Thursday, July 10, 2014

Kant and Lewis on our freedom

Kant (on one reading) holds that the initial conditions of the universe and the laws of nature depend on us (noumenally speaking). This reconciles determinism with freedom: sure, our actions are determined by the laws and initial conditions, but the laws and initial conditions are up to us. Kant also thinks that a further merit of this view is that one can blame people whose misdeeds come from a bad upbringing, because noumenally speaking they were responsible for their own upbringing.

Lewis holds that freedom is compatible with determinism, and in a deterministic world had one acted otherwise, the laws would have been different.

Everybody agrees that the view I ascribe to Kant is crazy (though not everybody agrees that the ascription is correct). But Lewis's view is supposed to be much saner than Kant's.

How? The obvious suggestion is that Lewis only makes the laws depend counterfactually on our actions (assuming determinism) while Kant makes the laws depend explanatorily on our actions. But that suggestion doesn't work, since Lewis's best-systems account of laws makes the laws depend on the law-governed events, and so it makes the laws depend not just counterfactually on our actions but also explanatorily: the laws' being as they are is grounded in part in our actions. So both accounts make the laws explanatorily depend on us.

Admittedly, Kant also makes the past, not just the laws, depend on our actions. But that's also true for Lewis, albeit to a smaller degree, because of his doctrine of small miracles...

16 comments:

Miloš said...

Peter van Inwagen glose over Lewis's view in great detail in one of his latest papers on free will: http://andrewmbailey.com/pvi/Freedom_to_Break_the_Laws.pdf

Mike Almeida said...

Lewis holds that freedom is compatible with determinism, and in a deterministic world had one acted otherwise, the laws would have been different. Lots of incompatibilists deny that the past depends counterfactually on the present in the way Lewis describes. But surely we cannot deny that the future depends counterfactually on the present. But incompatibilists have to deny that as well. Since, if the future depended counterfactually on the present, and determinism were true, then the past would also depend counterfactually on the present. But entailing that the future does not depend counterfactually on the present is a reductio of the view. Hence the falsity of incompatibilism.

Alexander R Pruss said...

Mike:

That's a neat argument! I haven't seen it before.

However, I don't think the incompatibilist needs to deny that the past depends counterfactually on the present in deterministic worlds. She needs only hold that the past doesn't counterfactually depend on actions *within our power*.

For instance, certainly the past depends counterfactually on actions not within our power, like the action of backwards causing some event.

Mike Almeida said...

Alex,

Interesting objection. Speaking generally, the incompatibilist will hold that it is not within my power to alter the past. If P = fact f held before I was born and L are the laws then, I *think*, most incompatibilists would say that all of the counterfactuals are false which state that, for some A, were I to do A, P would be false, and also false that were I to do A, then L would be false. All of the worlds in which I do anything are worlds that share a past and laws with ours. Were I to do A, something impossible would happen. That's why I can't perform A.

Brian Cutter said...

We could avoid this consequence of Lewis's view without deep changes to the Lewisian picture simply by denying Lewis's doctrine that laws must be *true* propositions. For example, we should allow that Newton's law of gravitation might be a law in a world in which it has some exceptions. Imagine a world in which it has a zillion true instances, but one exception, when a monk levitates immediately after praying fervently that he would do so. It seems we should say that Newton's law of gravitation is a law in that world, even though it is false. Likewise, suppose it is a law that momentum is conserved, and that if I had done otherwise, momentum wouldn't have been conserved. We should say, contra Lewis, that if I had done otherwise, it still would have been a law that momentum is conserved. On this picture, the laws---that is, which propositions are laws---don't counterfactually depend on my actions (though whether the laws are true may).

I don't think this view of laws would amount to a deep revision of the Lewisian picture, in part because it needn't be opposed to the spirit of Humeanism about laws. The basic idea of Humeanism is that the laws are certain simple, informative generalizations. And a generalization can be (simple and) very informative even if it's false (so long as it's nearly true, i.e. doesn't have too many counterexamples). In the first world described above, one would get along very well making predictions on the basis of Newton's law of gravitation.

Alexander R Pruss said...

Brian:

That's a neat move. But if you define laws in this way, then many a libertarian will agree that compatibilism is true, and that's too easy.

Suppose that I live in a world where lots of stuff happens and everything but one action of mine is causally determined, and that one action is free due to a miraculous power I have. Most libertarians will agree such a world is possible. But on this looser account of laws, if we tell the story the right way, this world will have deterministic laws--the exception due to the one action will be irrelevant to what the laws are. Thus, this will be a world with deterministic laws and freedom.

But this is too easy a victory for the compatibilist.

What the compatibilist wants to say is that it is possible to be free in a deterministic world without anything spooky like miraculous powers to act contrary to the laws.

It's actually pretty hard to formulate compatibilism that makes all this clear.

Anonymous said...

Dr. Almeida,

Would you mind expanding upon the following quote:

"Since, if the future depended counterfactually on the present, and determinism were true, then the past would also depend counterfactually on the present."?

Brian Cutter said...

"Suppose that I live in a world where lots of stuff happens and everything but one action of mine is causally determined, and that one action is free due to a miraculous power I have. Most libertarians will agree such a world is possible. But on this looser account of laws, if we tell the story the right way, this world will have deterministic laws--the exception due to the one action will be irrelevant to what the laws are. Thus, this will be a world with deterministic laws and freedom. But this is too easy a victory for the compatibilist."

I agree that this is too easy a victory for the compatibilist, but I don't think the victory is made easy by dropping the requirement that laws be true propositions. For we can get cheap victories of this sort while maintaining this requirement. Just take the scenario you've described, with the added stipulation that you don't exercise the relevant miraculous power. I.e. you have a power such that, were you to exercise it, some of the actual deterministic laws would be false (whether or not they would still be laws), but in fact you don't exercise it, and so the deterministic laws are true. Then we still get an easy victory for the compatibilist, for the laws will be true in this scenario. (I assume that the mere *having* of the "miraculous power" wouldn't amount to the breaking of any law, so long as it is not exercised, though this could be disputed. I suppose this depends on whether the laws comment on more than the distribution of categorical features.)

But, then, one wants to say: isn't this cheap victory precisely the victory Lewis has fought to win? After all, suppose that in fact determinism is true (and the deterministic laws are true). Lewis would say that we nonetheless have powers or abilities such that, were we to exercise them, some of the actual deterministic laws would be false. But then it's hard to see how this is different from unexercised "miraculous powers."

I suppose this is all to agree with your claim that "It's actually pretty hard to formulate compatibilism that makes all this clear."

Heath White said...

Mike,

That is an interesting argument and I wish it worked but I can’t see that it does. Maybe I am missing something.

F = “the future depends counterfactually on the present”
D = “determinism is true”
P = “the past depends counterfactually on the present”

Your key claim is “Since, if the future depended counterfactually on the present, and determinism were true, then the past would also depend counterfactually on the present” and I would render this as (F and D) -> P. Incompatibilists wish to deny both D and P.

But from
1. (F and D) -> P
2. ~P
We get
3. ~(F and D)
4. ~F or ~D
And adding
5. ~D
Is compatible with any truth value for F.

Am I missing something?

Mike Almeida said...

Heath and Bret,

The argument would go this way. Take any counterfactual that is intuitively true and according to which the future depends on the present: e.g., if I were to press the button, the door bell would ring, or P > R. Now suppose we are in a deterministic world and the incompatibilists are right: the past is settled and does not depend on the present. If that were true, then P > R would be false as well. If I were to press the button, then the door bell would ring, R. But the occurrence of R in the future would entail that the past was different, since, assuming determinism, the laws of nature L and the past history H entail ~R. So, incompatibilism is true only if the future is not counterfactually dependent on the present. But that's as clearly false as anything is. So, incompatibilism is false.

Alexander R Pruss said...

Mike:

But there are true counterfactuals about what would happen if something impossible happened. For instance, "If a man were an ass, he'd have four legs", as Aquinas says. That there are such true counterfactuals is uncontroversial. (What is controversial is the Lewisian thesis that all counterfactuals with impossible antecedents are true.)

Among the counterfactuals with impossible (at least nomically impossible) antecedents will be backtracking counterfactuals such as "If I backwards caused there to be a pink striped dinosaur, there would have been a pink striped dinosaur."

Likewise: "If I were a dolphin, I would have been born to a dolphin."

So the incompatibilist thinks that in the deterministic world, all counterfactuals are counterfactuals with impossible antecedents. And at least some of these counterfactuals are true.

Alexander R Pruss said...

In the last sentence, "impossible" is used loosely ("impossible given the laws and the past").

Mike Almeida said...

I'm not prepared to say its uncontroversial that there are non-trivially true counterpossibles. I'd like to see a coherent and intuitive semantics for them. But I think it's mistaken to hold that incompatibilists allow for true backtracking counterpossibles. The reason backtrackers are false for incompatibilists is because the consequent of those conditionals is impossible. The past is in some sense necessary, on this view, and there is a modal asymmetry on this score with the future. But then let P be the past. We know that, for relevant N, NP. But all backtrackers state that, for some present action A, A > ~P. But we know that NP. So, we know it's true that A > P. Therefore, it is false that A > ~P, for all backtracking conditionals.

Mike Almeida said...

There is another problem for the incompatibilist argument that is worth noting. On that argument, the backtracker A > ~B is false because NB: B is accidentally necessary. But it is only contingently true that NB; it is true in our world, but certainly not necessarily true. So, if it were the case that I did A, it would not be true that a past necessity, NB, would be violated. It would be the case that NB is not true. I would have actualized a world--there are no doubt many, if the past is not metaphysically necessary--in which the past includes ~B. So, even if we concede that the past is inviolable--even necessarily inviolable--we can't reach the conclusion that backtrackers are all false. Backtrackers are false only if the past is metaphysically necessary. But of course that's false.

Alexander R Pruss said...

Mike:

I didn't say that it's uncontroversial that there are nontrivially true counterpossibles, only that it's uncontroversial that there are true counterpossibles. :-)

I still don't see why the incompatibilist should deny true backtracking counterfactuals. Rather, the incompatibilist should deny that, in ordinary cases (putting aside cases of miracles, time travel, etc.), there are true backtracking counterfactuals *with antecedents within our power*.

That there is a true backtracking counterfactual with antecedent A and consequent B is strong evidence that A is not within our power. For instance, were I to have my third birthday tomorrow, I would have been born about three years ago. This backtracking counterfactual is evidence that it's beyond my power to make me have my third birthday tomorrow.

Mike Almeida said...

I still don't see why the incompatibilist should deny true backtracking counterfactuals. Rather, the incompatibilist should deny that, in ordinary cases (putting aside cases of miracles, time travel, etc.), there are true backtracking counterfactuals *with antecedents within our power*.

Suppose A > ~P is a backtracker. What explains why A is not within my power? It is certainly nothing intrinsic to A, A might be flipping a light switch, waving my arm, etc., All things we'd expect I can easily do. What explains why I cannot do A is that it would entail that the past P is different. But this cannot happen (we're told), since we don't have even 'counterfactual power' (whatever that completely misleading phrase is supposed to mean) over P. Well, why is that? Because NP is true, for some notion of N. But if NP is true, then 'A > ~P' is false and 'A > P' is true. This is because (i) is true.

(i) NP --> (A > P)

On the other hand, if it were true that A > ~P, then we'd have no reason, based on the past, to think that I could not do A.