Friday, February 29, 2008

Incompatibilism and the argument from cosmic justice

A redated post

My argument is going to be an argument from cosmic injustice. No matter how the "freedom" is described, in the final analysis you have to have a certain amount of cosmic luck to have that sort of freedom and to exercise it rightly. (Or maybe good predestination, if you are operating within a theological framework. On my view compatibilism turns the argument from evil into a really good argument). But still, the ultimate responsibility goes up the causal chain to the original source.

I don't know how you could develop a Frankfurt case that would dispel my sense of cosmic injustice that Jeffrey Dahmer is blamed for his bad actions, and Mother Teresa is credited with her good actions, in spite of the fact that Teresa and Jeffrey just happened to be on the end of causal chains that produced their respective actions.

I heard about Calvinism before I heard about compatibilism, but my objections to both are the same.


 James A. Gibson said...

I'm going to set aside the comment about evil and compatibilism - what you really want to refer to here is theological determinism, not compatibilism itself - since even Plantinga thinks the free will defense is shallow as a theodicy (c.f. Plantinga's "Supralapsarian" paper in van Inwagen, 2004). I can't see why what Plantinga says in that paper is not also applicable for a Calvinist to pick up in response to the logical problem of evil.

So the argument, I take it, is this: if S and S' are both determined to do their respective actions by something - God, nature, whatever - out of their control, such that there is some fact or set of facts which entail that S and S' will do actions x and y, respectively, then S and S''s actions are morally equivalent. But S's doing x makes S blameworthy and S''s doing y makes S' praiseworthy. Thus, our differing treatments of S and S' are unjust.

If this is the argument, I wonder why anyone not already committed to incompatibilism would accept the first part: that some set of facts entails S and S''s actions and neither were responsible for these facts together entail S and S' are on a moral par? This looks like an assertion of the position supposed to be proved.

Also, it seems that there are good reasons to think the direct jump from determinism to no moral responsibility can be answered by compatibilists. Again, Fischer has a case in the book I referred to earlier (actually, I think Mark Ravizza wrote that chapter). Assume at t0, a bomb has been placed on top of a mountain on one side of a valley to go off at t3, setting off an avalanche to kill the town below at t4. On the other mountain on the other side of the valley, a sniper has been sent in to trigger an explosive that will go offer at t2. That explosive will not be effective at killing the town until t4. Whether the sniper triggers the explosive at t2 or not, the town will be destroyed at t4. But in fact, at t2 she triggers the explosive. Is she responsible? I say, of course. But also notice, it is inevitable that the town is destroyed at t4 and yet this is something for which she can be responsible. So there is some plausibility to the idea that people can be responsible for events that are inevitable. What matters is the way in which the act is brought about that is salient over whether an agent is responsible or not.

Anonymous said...


I don't know the literature you're referring to, and have no way to access it, but the example you provide of someone being responsible for an inevitable outcome is very poor and won't do the work you are requiring of it.

Firstly, it's a straightforward case of over-determination, and intuitions on over-determination vary.

My own intuition is that is isn't the sniper's fault that the town was destroyed at t4. Having sid that, we can easily separate out the evaluation of the agent and the evalulation of the consequences of the agents actions ... the agent will get a bad evaluation in this case regardless of the fact that the town would have been destroyed in any case. The sniper certainly attempted to destroy the town, and that is all that matters for evaluating the agent.

Since Victor is most concerned with evaluation of agents and their actions, not of the consequences of their actions, you need an example to meet such a case. I don't see one on the cards.


 James A. Gibson said...


I can hardly see what the relevance is regarding varying intuitions about overdetermination cases. Obviously, the case isn't a proof. But I can't see how that makes the case a poor one.

You also object to my presentation of the case because I have presented one for which an agent is responsible for a consequence, not an action itself. Perhaps this is why the case is poor? Again, I can't see why the case needs to do all the the heavy lifting required to move an incompatibilist over for the case to be efficacious in some ways. I reject the desideratum you provide at the end of your post. Here is why: if you grant that the case shows the sniper is at least *partly* responsible for the event at t4, then you've granted that the mere fact that some future event is inevitable does not preclude one's being partly responsible for it. So the mere fact that an event like the sniper's attempting to phi at t2 is inevitable is no reason to think it is something the sniper couldn't be responsible for either. At this point, you'll need to argue the sniper couldn't be responsible because determinism would rule out some necessary condition for responsibility (e.g., sourcehood, PAP, whatever your heart desires). Alternately, argue the sniper is not responsible for the event at t4.

Anonymous said...


I'm convinced that compatibilism is only a problem if punishment or vengeance is involved (e.g. in Christian dogma). It seems to me that you cannot punish Dahmer for his actions if they are determined.

Moral praise or disapproval, however, seems perfectly allowable for the compatibilist.

Imagine a list of all of Dahmer's qualities. On that list are qualities such as sadistic, mean-spirited, self-centered, etc.

Now, imagine Mother Theresa's list (I have some issues with her, but for the sake of argument, they'll be put aside). Among her qualities are gentleness, humility, meekness, etc.

I look at Dahmer's list, and I disapprove. I look at Theresa's, and I approve. I praise Theresa's character and disapprove of Dahmer's.

I'm judging the character each of them possess. It doesn't matter how that character was formed. One receives praise and the other disapproval.

Praise or disapproval of someone's character, then, does not require me to consider how that character was formed.

Only when I seek to punish or seek vengeance should I consider how someone's character was formed. I think your sense of cosmic injustice is valid if one were to be punished for something she did only because she sat at the end of a long causal chain of events. This is a problem for Christians who believe that God will punish or seek revenge on his enemies.

The problem disappears, however, if one thinks only in terms of restraint or correction. One can, rightly, restrain Dahmer from doing any further harm, but to punish would be unjust.

If you did not believe it God was just for punishing, would you still think determinism was problematic for Christianity? In other words, if you didn't have a prior commitment to believe that God is going to punish some in hell for their deeds would you still be looking for a way around compatiblism?

Anonymous said...


Firstly, intuitions about over-determinisism are clearly relevant.

Plausibly, if some event occurs, one cannot be morally responsible for that event unless one causes that event. Since the event is over-determined, on several influential accounts your sniper did not cause the event of the town's destruction. There's the relevance of over-determination.

I explicitly denied that the sniper is responsible for the town's destruction. I think the sniper has done something bad, but the bad thing they did wasn't destroying the town. Destroying the town would of course be bad, but since the sniper didn't do it (wasn't the cause of it), that isn't the bad thing they did.

Your best bet here, James, is to say that the analyses of causation that give this result must therefore be flawed. That's a perfectly legitimate response, but it won't get you very far. It will only get you the result that the sniper is a cause of the town's destruction. All well and good. This would allow that a person can be responible for events where the conditional "if I'd have done otherwise, there'd have been a different outcome" is false.

What you really need is a case of responsibility where the antecedant of the conditional like the above is causally impossible, ie where a person is responsible even though "they couldn't have done otherwise". All you've provided, at best, is a case where someone is responsible even though "the outcome couldn't have been otherwise". These are very different things.


Anonymous said...


At times I've thought along similar lines myself. In summary, the thought is that retributive punishment is out of the question given determinism but that we can still make sense moral approval and disapproval.

Well, I agree about approval and disapproval, but praise and blame cannot be equated with these two. At points you seem to equate them, while elsewhere distinguishing them.

Praise and blame must go if we accept determinism. They are notions of "desert" of what a person "deserves", and must kept or rejected along with retributive theories of punishment.

There are, of course, other theories of punishment along the lines of (i) expression of disapproval, (ii) deterant and (iii) reformation of character.

There are things to be said for and against each of these other theories of punishment. Personally I find them all to be ultimately inadequate unless supplemented by the idea of desert. This applies to punishment here on Earth just as much as to punishment in the hereafter. Incidentally, C.S. Lewis wrote a good paper on punishment, defending the importance of the retributive element.

Still, I think the main reasons for Christians typically endorsing incompatibilist varieties of freedom are as follows, and in this order:

(1) It makes sense of the subjective experience of choice.
(2) Notions of praise and blame require incompatibilist freedom, and we are wedded to these notions.
(3) The Free Will Defence (FWD) relies on such agency being at least possible, and we feel the FWD "gets something right" (by giving room to say that bad stuff cannot necessarily be traced back to God).
(4) The existence of Hell is part of the Orthodox Christian view. Punishment in Hell requires praise and blame and these require incompatiblist freedom.

There are other reasons, but these are the ones which occur to me just now.

Reasons (1) and (2) are perfectly good reasons and are available to people of any religion or none. But as (3) and (4) indicate, there more reasons to adopt such varieties of freedom if we also accept Christianity.


 James A. Gibson said...


OK. This is a better response. I want to think about your criticism, but you'll need to fill it out more. I'll grant you that it is a necessary condition of one's being responsible for an event that one be the cause of that event. I also understand that this is a case of overdetermination. Everyone in the debate realizes that and makes that explicit.

What I can't see is how the sniper is not a cause of the event at t4. What are the analyses of causation ruled out by the case if it is assumed true that the sniper is a cause? I'll give you this: if the case conflicts with several plausible theories of causation, then I believe this is a significant difficulty with the case. I suspect Fischer would make this admission too.

I also need to comment on your last paragraph. When you claim that I need to provide a case in which the antecedant is impossible, such that the agent could not have done otherwise, I think you've changed the debate to different dialectical space. It is important to see that van Inwagen, for instance in his Essay on Free Will, provides two ways in which determinism precludes responsibility. The first is a direct argument from determinism to the lack of responsibility. The second is indirect, from determinism, to that no one could do otherwise, to that no one is therefore responsible. The case I provided is a response to the direct move and *not* the indirect move. So the requirement you're asking for changes the debate to the indirect argument and it is *there* that the Frankfurt cases are relevant. I've understood Victor's challenge as the direct argument.

Anonymous said...


Thanks for your response. In terms of analyses of causation, I am mostly thinking of counter-factual analyses a la David Lewis. There is a paper of his in the Sosa/Tooley collection and elsewhere. The analysis looks roughly like this:

Event c causes event e iff c and e occur and if c hadn't occurred then e wouldn't have occurred either.

This counterfactual David Lewis then interpreted according to his usual metric; so that "If c hadn't occured, then e wouldn't have occurred" is equivalent to "In the possible worlds most similar to our own in which c does not occur, e does not occur either".

On this analysis it will turn out that your sniper did not cause the town's destruction, and that precisely because the destruction would have occurred whether or not the sniper had "sniped".

On the distinction between the direct and indirect refutations of compatibilism, you are probably correct and I may be changing the terms of the debate here. Still compatibilists must answer both the direct and the indirect challenges. The indirect case is supposed to be dealt with by the Frankfurt cases? Could someone give an example ... it's been a while since I read any of this stuff.

But even on the direct challenge I'm unsure your example will help. You describe the direct route as going straight from determinism to lack of responsibility. Now presumably there are intervening premises and inferences here, but they don't go via the "on determinism they couldn't have done otherwise" claim.

There are several ways this direct argument might go. Victor's argument at the beginning of this thread could be taken several ways. The crucial interpretations will, I think, hinge on something like the following principle:

(I) If I am responsible for event e, then I have control over whether or not e occurs.

This principle itself has several interpretations. Your example is, I take it, supposed to be a counter example to this principle.

Does the sniper have control over whether the town is destroyed? We both agree: No, it will be destroyed whatever the sniper's choice. And is the sniper responsible for the event? If so, then this is a counter example to the principle and if Vic's argument is based on that principle it will fail too. But I'm tempted to say no here. I admit that others will find this view puzzling, but I find my intuitions lining up behind the analysis of causation given above. Note, however, that as described in my earlier post, saying the sniper isn't responsible for the destruction of the town doesn't prevent me from finding fault with the sniper's behaviour.

Before I sign off, I must admit to having read very little of Van Inwagen, or of other literature on this issue ... so please excuse me if I'm simply ignorant here.


 James A. Gibson said...


Right. I thought you might raise a counterfactual account of causation. A problem, however, is that the account does not do adequate justice to cases of overdetermination, in which there are multiple causes alleged to be sufficient to bring about an event, but no particular cause of which is necessary. Take, for example, the case in which both Lee Harvey Oswald and John Doe simultaneously shoot JFK from different vantage points and each hits JFK at the same time, each fatally wounding JFK. Either one is the cause and the other is not, both are causes, or neither is the cause. The last option is implausible in that at least one of them brought about the death of JFK. The first option is implausible because what would be the relevant difference in choosing Oswald over John Doe? None the I can see. The second option is the most plausible. Both are causes, each sufficient to bring about JFK's death. So your objection would work against a case like this as well, or any case of overdetermination for that matter. The account of causation for overdeterminative causes will be more complex than the one you provided.

You are right that compatibilists must deal with the indirect and direct arguments. Here, I have only tried to answer the direct argument. Since I don't want to give you a novel in a blog comment, I will just refer you to one example of Frankfurt cases. See Philosophical Perspectives, volume 14. Derk Pereboom has an interesting case there. Or for more on Frankfurt cases generally, see Fischer's article in Ethics, 1999 (I think), or the Wideker / McKenna volume entitled Moral Responsbility and Alternate Possibilities - I might have that backwards.

Back to the direct argument. I think principle (1) is not the one I am objecting to since I think the sniper has a kind of control, but simply not the kind of control that requires alternate possibilities. I referred to this in a different thread as "guidance control." Rather, I am objecting to the following principle:

(1') If I am responsible for e, then I am responsible for e-1...e-n + whatever it is sufficient to bring about e.

I grant you that the sniper does not have control to make false that the town is destroyed, but that fact is irrelevant to the sniper's responsibility. All I have been doing is laying out the argument in Fischer & Ravizza (1998), chapter 7. I would look there too.


Anonymous said...


Yes the counterfactual account of causation has consequences for cases of overdetermination which some people find unpalatable. Cases were the overdetermination is completely semetrical are particularly difficult. But then, other theories of causation also have their problems, but we go for the one we find captures the most important of our intuitions ... for some this turns out to be the counterfactual account. David Lewis is a hard philosopher to ignore in this respect. Still, I think you're right that this is a weak point for the theory ... it's just that if we do accept the theory then we have to accept its less welcome consequences too.

As for the principle you say you're objecting to, namely,

(1') If I am responsible for e, then I am responsible for e-1...e-n + whatever it is sufficient to bring about e.

I have little idea what this is supposed to mean. As it's stated it is of course false. Granting that the sniper is responsible for the destruction of the town, it doesn't follow that he is responsible for whatever else occurs which is sufficient for the destruction of the town. In particular he is not responsible for the explosion of the bomb at t0. But surely Victor's argument doesn't require this.

But perhaps your principle was not intended in this way. Allow me to reformulate it:

(1'') If I am responsible for e, then I am responsible for the existence of some causal chain sufficient to bring about e.

On this principle we can accept that the sniper is responsible for the destruction of the town, but it will also suffice for Victor's argument. On compatibilism agents decisions may be part of a causal chain sufficent for an event, but the agent isn't responsible for the existence of that causal chain, rather that causal chain simply happens to "run through them".


Sturgeon's Lawyer said...

Okay, then: let's focus on the Dahmer/Teresa dichotomy, assume non-freedom/predestination, and see where we get from there. In all cases, insert a per hypothesi wherever needed.

1. Whether or not it "makes sense" from a moral point of view to blame D. and laud T., for us to do so makes sense in the universe of discourse. We do it because we are not free to not do it.

2. If all behavior is determined by a set of sufficient causes, then the response of the environment to an agent's past behaviors becomes a part of the causal chain which determines that agent's future behaviors. This includes the behavior of other agents in the environment, such as reward and punishment.

3. In any case, even if we accept that all behavior is determined by sufficient causal chains, those chains are complex beyond our ability to comprehend. But, we have observational data that suggests that acting as if agents had free will produces results (in the future behavior of those agents) at least as desirable (from our point of view) as any other approach. Thus it makes sense operationally to punish undesirable behavior and reward desirable behavior.

4. In any case, the question is moot. The question of whether we "should" reward or punish behavior is meaningless in a discursive universe where all behavior is causally determined. There are no decisions to make, and so we will do what the history of the universe, as it impinges causally upon us, has determined we will do.

None of this (it seems to me) is an argument in favor of determinism; it most, it is an "apologetic" for determinism, a showing that determinism is not logically unsound, suggesting in particular that the argument from cosmic justice is not so much wrong as irrelevant to the case insofar as it makes assumptions that do not hold in a deterministic universe.

 James A. Gibson said...

Hi Steve.

Yes, David Lewis is a hard philosopher to ignore, but he isn't a hard philosopher to disagree with because some of his views strike many (myself included) as crazy! You may recall that Lewis is associated with a view of compatibilism according to which agents have the causal power to change the past. I am more convinced that the town is destroyed in virtue of there being a prior cause than I am that Lewis' theory of causation is sufficient as you formulated it. And I see no reason why someone couldn't ascribed the causal destruction to the sniper than to the other avalanche on the opposite mountain. So I am placing my bets on the metaphysical possibility of overdetermination (as also brought out in the JFK case) than in the account you provided. The account is too thin to cover all the sorts of cases we want to describe, most relevantly these.

About (1'). You are right to interpret it as (1''). But I should not have wrote (1'). I should have wrote (1'''):

(1''') If I am responsible for e, then I have the power to make it the case that ~e.

That principle precludes inevitability for responsible agents, and it is that principle which is more pertinent to the direct argument. I think it is more like this principle that Victor was invoking rather than (1') or (1'').

I also grant you that compatibilists will have a problem with the passive / active distinction. But there is plenty in the literature to answer that, I think, such that the causes simply don't just "run through the agent." See, for instance, "The Problem of Action" Harry Frankfurt, and work by David Velleman and Michael Bratman.

 James A. Gibson said...


Here is a paper that might be of interest to you:

Jonathan Schaffer, "Overdetermining Causes," Philosophical Studies 114 (2003): 23-45.

I have only given this paper a glance and plan on reading it later this afternoon. His abstract considers three events, the throwing of one rock, the throwing of another, both of which hit the window simultaneously, and the breaking of the window. He says that he argues "that the throwing of each individual rock is a cause of the window shattering, and generally that *individual overdeterminers are causes* [his emphasis]." I suspect there will be much relevant in this paper to our discussion.


Anonymous said...


We're starting to get somewhere here, I think. I'm happy to grant that sniper is responsible for the destruction of the town. (Certainly, I think you can construct other cases where the causes are more symetrical, and I'll pretty much have to grant you this.) It's also true that the non-destruction of the town isn't something in the sniper's power to bring about. However, I still don't think the example will do the work you want it to do. (And I'm also struggling to see how your 1''' keeps the focus on the direct rather than the indirect route.)

It's open to the opponent of compatibilism to adopt the following account:

An agent is responsible for some event e only if either
(i) the agent brings about e but could have brought it about that not-e, or
(ii) the agent is responsible (as under (i)) for some event e' which is causally sufficient for e.

The fact that there is some other event for which the agent is not responsible which is also causally sufficent for e, will not prevent this from being a "libertarian" account of responsibility, as via clause (i) it's still ultimately rooted in the PAP.


 James A. Gibson said...

Hi Steve.

I think that the details here are becoming confusing. I will go back sometime this week or next week and review the material on the various versions of the consequence argument. The dialectical horizon is becoming more dark to me and now I am not sure what arguments to shoot at. I suspect this is why I am probably not being clear enough. If I find anything relevant, I'll either throw something up here or on my blog.

Until then...

Lippard said...

I have to comment on the additional ironies of the examples given:

1. Jeffrey Dahmer was raised in a fundamentalist Christian household, and re-converted while in prison shortly before being beaten to death by another inmate. His father, Victor Dahmer, is an evangelist for young-earth creationism. According to born-agains, Dahmer is in heaven.

2. Mother Teresa, while she raised money to care for the poor in India, was not highly regarded in India and engaged in many actions that are contrary to her image in the United States. For example, her private plane killed six people in Tanzania in 1986, including one of her nuns, which she simply declared was God's will. Christopher Hitchens wrote a book giving the other side of the Mother Teresa story, titled _The Missionary Position_ (and you can see a Hitchens essay online criticizing her as a fraud). Many born-agains believe that, like the Onion has described, Mother Teresa is in hell (but not by accident), since she was a Catholic, not a born-again.

Lippard said...

Correction, Jeffrey Dahmer's father's name was Lionel, not Victor. I hope I wasn't confusing him with the owner of this blog!

Mike Darus said...

I am enjoying your discussion. I noticed another irony in the type of determinism you are describing. There is an asumption that at the end of the day, even though the actors in the scenarios make no real free choice, as the observer, you do. You grant yourself the freedom to determine whether reward and punshishment are indicated. What makes you think your decision is not determined?

Anonymous said...

Hi Mike,

Thanks for sharing your point, you wrote:

“I noticed another irony in the type of determinism you are describing. There is an assumption that at the end of the day, even though the actors in the scenarios make no real free choice, as the observer, you do. You grant yourself the freedom to determine whether reward and punishment are indicated. What makes you think your decision is not determined?”

That’s funny! :-)

Reminds me of the time I was discussing exhaustive determinism with a theological determinist/calvinist at a restaurant. The waiter comes up and the calvinist asks more
questions about what the available options that day were. So the server is answering these questions about the available choices for the calvinist, so I couldn’t resist. I asked the server: do you believe that you have free will? He answers: “of course, I’m explaining the choices this guy [the determinist] has so he can make a choice.” I was laughing so hard, and for some reason the determinist changed the subject and we didn’t talk about determinism any more that day! :-)

Fact is, the determinist can claim determinism all day long, but he grants himself freedom to determine all sorts of things, like what he will CHOOSE to order for lunch . . .