose A.

Thus, by trivially adjusting the antecedent circumstances God could guarantee that, as I said before, free creatures act and choose in exact accordance with his will. And Zimmerman thinks that the mere fact that Molinism and its ‘brute’ SCFs imply the possibility of transworld manipulability (which he takes to be incompatible with creaturely freedom) is reason enough to reject it.

You seem to be suggesting that such manipulation is irrelevant to one's freedom so long as the circumstances in which one is situated are indeterministic (i.e. not determined by past + laws). But I guess Zimmerman would simply disagree: as we all know, mere indeterminism is necessary, but not sufficient to guarantee libertarian freedom. Rather, there is some additional condition that must be met (likely having to do with control or self-determination) and, presumably, this condition is what is intuitively being violated in the above scenario.

Zimmerman himself acknowledges that many Molinists would be prepared to accept the above and still see no problem in countenancing TM merely as exceedingly improbable. They will retort that the existence of SCFs and their use by God in his providential planning of a world does not render a person unfree, and the same is true of a voodoo-world. Creatures remain free so long as God is unable to “pick and choose” which SCFs are true and, according to Molinsim, he can’t— SCFs and their brute truth-values are given to him, not determined by him.

Again, however, Zimmerman simply disagrees: he thinks that transworld-manipulable creatures clearly are not free. A relevant passage: “The harder I think about the circumstances of a creature afflicted with transworld manipulability, the harder it is for me to believe it really would be free. How could I still be free, if someone else possessed the means to determine the outcome [i.e. ‘decide what it will be’] of every possible kind of choice with which I might be confronted?” (29).

I myself have some fairly serious reservations about Zimmerman’s argument, especially concerning the dialectical situation (and mainly having to do with the fact that the means of God’s control in normal and voodoo worlds - so-called “weak actualization” control - remains the same in kind/nature, differing only in degree, but the latter seems irrelevant to the freedom question), but this post is long enough…

Thanks Daniel, quoting Zimmerman,

Again, however, Zimmerman simply disagrees: he thinks that transworld-manipulable creatures clearly are not free. A relevant passage: “The harder I think about the circumstances of a creature afflicted with transworld manipulability, the harder it is for me to believe it really would be free. How could I still be free, if someone else possessed the means to determine the outcome [i.e. ‘decide what it will be’] of every possible kind of choice with which I might be confronted?” (29).

It seems to me mildly question begging to describe this as "manipulation", but that is a minor point. Here is a way to test whether the relevant agents are free or not. Suppose for reductio that the manipulation undermines freedom. We should expect that manipulated agents are also not responsible (or not fully so) for what they do. Here's a counterexample. You threaten to dump house paint on my car, were I to retrieve my newspaper before 10am Friday. Your threat is genuine and I know that. Let this be a counterfactual of freedom,
I retrieve my newspaper before 10am []-> You dump paint on my car.
Suppose I refuse to curb my paper-retrieving habit despite your threat. I go and retrieve the paper at 9:55am. You thereafter pour the paint on my car.
Did you pour the paint freely? I say yes, and I say that you are fully responsible for the action, despite the fact that I knew what you would do were I do retrieve my paper before 10am, and I did it anyway. I don't think that constitutes manipulating your behavior. So it is not enough for manipulation that I know that A []-> You freely do B and I do A anyway. It is not enough either, it seems to me, that this is true of all of the counterfactuals of freedom relative to a particular agent. At the very least, God would have to be intending to manipulate the action of free agents, and it is hard to see how/why God might have such an intention. I should note that Zimmerman's point seems vaguely similar to a point made by Luke Gelinas on PB a while back.

Mike:

Richard Gale (I am thinking of what he says about Molinism in On the Nature and Existence of God as well as of conversation he and I had on his argument) distinguishes between one-time control and control of a whole life. For instance, it might be that Jones has such a character that he not only wouldn't but couldn't refuse a minor request from a friend that costs him almost nothing, and the friend might know this. Nonetheless, granting the request is something Jones is responsible for.

But if, on the other hand, Jones' whole life were under the control of someone else, then Jones would not be free.

Why does this make a difference? Well, I'd say that if it's a case of a single action, then Jones is going to be at least derivatively free--i.e., his action is going to flow from a character that was partially determined by earlier choices that themselves were libertarian-free.

Principle: If A has control over all of B's choices, then B is responsible for none of her choices.

Now, Zimmerman contends that it is the whole life that is under divine control.

I am still unconvinced, since presumably one of the points of Molinism is to allow God to exercise control over the whole life of a person, and so the Principle may simply beg the question.

Moreover, Zimmerman hasn't shown that God has to exercise the said control. Double effect considerations might creep in. God might know that putting Jones in situation C will result in Jones' mowing the lawn, but nonetheless he might only intend that Jones be in C, not that Jones should mow the lawn.

Mike, I think (something like) your objection can be effectively generalized in the following way (I wrote a paper advancing this objection, among other things, a while ago).

First, it may be noted that Zimmerman’s argument seems to ultimately boil down to an expression of his incredulity concerning the compatibility between freedom on the part of the controlled subject and what we might call “strong weak actualization control” (or SWA-control), according to which one who enjoys SWA-control over a subject S is able to weakly actualize (and hence securely bring about) any choice made by S as desired. In this way, Zimmerman insists, anyone (such as God) who enjoys SWA-control over another (perhaps akin to Gale’s “control over a whole life”) thereby abrogates the freedom of the controlled subject.

But, it may strike one as odd that Zimmerman originally is willing to grant (at least “for the sake of argument”) that weakly actualizing a state of affairs involving free choice is freedom-preserving—he even goes so far as to say that this supposition, one that is not unique to Molinists, is “plausible”—but now implicitly renege on it when SWA-control is under consideration. Taking an exercise of ordinary weak actualization-control (or OWA-control) to be an intentionally bringing about of circumstances C such that one knows how a subject S will act in C and bringing about C because one wants S to act in that manner (and the possession of OWA-control just is the ability to so exercise control), we can say that Zimmerman is willing to admit the compatibility of OWA-control with freedom on the part of the controlled subject, but not that of SWA-control. The irony here should be apparent. After all, the difference between SWA- and OWA-control is not one of means, but merely of extent. The nature of the control is isomorphic; rather, SWA-control is just a scenario in which one possesses extensive OWA-control. So if one shares Zimmerman’s intuitions about SWA-control being incompatible with freedom on the part of the controlled subject, one should cast a suspicious eye on instances of OWA-control as well, since the same kind of control is being exercised there as in SWA-scenarios, just to a (far) lesser extent. But insofar as instances of OWA-control are intuitively freedom-preserving (and this much seems widely accepted, even initially by Zimmerman), so one should not balk at instances of SWA-control. In sum, the extent seems irrelevant to whether freedom is compatible with the means of control.

Of course, there seems to be logical space here between “one-time” OWA-control and “whole life” SWA-control that Zimmerman seems to be appealing to. That is, perhaps he can be read as suggesting that while instances of OWA-control are uncontroversially freedom-preserving, the degree or extent of freedom enjoyed by the controlled subject diminishes in proportion to the increase in the degree or extent of OWA-control. So perhaps there is a sliding scale describing the relationship between weak-actualization control exercised over S and the freedom possessed by S. But that seems plausible only insofar as one thinks of freedom as a degreed phenomenon—as something that is or can be possessed in degrees. And there is certainly a ring of plausibility to that supposition. Freedom is something that can be increased and diminished as well as gained and lost. But some might think that there would be the need for a cutoff at some point, marking the boundary between worlds in which most or all the creatures are free (i.e. non-voodoo worlds) and those in which most or all are not free (i.e. voodoo worlds). And if so, then Zimmerman would thereby be committed to some limit principle or threshold concerning where WA-control conflicts with the freedom of the controlled (let’s call it LIMIT*). But if that is the case, then arguments parallel to those that Zimmerman presents against LIMIT—the original limit principle that he thinks Molinists are objectionably committed to with respect to which distant facts are relevant (when embedded in the antecedent of a SCF) to the truth of SCFs concerning creaturely action, but which he rejects as implausible on the basis of there being “as no arbitrary metaphysical cutoffs”, etc.—would seem to work against LIMIT*.

Considerations involving divine “intention” and potential double-effect are also very interesting and relevant, although I’m not sure what Z-man would say.

But lastly, I should note that Zimmerman is aware of some of these criticisms (I’ve had minor correspondence with him), so I expect his revised paper (since I believe it is still a work in progress) will take them into account and, presumably, offer responses.

Of course, there seems to be logical space here between “one-time” OWA-control and “whole life” SWA-control that Zimmerman seems to be appealing to.

Reconsider the whole-life-manipulation hypothesis. I think it's mistaken for at least one clear reason. I am supposed to be worried that there is a world W such that, for every SCF of the form C []-> I do A, there is some event e that God might strongly actualize such that (e & C) []-> I do ~A. I allegedly lose my freedom under such an assumption (or, at least if God does in fact bring about the e's.). Here is why I am not worried. Since I have libertarian freedom in W, I know that there is some W'(not among the closest worlds to W) in which it is true that, for each action A, (e & C)& A, and so true that (e & C) []-> A. We lose no generality in assuming as well that, in W', it is true for all actions ~A, C []-> ~A. So, it is true in world W' that (e & C) []-> A and C []-> ~A, (e & C & A). Yet it is evident that I am free in W', since in W' I have acted contrary to the attempted manipulation in W.

Once again, in brief. There is a world W' (not among the closest worlds to W) in which I exercise my freedom contrary to God's attempt to manipulate me in W, and it is true there (in W') that e & C & A. It is therefore true in W' (assuming strong centering) that (e & C) []-> A. In at least one of those W' worlds it will also be true, for all actions ~A, that C []-> ~A (since weakening antecedents is invalid). We have in W' what Zimmerman would call a world in which I am manipulated. But that's false! In W' I have acted contrary to the attempted manipulation in W! So I am free in W'.

Argh, I just noticed that the counterexample above works only if we assume weak (not strong) centering in W'. Only then do A & C not entail C []-> A. But that's too much to assume.

Let me try a modified example that, I think, works as well. Suppose we are in W and it is true that (e & C) []-> A and C []-> ~A (for a whole life). This is the situation in which I am allegedly not free. It is consistent with everyhting we know about W that there is some W' accessible from W (but not among the closest worlds to W) such that (e & C) & ~A. Let's keep strong centering. It is true in W' then that (e & C) []-> ~A and C []-> ~A. It's pretty clear that, in W', I am free. God's actualizing e did nothing to affect what I would do. But then it is true in W that (i) I am not free and (ii) I can strongly actualized a world W' in which all of my actions are free. But if I can (in W) strongly actualize a world W' in which all of my actions are free, then I am free in W.

' P(Curley would accept a $10,000 bribe in circumstances C)'

Does a complete description of the circumstances which exist when Curley is offered a bribe include :-

1) An omniscient being who knows Curley will accept the bribe

2) An omniscient being who knows Curley will refuses the bribe

3) An omniscient being who does not know whether or not Curley will accept the bribe

4) No omniscient being exists when Curley is offered the bribe.

One of those 4 must be included , if you want a complete description of the circumstances which obtain at the moment Curley is offered a bribe.

But is there a fact of the matter about which of those 4 conditions obtains?