Libertarians must concede that there is always some chance that a moral agent chooses to perform A at t and fails to perform A. But not all chance is bad for libertarianism—that is, not all chance (or luck) undermines control. Here’s a (rough) idea for eliminating the bad sorts of libertarian luck. Duncan Pitchard argues that the kind of luck that undermines knowledge in Gettier cases (and Zagzebski cases) is veritic luck. He proposes a safety condition to distinguish cases of believing p that involve veritic luck from cases of knowing p. His specific condition states that,
V. If a believer knows that p, then in nearly all, if not all, nearby possible worlds in which the believer forms the belief that p in the same way as she does in the actual world, that belief is true (Epistemic Luck, 163).
But a similar condition (L) might well exclude the bad sorts of luck or chance from the libertarian accounts of free action.
L. An agent does not perform A by chance at t, only if, in nearly all, if not all, nearby possible worlds in which the agent chooses to act in the same way as she does in the actual world, the agent does A at t.
(L) is compatible with libertarian free will, since it is possible that the agent chooses to do ~A. It is even possible for the agent to choose to perform A, and fail to perform A. Since it is possible to do so, there is always some chance of it happening. (L) does not commit us to accepting true counterfactuals of freedom, since it does not require that the agent does A in all of the closest worlds. (L) expresses the rough idea of what I’m after—a safety condition for libertarian free action. It seems like one interesting way to diminish the bad sorts luck in libertarian accounts of freedom. Of course libertarians also demand dual ability and dual control, so we need (again, roughly) conditions D1 and D2.
D1. An agent S has dual ability relative to action A at t iff. S has the categorical ability to do A at t and the categorical ability to do ~A at t.
D2. An agent has dual control relative to action A at t iff. S controls the actual choice to do A at t and would control the choice to do ~A were the categorical ability to do ~A at t manifested.
The condition (L) ensures that (D1) is also satisfied. But it does not guarantee that (D2) is satisfied. To satisfy (D2) we might add (L1).
L1. An agent S has dual control relative to action A at t, only if (i) S performs A at t, (ii) S does not perform A by chance at t and (iii) S would not perform ~A by chance at t, were the categorical ability to do ~A at t manifested.
So, (L1) makes it a necessary condtion of control relative to performing A that performing A does not involve the bad sorts of chance or luck precluded in (L). A resolution to the chance objection to libertarianism might take this sort of shape.
I would think what libertarians care most about is the choice itself, rather than the action. Now, the chance objection is best formulated not as an objection that a libertarian-free action is a result of chance, but rather as an objection that a libertarian-free choice is a result of chance.
Moreover, I think the failure of (L) does not undermine responsibility. Fred freely murders Maurice in the actual world by giving him a fairly slow poison at noon, from which Maurice died at 1 pm. Suppose that unbeknownst to Fred, a day earlier Patrick hired an assassin to shoot Maurice at 12:30. The assassin is very reliable, and does her job 99% of the time. However, as it happened, at 12:20 she slipped on a banana peel, and accidentally shot herself. (She survived just long enough to repent of her murderous life.) In 99% of nearby worlds, the assassin succeeds in shooting Maurice at 12:30. Hence, in most nearby worlds, Fred does not murder Maurice.
Clearly, (L) fails. Moreover, it seems somewhat plausible to say that Fred only murders Maurice by chance. But he is still fully responsible for murdering Maurice.
"Libertarians must concede that there is always some chance that a moral agent chooses to perform A at t and fails to perform A." -- This isn't true in the case of God, so let's restrict quantifiers to finite agents. But with that restriction, don't compatibilists have to concede the same thing, assuming we live in a chancy world?
I would think what libertarians care most about is the choice itself, rather than the action.
They care about both. (L) can be generalized in the relevant way.
In 99% of nearby worlds, the assassin succeeds in shooting Maurice at 12:30. Hence, in most nearby worlds, Fred does not murder Maurice.
That doesn't follow. Whether she kills him in any nearby world depends on how similar those worlds are (to ours) in which she succeeds in killing him. The probabilities do not tell us how close those worlds are. She might be successful 99% of the time and it might still be the case that, in order to succeed this time, it would take a large or gratuitous change. Just as there would be no minor changes in the location of air molecules in Guatemala (these are simply gratuitous under the counterfactual hypothesis) there would be no changes in the location of the banana peel. All of the most similar worlds would have the well-placed peel and all of the most similar worlds would have the molecules in Guatemala in their actual location.
Let's modify the banana peel hypothesis, then. Let's suppose that the banana peel has to be placed precisely where it is for the assassin to slip on it. She is a very careful assassin. She watches where she walks. But there is one precise position where a combination of shadows makes her overlook it.
I am not sure how to modify (L) to work in the case of choices.
Let's modify the banana peel hypothesis, then. Let's suppose that the banana peel has to be placed precisely where it is for the assassin to slip on it. . .
Alex,
How does that help? It will still be a gratuitous change to have the peel in some other position, given the counterfactual hypothesis. In all of the most similar worlds in which Fred decides to murder, it will have the improbable position is does have.
I am now confused what you count as nearby worlds and what the "counterfactual hypothesis" is.
Right, it's not technically a counterfactual hypothesis. The agent does the same thing in the nearby (possible, non-actual) worlds we are considering as he actually does. If he chooses to do A, you ask whether there are any worlds among the nearest in which he chooses to do A--just as he does in the actual world--and fails to do A. If there are such worlds, then he action is chancy. If not, not. Maybe it is best (though I hadn't thought of it) to think of the actual world as non-centered, so that there are other worlds in which the agent chooses to do A that are as close to our world as our world is to itself. Look at those worlds to see whether there are any in which he fails to do A.
I thought the point of this was to say why agents freely (or intentionally) do things they are not in complete control of. E.g. Austin points out that if you make a golf putt, we say you did it freely or intentionally, even if there was a chance you would miss; if you make a hole-in-one off the drive, we are somewhat less likely to say this. I think the safety condition captures this idea nicely: if you tried making the golf putt the same way, most of the time you would still make it; if you tried making the hole-in-one the same way, most of the time you would fail.
How to apply this to choices? Maybe the idea is that the process of deliberation leading up to the choice is the "trying in the same way": your choice is free or intentional iff (if you reasoned the same way, most of the time you would make the same choice). Interesting.
And no I don't think these observations are confined to libertarians; compatibilists could agree with them too.
I thought the point of this was to say why agents freely (or intentionally) do things they are not in complete control of. E.g. Austin points out that if you make a golf putt, we say you did it freely or intentionally, even if there was a chance you would miss;
I think that's part of the idea. But there's also the persistant charge against libertarianism that it makes our actions/choices in general too chancy for genuine freedom. How do we distinguish cases in which there is too much chance from cases in which there isn't? I think the safety condition in (L) gets us close to what we want.
Dr. Almeida,
I have been thinking about your argument and wondered if you could clarify something for me to aid me in my reflection?
"L. An agent does not perform A by chance at t, only if, in nearly all, if not all, nearby possible worlds in which the agent chooses to act in the same way as she does in the actual world, the agent does A at t."
My first question is: Don't all these possible worlds have to have the exact same past up to t?
My first question is: Don't all these possible worlds have to have the exact same past up to t?
Yes, I think that's right. What libertarians are worried about is a world being the same up until t (same deliberations, same choices, same decisions, etc.) and the agent doing A in some close worlds and ~A in other close worlds. We want it to be possible for the history until t to be the same and the agent perform ~A, we just don't want that to be true in any of the closest worlds to ours. In short, it would take some strange event occurring at t--perhaps simultaneous with the performance of the action--for the agent to perform ~A at t.