An argument against the possibility of transworld depravity
January 31, 2010 — 21:36

Author: Alexander Pruss  Category: Molinism  Comments: 24

Transworld Depravity (TWD) is the thesis that possibly every feasible world with significantly free agents contains moral evil. I will offer an argument, assuming Molinism, that TWD is necessarily false. I don’t think the argument is all that strong, but I hope it will push Molinists to think about a certain interesting (to me) issue.

In order to get Adams to accept some counterfactuals of creaturely freedom (CCFs, denoted with →), Plantinga offered this example. Actually Curley takes a bribe of a certain amount. Surely, then, it is true that were Curley to have been offered a larger bribe, he would have taken that, too. Adams agrees.

One might not unreasonably take Plantinga’s example to support the following thesis:
(*) Necessarily: If x actually freely chooses A in circumstances C, then had x instead been in circumstances C* instead of C such that D(C*,C,x,A), then x would still have freely chosen A.
Here, D(C*,C,A) says that circumstances C* are a variation on C (this minimally implies that they occur in the same spatiotemporal location, but more may need to be added), and they dominate circumstances C for x in respect of A in the following sense: (a) the agent is non-perverse and hence without the least inclination to act unreasonably for the sake of acting unreasonably, (b) any consideration operative for x in C in favor of A is also operative for x in C* in favor of A in at least as strong a form, and (c) any consideration operative for x in C* against A is operative in C against A in at least as strong a form.

One might then generalize (*) to:
(**) If C and C* are sufficiently determinate circumstances for a free choice, then (C → x freely does A) & D(C*,C,x,A) entails C* → x freely does A.

Suppose (**) is true. Imagine circumstances C where there is only one free agent, Eve, who makes only one free choice: whether to eat a yummy apple or to dance a merry jig (no other options are available, and it is not possible to do both), and this choice is significantly free because God forbade Eve to eat the apple. Eve has no inclination to disobey God or act unreasonably as such. Eve, however, has a desire to eat the apple on account of its yumminess or to dance the jig on account of its merriness. Call these circumstances C. Now, let C* be circumstances just like these, except that God instead forbade Eve to dance the jig.

Now, suppose TWD holds. Then, C→(Eve freely eats apple) and C*→(Eve freely dances jig). But this contradicts (**), since C* dominates C in respect of apple-eating for Eve. Why does domination hold? Well, any operative consideration in favor of apple-eating in C (namely the yumminess of the apple) is present in C*, and any operative consideration against apple-eating (namely the merriness of the jig) in C* is present in C. The only difference is that the fact that God forbids the apple-eating in C but it is the jig-dancing that is forbidden in C*; but given that Eve has no inclination to act unreasonably or disobediently as such, this does nothing to contradict C’s being dominated by C* in respect of apple-eating (that God forbids apple-eating in C either counts for nothing or counts against apple-eating in C, etc.)

The argument for (**) from (*) is, roughly, as follows. Suppose C→A and ~(C*→A). Then, were C to be actualized by God, it would still be the case that: C→A and ~(C*→A), since God is unable to affect the truth values of CCFs. But by (*) it is impossible to have C and C→A and ~(C*→A), since it is not possible to have C and A and ~(C*→A).

So, the defender of the possibility of TWD needs to deny (*). But I think (*) is pretty plausible. In fact, it seems just to be a consequence of what “is operative for/against A in at least as strong a form” means.

Suppose, for simplicity, that C and C* differ only in respect of the fact that C* contains one or more additional considerations in favor of A. (In fact, that is the only case my counterexample to the possibility of TWD needs: in C* there is an additional consideration in favor of apple-eating, namely that jig-dancing is forbidden by God.) It is surely very plausible to say that, necessarily, given the above, if x had an additional reason in favor of A, she would (still) have chosen A, at least assuming the non-perversity condition that she has no inclination to act unreasonably for the sake of acting unreasonably. (Some people think the non-perversity condition is always satisfied. But that’s controversial.)

So, the Molinist defender of the possibility of TWD needs to deny (*). But the intuition in Plantinga’s Curley example is pretty strong, as are, I think, the above arguments. (I am thinking that the Molinist’s best move here is to say that her counterfactuals are different in kind from the counterfactuals used to test what considerations are operative.)

What this points out is that the possibility of TWD is, basically, a sort of logical independence thesis for CCFs: Even if C and C* are related by domination, there are no entailments between C→(x does A) and C*→(x does A).

Here is another interesting result. Even if (*) is false, the intuition about Curley is pretty plausible. I think this means that if C→(x does A) and D(C*,C,x,A), then it is very probable that C*→(x does A). But, if so, then my argument establishes that it is very probably that TWD is false, assuming Molinism.

Comments:
  • Mathis

    Has anyone here read Richard Otte’s paper “Transworld Depravity and Unobtainable Worlds”?
    http://people.ucsc.edu/~otte/articles/otte.twd.pdf

    February 1, 2010 — 8:14
  • I’ve skimmed it. It’s quite a nice paper. It’s also surprising that something like this hasn’t been written earlier–that’s a mark of a really good paper. I also think it’s basically right, though I don’t know if it solves all the related problems. It may solve the problem of what happens when God pre-announces free actions (modulo some particularly thorny problems for Molinists when the prophecy itself is a part of the circumstances of action). But it leaves some similar problems. Here is one: What if God, prior to the action, announces not what the action is, but what the truth values of the relevant conditionals of free will are? And another: What if God, prior to the action, does not announce what the action is outright, but makes several announcements that only together entail the action? Which of the announcements are to be kept in the initial segment?
    However, if (**) holds, then even after all of Otte’s improvements to the FWD have been made, we have a counterexample. In Otte’s terminology, if (**) holds, then, necessarily, a morally perfect world is obtainable (but it is a contingent matter of fact which world that is; so, a Molinist God could create a morally perfect world, but a God who has simple foreknowledge might well be taking a risk of creating an imperfect world).

    February 1, 2010 — 8:36
  • Mike Almeida

    What if God, prior to the action, announces not what the action is, but what the truth values of the relevant conditionals of free will are?
    Why think the announcements are not restricted by the possible facts? We can just as easily say that those conditionals are prevolitional, so he’d have to announce in that world the truth-values they have in that world.

    February 1, 2010 — 9:14
  • Mike Almeida

    (*) Necessarily: If x actually freely chooses A in circumstances C, then had x instead been in circumstances C* instead of C such that D(C*,C,x,A), then x would still have freely chosen A.
    Plantinga would deny (*), and for good reason. If (*) is supposed to hold for every C*, then since C itself is such a C*, it is necessarily false that God actualizes C and x does ~A. So x is not significantly free. But x is significantly free, hence (*) is false.

    February 1, 2010 — 9:28
  • Mike:
    Suppose C=C*. Then, D(C,C,x,A) holds trivially. So, in that case, (*) says:
    Necessarily: If x actually freely chooses A in circumstances C, then had x been in C, x would have freely chosen A.
    This is a case of Centering. However, it is a case of Centering that a Molinist ought to accept. Here’s why. Molinists accept that when the antecedent is sufficiently determinate and the consequent reports a free choice, Conditional Excluded Middle (CEM) applies. But CEM entails Centering. Here’s the argument: By modus ponens, (C→~p)&C entails ~p. Therefore, C&p entails ~(C→~p). But Necessarily(C→p or C→~p), by the restricted version of CEM. Thus, C&p entails C→p.
    Of course, if the “necessarily” had been around the consequent, the Molinist would reject this. But the “necessarily” is around the conditional, not the consequent.

    February 1, 2010 — 9:55
  • I fixed a mistake in condition (c) of the definition of D. I had “at most” where “at least” was called for.

    February 1, 2010 — 10:01
  • Mike Almeida

    Necessarily: If x actually freely chooses A in circumstances C, then had x been in C, x would have freely chosen A.
    That’s not trivially true. Let @ be the actual world at which x in C freely chooses A. Let W be a possible, non-actual world in which C also obtains and x freely does ~A. The antecedent in (*) is also true at W. If x actually does A, then it is necessarily true that x actually did A (cf. if it is raining here, then it is true in Pakistan that it is raining here). W is not the actual world despite the fact that it is true in W that it is actual. And since x is significantly free, we know there is such a W in which C obtains and x freely does ~A. So we have a world in which the antecedent of (*) is true, and the consequent is false.

    February 1, 2010 — 10:52
  • Good point. I meant “actually” to have narrow scope (and that’s how my arguments take it). So, for explicitness, just drop the “actually” in (*).
    This isn’t strictly speaking a mistake on my part, because in English we do use “actually” with narrow scope. “You said you studied biology but actually you didn’t. Were you to have actually studied biology, you would have known that p.”

    February 1, 2010 — 12:09
  • Mike Almeida

    Take a C* that differs from C only in it’s relational properties. C* occurs in the context of an ideal world, and C does not. We have assumed that x is significantly free and not TWD, so there is some C* in some context B such that were C* []-> freely does not eat the apple. Otherwise, x is not significantly free.
    A different worry. The assumption that x is TWD is a fact that is operative in C and C* and that explains why x goes wrong in each set of circumstances. This can appear to be an appeal to the brute fact that x goes wrong in each set of circumstances to explain why he goes wrong. But Plantinga does have an argument. His argument is that it seems possible that libertarian free creatures are such that they would go wrong no matter the circumstances in which they are created. And indeed that certainly does seem possible given libertarian free agents. (*) seems intuitive only if you don’t believe there are libertarian free agents in Plantinga’s sense (not Kane’s sense).

    February 1, 2010 — 14:27
  • The first worry is handled by appropriately restricting what goes into C / C*.
    The “seems possible” intuition has two readings:
    1. (C)(if a libertarian significantly free creature x is created in C, then possibly: x goes wrong in C)
    2. Possibly: (C)(if a libertarian significantly free creature is created in C, then x goes wrong in C)
    (1) is compatible with (*) (with an appropriate restriction on the quantification over C)
    (2) is incompatible with (*).
    I think that if one has the intuition (2) even after disambiguating, then that intuition is simply due to not having thought enough about circumstances like C and C*.

    February 1, 2010 — 15:13
  • I just posted a draft of a paper which argues all this a little more rigorously.

    February 1, 2010 — 16:36
  • Mike Almeida

    The first worry is handled by appropriately restricting what goes into C / C*.
    That is not non-question-beggingly true. We know that if x is libertarian free, then there is a world in which C holds (i.e., all of the intrinsic facts in C that God can strongly actualize hold) and x does not freely eat the apple. And we know that for any C*, there is a world in which x eats the apple and a world in which x does not. That simply follows from libertarian freedom as Plantinga understands it. So you’d need some other argument to show that Plantinga’s view of libertarian freedom is mistaken. But there is no such argument here. On the contrary, we have a principle that begs the question against the possibility of Plantingan libertarian freedom.

    February 1, 2010 — 20:22
  • Mike Almeida

    2. Possibly: (C)(if a libertarian significantly free creature is created in C, then x goes wrong in C). (2) is incompatible with (*). I think that if one has the intuition (2) even after disambiguating, then that intuition is simply due to not having thought enough about circumstances like C and C*.
    I guess I don’t think Plantinga’s view could be plainer in NN. It’s clearly the wide scope reading; the narrow scope reading is not relevant to his discussion. If you think hard about what libertarian freedom minimally entails, you’ll get the intuition that (2) is true. The least you can say about libertarian freedom is that it entails branching futures at every point of significant choice compatible with the same past.

    February 2, 2010 — 7:23
  • Mike:
    Could you spell out the branching futures assumption more clearly? Thanks!

    February 2, 2010 — 8:43
  • John A.

    “In order to get Adams to accept some counterfactuals of creaturely freedom (CCFs, denoted with →), Plantinga offered this example. Actually Curley takes a bribe of a certain amount. Surely, then, it is true that were Curley to have been offered a larger bribe, he would have taken that, too. Adams agrees.”
    Maybe Curley has a reason not to accept a larger bribe. He may think that the smaller bribe will be easier to hide while the larger one will bring undesired attention to him. If this is possible then that claim that he would surely take a larger bribe is false.

    February 2, 2010 — 10:45
  • John:
    That’s exactly right. So to get the Curley case to work, we need to assume domination of the reasons for taking the $35K bribe by the reasons for taking the $36K bribe. In the case you give, there is no such domination. I discuss this in the draft paper.

    February 2, 2010 — 11:39
  • John A.

    Alexander
    I will read the draft (I am looking forward to it) but one question. If we add by 1K increments and conclude that Curley would accept more if it had been offered instead of the 35K, why not subtract in 1K increments and argue that Curley would take less if offered? This seems like a ‘paradox of the heap’ problem where there is no reason to stop going in either direction so that we would end up with a world w/o evil by subtracting at one end and an infinite number of worlds going towards more evil by adding if the k’s represent increments of evil. If that is true then TWD would be false because there would be one world w/o evil because Curley has no reason to accept the bribe.
    Anyway. I am looking forward to reading your draft.

    February 2, 2010 — 15:00
  • Mike Almeida

    Could you spell out the branching futures assumption more clearly? Thanks!
    It’s a minor assumption, Alex. Sometimes libertarian freedom is actually defined by appeal to branching futures. Similarly, for obvious reasons, are indeterministic events. Let an event/action e be undetermined at t just in case the laws of nature L and history of the world prior to t is consistent with e at t and consistent with ~e at t. You then have two possible future branching from (consistent with) a single past. The same minimal condition holds for libertarian events/actions. It holds too, as far as I can tell, for forms of libertarianism which appeal to agent causation, or to events occuring “in” the agent, and the like. All of that is consistent with the existence of two possible futures branching from a single past.
    There is a much stronger argument from libertarianism to the rejection of would-CCF’s. Though Hawthorne argues (and I agree) that this shows more about the problems with Stalnaker-Lewis semantics than it does about which counterfactuals are true.

    February 2, 2010 — 16:22
  • Sure, but I don’t see how that refutes (*).

    February 3, 2010 — 7:55
  • Mike Almeida

    If for each point of choice n and action A with past P, the future branches into (at least) two possible histories h and h’ such that A occurs in h and ~A occurs in h’ and such that h and h’ share P, then (*) is false. There would have to be some P = C* such that P has possible futures h and h’, A occurs in h and ~A occurs in h’. That’s the idea. This is why I think (*) is question-begging. But maybe you can show it does not beg the question against libertarianism.

    February 3, 2010 — 8:34
  • Everybody:
    I uploaded a new version of the paper (which I’ve also just submitted to a journal).
    Mike:
    I deny the first sentence in your comment, but I affirm its antecedent.

    February 3, 2010 — 20:55
  • John A.

    Alexander
    Where did you upload it?
    Thanks

    February 4, 2010 — 7:50
  • Mike Almeida

    Mike: I deny the first sentence in your comment, but I affirm its antecedent.
    Here’s a reductio.
    (*) Necessarily: If x freely chooses A in circumstances C, then had x instead been in circumstances C* instead of C such that D(C*,C,x,A), then x would still have freely chosen A.
    1. Assume (*) for reductio.
    2. Let W be a C-world in which x freely chooses ~A in C.
    3. We know that (*) is true in W, since the antecedent of (*) is false. We know there is a world in which the antecedent of (*) is false, since the antecedent is not a necessary truth.
    4. It is true in W that C []-> ~A (centering, 2,3)
    5. It is true in W that C []-> A (from (1), (*)) !@# Contradiction.
    6. (*) is false.

    February 4, 2010 — 9:05
  • The paper is at http://alexanderpruss.com/papers. I’ve also sent it off.
    For the record, Al tells me he denies CDP (a more careful version of (**) in this post).

    February 22, 2010 — 10:25