Neil Levy

Neil Levy (University of Melbourne), “Why Frankfurt Style Cases Don’t Help (Much),” with commentary by Kevin Timpe (University of San Diego).  Both the paper and the commentary can be found here.

Comments

[Jussi Suikkanen]

Neil,

I really like the paper. Very interesting and clear. I wonder though if there already are certain other *Frankfurt* cases or FSC2s that put pressure against your PAP*.

So PAP* states that:
'An agent is responsible for his or her choice only if he or she possessed the the power-sense ability to do otherwise at t'.

The agent then has the relevant power-sense ability to do otherwise when she has the right powers, the requisite talents and knowledge, to do something else.

Now, one type of cases Frankfurt discusses where I'm not sure the agent would have such abilities are the Luther type of cases. In these cases the agents have volitional necessities that drive them towards one choice in a situation. Being the agents they are, it is inconceivable that they willingly do anything else - if they did, it wouldn't be them who were acting. I think there is an important difference here to the willing addict cases. They have the power-abilities to do otherwise because their drug-volitions are not their essential features. I wonder if in these cases the agents then lack the right powers to do anything else. Luthers can no other. Yet the intuition is supposed to be that these agents are morally responsible. This is because the volitional necessity is something the agent identifies with, to have these volitions is what she stands for, what it is to be her.

So, anyway, here you might have a Frankfurt case where the agent lacks even the power abilities but still the intuition is that she is morally responsible. This would imply that PAP* is false too.

[Neil]

Jussi,

I'm about to get on a plane, so no time for a considered response. But there's always time for an unconsidered one - here it is. A tracking back condition can handle volitional necessities, just as Kane suggests. That is, if the agent is responsible in these kinds of cases (and my view is that agents are rarely responsible) it is because that PAP* was satisfied at some appropriate time when the agent formed her character. There are, as Manuel points out, epistemic conditions which entail that tracking back conditions rarely do the work that philosophers ask of them. That's fine by me: I agree that tracking back conditions rarely will do the requisite work. When, and only when, agents are responsible, even though experience genuine volitional necessity, the condition might still explain why.

[Kevin Timpe]

For what it's worth, I agree with Neil about the way that his account, like others, can handle such cases.

[Terrance Tomkow]

Levy says that FSC’s sometimes show something interesting about responsibility and possibility.

To see if this is so, let us remind ourselves that for an FSC to convince us of anything there must be situations meeting the following requirements. There is someone, Jones, who does not do X and we *all agree*:

    1. Jones is morally responsible for not doing X.

And where we also agree that:

    2. Jones could have done X.

Now, to turn this into an FSC we add a "Black" whose powers are such that we are convinced that, with this addition, 2 is false but 1 remains true. (If some philosophical doctrine you hold prevents you from ever simultaneously endorsing 1 and 2, then Frankfurt stories have nothing to tell you.)

Now, Levy says that there is a sense of "can" in which FSC's work and a sense of  “can" in which they don't. He says FSC’s fail if we are using the "power sense" of "can" but they (may) succeed if we use the "occurent" sense of can. He never defines these senses but he offers us examples. Can Jones play the piano right now? Yes, he's been playing masterfully for years and has no physical or mental impairments that would keep him from playing. This is the "power sense" of  “can”. But there is no piano around right now. So, can he play the piano *right now*? No! This is the "occurrent" sense of  “can”.

Levy says, in effect, that the reason you can't tell FSC stories in the power sense is that to make (2) false you are going to have to mess with Jones’ intrinsic properties in ways that will also make (1) false. E.g. You can rob Jones of the power to play the piano by cutting off his hands, but we don't hold people without hands responsible for failing to play the piano.

I think that's right. FSC’s fail for this sense of "can".

But now where are the stories in which FSC’s are supposed to succeed? Stories told with the occurrent sense of  “can’? Let's try the piano playing case. We can make (2) false right now, in the occurrent sense, by making sure that there are no pianos around right now. But of course, if there are no piano's around right now we wouldn't hold Jones responsible for not playing the piano *right now*. So this doesn't work as an FSC: (2) is false, but so is (1) for the relevant X.

Well, we could have a piano nearby but have it rigged with explosives that would go off the moment someone began to play; or we could put in a mechanism that would wreck the piano if Jones approached it; or we could put something in Jones brain that would blow up the piano if he decided to play it, or maybe that would blow it up if he was even about to begin to decide. Clearly there are lots and lots of ways of making it false that Jones can play the piano right now and you will find many permutations in the low-rent science fiction genre that calls itself "the Frankfurt literature".

But notice something. No matter how you have gimmicked things up, however you make it false that Jones can (occurrently) play the piano you also make it false that Jones is responsible for not playing the piano. We don't hold Jones responsible for not playing the piano right now if there is no piano for him to play, right now, or if the only piano is broken, or gimmicked to break as soon as he tries to play it or... Whenever you make (2) false in the occurrent sense, you also make (1) false. (Of course you might hold him responsible for not even *trying* --- but that is a different X.)

And so it is, with Levy's (only) other example. Black stands by ready to intervene if the micro-chip in Jones’ brain reveals that Jones is even beginning to be " in the process of deciding to vote for Hilary Clinton". In that event Black will intervene and keep Jones from voting for Hilary. That would certainly rob Jones of the occurent ability to vote for Hillary. But in this case, whatever we might hold Jones responsible for, it surely *wouldn't* be his not voting for Hillary. After all, Jones COULD NOT HAVE VOTED FOR HILARY because if he had even *begun* the process of deciding to vote for her, Black would have prevented him from voting for Hilary. Of course you might want to hold him responsible for not even *beginning* to decide to vote for her. But note that in the story as told, nothing robs Black with the occurrent ability to do *that*.

So Frankfurt stories don't work-- i.e.. don't show us anything interesting about morality and "can" -- for the power *or* the  occurrent sense of "can".

This is because Frankfurt stories don't work, period. See Kadri Vihvelin’s remarks of last week for an explanation of why this so .

[Neil]

Terrance,

I'm content to limit the scope of the paper, so that it is not intended to show you anything interesting if you don't think that FSEs work *at all*. That leaves me a large and respectable audience. So I don't think I need to reply to your line of argument. But here's a few thoughts, anyway.

You write:

But in this case, whatever we might hold Jones responsible for, it surely *wouldn't* be his not voting for Hillary. After all, Jones COULD NOT HAVE VOTED FOR HILARY because if he had even *begun* the process of deciding to vote for her, Black would have prevented him from voting for Hilary.

End quote (Can't get comments to accept html tags).

Now, surely that just begs the question. You say we can't hold Jones responsible because he lacks alternative possibilities. But the point is to show that he can be responsible despite lacking alternative possibilities.

Second, you write that nothing in the story prevents Jones from *beginning* to vote for Hilary. Well, that might be right. I was presenting a skeletal FSE, exepcting my audience to trick it up to block all the obvious objections. In particular, I expected them to fill in the details so that there can only be a flicker of freedom: a non-robust alternative. Here's one, adapted from Pereboom: it is a necessary, but not sufficient, condition of Jones deciding (or beginning to decide) to vote for Hilary that Jones thinks about the US health system. If and only if Jones he considers the health system might he go on to seriously consider voting for him (because of quirks about Jones' psychology). But thinking about the US health system - that is, having the thought cross his mind - is not something that Jones does volunarily. Nor, it seems, is something in virtue of which we can hold Jones responsible. Having this thought is Black's trigger for intervention: though he knows that having the thought will not cause Jones to decide to vote for Hilary (since, I now stipulate, indeterminism holds), he also knows that Jones will definitely not decide to vote for Hilary in the absence of thought. Jones did not possess APs, in one sense; Black ensures that. But Black plays no role in the actual sequence at all, and therefore seems irrelevant to Jone's responsibility.

Finally, the agrument with Kadri should be carried on elsewhere. Two quick points. One, her defence of agents' abilities seems very close to mine (she, too, seems to think that Black is a fink). But as I said in the paper, that only tells us about agent's abilities, not about the possible worlds to which they actually have access. Second (since Kadri in her response) puts a lot of weight on the backtracking conditional response, she owes us a response to John's claims that these conditionals are fine under certain conditions.

[Jussi Suikkanen]

Neil,

thanks for the reply and hope you had a good flight. I'm very scared of flying and for this reason this conference is quite ideal for me. Anyway about the reply. It might be, as you hint, that responsibility for an action a at t in the Luther cases might require a power-ability to do other otherwise (i.e. to form one's character) at an earlier time t-x. However, if this is admitted, then it already does the job against PAP* as it now reads. If our intuition about these cases is right, then, contra PAP*, no power abilities are required at t like the principle claims. I'm sure there is a way to tweak it.

[Jussi Suikkanen]

Also, come to think of it, if we go this far, then it might be that the compatibilists begin to win ground about the intuitions. At least I don't have clear intuitions that anything like PAP has to hold for the prior volitional necessities forming actions (or processes more likely). It is likely that those volitional necessities must have a right sort of history - not being a result of manipulation and the like (like I remember Fischer, Ravizza, Christman and Mele claim). But, at least the incompatibilist needs to start working at this point to show that something further than that is required.

[Terrance Tomkow]

Apparently even if it were revealed that, thanks to an evil scheme involving Karl Rove, microchips and the NSA, the majority of Americans were *unable* to vote for Gore in 2000, Levy would nevertheless hold the majority responsible for reelecting Bush. This seems odd. But never mind.

The Pereboom story Levy cites is a classic Frankfurt story insofar as it is vague in all the crucial details. Let's look at it closely.

In the story, Jones  actually does three things.

     (1) does not think of the US health system.
     (2)  does not even begin to decide to vote for for Hillary.
     (3)  does not vote for Hillary.

The alternate possibilities here-- the things Jones does not do-- are

     (1') Think of the US health system
     (2') begin to decide to vote for Hillary.
     (3') Vote for Hillary.

Now if this story is to work, Jones must be responsible for doing at least one of (1-3)  even though, thanks to Black, Jones can't do the corresponding member of (1'-3').  Set aside questions of responsibility; Levy will always accuse us of begging these. Let us, instead, focus on the question of alternate possibilities. Which alternative (1'-3') does Black's presence preclude?

We know it isn't (1') since Black does nothing to prevent (1') and does not cause (1). Black just stands ready to act in case (1') happens.

Does Black's presence mean that Jones cannot, in some sense of "cannot", do (2'). Well, one might argue like this:  In the story,  (1') is, a causally necessary condition for (2'). Moreover Jones has no voluntary control over whether or not (1') happens. Given that  (1') didn't  occur and Jones couldn't make it occur,  it follows that, in some sense, Jones *could not*  (2')? 

This is an interesting argument, but it is wholly beside the present point since you will note that in whatever sense these nomological facts make it impossible for Jones to (2'), Black has nothing to do with it.  He didn't design Jones' psychology and did not interfere with the initial conditions (1).

So is there some other sense of "can" or "could" in which Jones could do (2') if *only* Black were not on the scene?

Levy invokes two senses of "can": the "power" and the "occurent" sense. I take it, that Levy would agree that given that Black has not messed with Jones' intrinsic properties that Jones has the same power/ability to do (2') that he would have if Black weren't on the scene. Once again Black is irrelevant.

That leaves Levy's other sense of can. The "occurrent" sense.  Can Jones begin to decide to vote for Hillary *right now*? Well if "now" is any time after (1) the answer here seems to be "No".  Remember the story is that Jones can begin to decide to vote for Hillary only if he has had a prior thought about the health care system. But as of *right now*, Jones has not had that thought so he can't (2') right now.

The problem with this line of reasoning, as before, is that, even if it works, there is no work for Black to do in it. It's not *Black* depriving Jones of the ability to (2') right now. Its the prior non-occurrence  of (1') that renders him unable to (2') right now, and, in the story Black had nothing to do with the non-occurrence of (1')

So how about (3')? Could Jones have voted for Hilary? Well the story seems to tell us that if he began to vote for Hilary he would have voted for her since in *this* story (unlike Levy's first version) Black's intervention would be triggered by (1') not (2'). So it seems that if, right now, Jones did (2') then (3') would ensue and, for all we have seen so far, Black's presence made no difference to whether or not Jones can (2').

Of course one might wonder, when voting time comes, whether or not Jones can *right now* do (3'). After all, given (2) and given the workings of Jones' psychology it is nomologically impossible, *right now* for Jones to vote for Hilary. Maybe so, but, once again, this impossibility has nothing to do with Black.

Maybe the problem is I don't understand Levy's "occurent" sense of Can. Levy owes us a definition, particularly since a definition should be easy to give.  As David Lewis taught us, 'can" always means "compossible with certain facts F".  You get different "senses of can" by giving different specifications of the facts F.  Thus you might say that Jones can X in the *power* sense iff his doing x is compossible with all the facts about Jones intrinsic constitution.  This has its problems but will do as a first approximation. So how how shall we define the occurent sense of can? We can't say, "Jones can occurently do x right now" iff his his doing x is compossible with all the facts right now"  since among the facts right now is the fact that Jones does not do x. 

So what is the occurent "can"?  And what is it that Jones can't do because of Black? And where is the argument that Jones can't do it, whatever it is?

But look where we are!  Over thirty years after Frankfurt's paper we still haven't got *one* clear example of a Frankfurt story that works. Always there seem to be "details" that need to be filled out; a little more flesh that has to be added to the "skeletal" story no matter how much it is larded out.

Give it up.

[Fritz]

Like some others, I have argued both in print and in forthcoming work that Frankfurt stories aren't helpful for establishing compatibilism about moral responsibility and causal determinism. I have argued that they are even less helpful for purposes of arguing for compatibilism about freedom and causal determinism. But even though I stand by my arguments for these claims and think that some others who have defended similar claims in different ways (e.g., Ginet, Widerker) have also ably defended the claims, I do not think it's beyond the pale for philosophers who find something valuable in Frankfurt's work and extensions of his work to continue exploring the examples and the structure of different variants of the examples and the possible applicability of these examples to conclusions about moral responsibility and other notions. Reasonable minds apparently disagree about the force of the examples. Notice, for example, that one needn't think *at all* that compatibilist conclusions follow from reflection on the Frankfurt examples in order to think that the examples help us see something interesting about responsibility...

[John Fischer has occasionally accused me of being a bit angry and inflexible with my anti-Frankfurt argumentation... But perhaps he'll see I've mellowed a bit. or at least that in comparison with Mr. Tomkow I seem mellow about these points!]

The position that I hold (and Vihvelin holds, and a few others hold) is, as it happens, a *minority* position about the Frankfurt examples. Of course I believe it should be the majority position. But it's presumably our job to more carefully explain and defend our position and try to convince others that the examples don't work. Telling those who disagree to "give it up" and referencing and parrotting work almost all of them are already familiar with is unlikely to lead to progress. Doing so in a context where the contribution to the discussion is mostly off topic seems especially unlikely to lead to progress.

So Neil, I like your initial response: you have a wide audience for your discussion -- it's all of those who (mistakenly if a few of us are right but that's a *separate* discussion) accept the standard moral of the Frankfurt literature. Seems quite sensible of you to address that rather large crowd.

[Neil]

In reverse order...

Thank you for the equivocal encouragement, Fritz! I'm sure that something of value has come out of FSEs, even if it nothing more than revitalising a debate that seemed a little moribund, and sharpening philosophical reflexes. If, though, we have reached a dialectical stalemate, then FSEs have, in one sense, failed: they haven't broken the impasse in the way that Frankfurt-style compatibilists hoped.

Maybe I can say something to Terrance that, if it won't actually change anyone's mind, might focus the agreement. Terrance writes:

The alternate possibilities here-- the things Jones does not do-- are
(1') Think of the US health system
(2') begin to decide to vote for Hillary.
(3') Vote for Hillary.
Now if this story is to work, Jones must be responsible for doing at least one of (1-3) even though, thanks to Black, Jones can't do the corresponding member of (1'-3'). Set aside questions of responsibility; Levy will always accuse us of begging these. Let us, instead, focus on the question of alternate possibilities. Which alternative (1'-3') does Black's presence preclude?
We know it isn't (1') since Black does nothing to prevent (1') and does not cause (1). Black just stands ready to act in case (1') happens.
Does Black's presence mean that Jones cannot, in some sense of "cannot", do (2'). Well, one might argue like this: In the story, (1') is, a causally necessary condition for (2'). Moreover Jones has no voluntary control over whether or not (1') happens. Given that (1') didn't occur and Jones couldn't make it occur, it follows that, in some sense, Jones *could not* (2')?
This is an interesting argument, but it is wholly beside the present point since you will note that in whatever sense these nomological facts make it impossible for Jones to (2'), Black has nothing to do with it. He didn't design Jones' psychology and did not interfere with the initial conditions (1).

(end quote)

So: Jones is responsible for his voting just in case there was something concerning which he exercised dual control and the failure to actualise which makes him responsible for his failure (Is this right?). Call this (just to have a name) a responsibility-grounding event.

If that's the case, then Terrance is right, the counterexample can't get off the ground. Why? Because if the signal for Black's intervention is itself a responsibility-grounding event, then Jones is indeed responsible, but because his failure to actualise it is something over which he exercised dual control, then indeed he had alternative possibilities. But if, on the other hand, there is no responsibility-grounding event, if the signal for intervention is not something over which Jones exercised dual control, then he is not responsible. A dilemma (am I still on track?) Terrance's point is, therefore, that as a matter of fact the case fails because with or without Black Jones is not responsible: the contingent feature of his psychology that is the signal for intervention - his thought of the US health system - is not something over which he exercises dual (or indeed any) control. Given that that's the case, Jones isn't responsible at all. Black plays no role.

I can see why someone would want to say something like that. But it seems to me a big bullet to bite. If we take this line, we are committed to saying that an agent is responsible for an action just in case he exercised control (which here I deliberately leave vague - plug in your favorite account) over *all* the psychological antecedents of that action that are necessary conditions of it. For there doesn't seem to be anything special about Jones or his thought. But - unless we have a very strange view of the mind - we shall have to admit that whatever the nature of the universe - theologically, nomologically, or what have you - none of us ever exercises such control. All our conscious thoughts are the tip of a cogntive iceberg; we exercise control only over the small part that enters the global workspace, and even our exercise of control is itself only made possible by the working of subpersonal processes which we cannot, in principle, control. So if that's the requirement then determinism is irrelevant: none of us is responsible.

As I said, I can see why you might want to tell the story about the responsibility-grounding event. I can even see why you might want to bite the bullet (for what it's worth, I have days on which I am a no-free-will-either-way guy). But it's a big bullet to bite.

[Neil]

Haven't forgotten you, Jussi. Just decided to press post before I lost a long comment - again.

I'm not sure I've got the point. The claim isn't that the agent needs a power-sense ability at t; the agent retains that ability so long as the agent retains whatever bundle of dispositions are the basis of that ability (including when he is asleep). The claim is that agent requires the occurrent-sense ability at the time (which she fails to have in genuine cases of volitional necessity - if there are any).

Does that help? If not, you'll need to tell me more.

[Jussi Suikkanen]

Neil,

thanks. Now I've lost it completely. I thought the crux of the paper was that FSCs show that PAP is false but PAP* might still be true and unaffected by FSCs. PAP* states that:

'An agent is morally responsible for his or her choice at t only if he or she possessed the power sense ability to do otherwise at t'.

Now the point of the Luther cases is that one can be morally responsible for a choice at t without possessing the power sense ability to do otherwise at t. There are no possible worlds in which Luther does something else at the moment he stands and can no other. But, now you say that 'the claim isn't that the agent needs a power sense ability at t'. Looking at PAP* I thought it was. I'm probably missing something.

[Neil]

Someone's confused. It might well be me.

The Luther style case is a counterexample to PAP, understood as a claim about occurrent-sense abilities. I, like others, want to reply to this kind of case by introducing the backtracking condition (borrowed here from Kane). I think this kind of reply matters because PAP still plays some kind of role in responsibility attributions, so the availability of the standard reply matters here.

Now I take it that you want a Luther-style counterexample not to PAP, but to PAP*. I'm not sure that there is such a case. The fact that the agent cannot bring herself to phi at a time does not show that she lacks the power-sense ability to phi. What would? I may still be missing the point; I think I need a case to get my teeth into.

[Jussi Suikkanen]

I thought all Luther style cases would be counterexamples against PAP*. I'm going to assume in the following that there are volitional necessities. Of course that's a contested claim. But here's one:

'Evil Jones has a set up a cunning trap for me. If during the next hour I open my door, my very best friend is blown up. I know that this is the case and desire to do no such thing. Part of what it is to be me is to want to my friend to live - I could not get myself to kill her in any case being the person that I am. I'd rather die myself.

This feature of my will is my intrinsic, essential property. Now, PAP* claims that if I'm to be responsible for saving my friend's life I must have the power ability to do otherwise - to kill her. The relevant test for whether I have such an ability is to check resembling possible worlds that lack the properties that in the actual world mask the power to do the other option (like the device in the FSC and the addiction in the willing addict case). But in the Luther cases, like mine above, there are no such masking features that could be removed for the counterfactuals. You could try to remove the volitional necessity, but if it is an essential feature of me, then that world would lack *me* to be tested. And, in all worlds where I have that volitional necessity I save my friend - I can no other. If then the counterfactuals are the test for whether I have the power ability to do otherwise, then I have no such power. But, I take it that my friend would still thank me for saving her and think I'm responsible for it.

[Neil]

Okay, got you. Turns out it was me who was confused.

Well, there are several ways we can go. One is to deny that the values underlying the volitional necessity are intrinsic properties of me. That very rapidly takes us into issues of personal identity. At least in one sense, they're clearly not intrinsic properties - standard reidentification criteria for personal identity are going to give you that result. On another kind of account of personal identity - the characterization sense of personal identity - they are plausibly do count as intrinsic. So one move would be to hold that the reidentification sense is the relevant one. Though I think in general the characterization sense is usually the appropriate sense in normative contexts, we may be moving out of normative contexts and into metaphysical here - so the move has something going for it.

Another move would be to follow Smith: we can argue that sometimes to discover what an agent's intrinsic properties are by bracketing other properties that are also intrinsic. Just as the chameleon's shyness might be intrinsic, yet we bracket it to discover whether it's disposed to look green, so we can bracket the values underlying the volitional necessity to test for your power-sense abilities. I'm warming to this move: remember that we're not concerned here with which possible worlds are actually accessible to you. So the fact that you can't 'turn off' these values needn't worry us.

Finally, it may be that a backtracking account could still work: you may be responsible for having these intrinsic properties because of some earlier action of yours. At a first pass, it seems to me that this is only going to work if we can backtrack to a time at which you possessed the occurent-sense ability, and not just the power sense. I'm less happy with this move, because it may well be the case that the values are unsheddable and always have been - you may have been enculturated into them. But this move is still possible for me, given my quite radical restrictivism. I think people are responsible much less often than we normally think. It could turn out that despite appearances you're not responsible for your actiom.

You said above that you thought the account could be tweaked to deal with the problem. What's your suggestion?

[Jussi Suikkanen]

Neil,

thanks for a great reply of the options. I'm not sure what I can say. At this point things get very abstract and speculative. Here's few initial thoughts:

1. The personal identity option. I agree with you that we get into deep waters here. I think we face a bit of dilemma. On many characterisation sense of personal identity volitional necessities might count as intrinsic, essential features and the counterexample goes through. If you then go for a weaker view of identification, then the problem might be that there's not going to be cases where PAP* would be violated - looking across the possible worlds with those loose criteria (for instance physical) we might always (or at least way too often) find the agent doing something else than what she does in the actual world. This would mean that we would get too many responsible agents for the agents too often have the power-abilities. Maybe there is room inbetween the horns.

2. The bracketing move. I'm not sure how that would work here. If we bracket my volitional necessity of caring about my best friend, I'm not sure we would find a power to kill her underneath. If we remove such basic carings, I'm not sure there would be motivations in me left to be moved to do anything.

3. I actually had in mind with tweaking something like the third choice. So the PAP** would go roughly something like that the agent needs to have a power ability at t to do otherwise or failing that she would have needed at an earlier time a power ability to avoid making herself such that she does not have the power ability at t. I too worry that in adopting basic carings there may not be usually a chance to do otherwise.

hmh. I think I need to think about this more. Maybe one day I'll even put something down on the paper... A lot of thanks anyway for this discussion.

[Kadri Vihvelin]

Neil,

Exactly!

The dilemma you describe for Frankfurt stories in your reply to Tomkow is precisely the one I was trying to demonstrate in my 2000 paper. Let me put the case once again, using your terminology which I think is much more evocative than mine.

There are two different ways to tell a Frankfurt story. Either:

(i) Black’s intervention would be triggered by a "responsibility grounding event".

or

(ii) Black’s intervention would be triggered by some reliable prior indicator (which is not itself a responsibility grounding event).

The problem with stories of the (i) sort is that:

(a) They don't rob Jones of the ability to refrain from the responsibility grounding event (or responsibility for it).

(b) While they rob Jones of the ability to cause alternatives to the events which are causally downstream from the responsibility grounding event, and while we sometimes will hold Jones responsible for some of these events, the responsibility always seems inherited from the responsibility grounding event and the fact that Jones had alternatives to it. This is just as we would expect if PAP expresses a deep connection between responsibility and alternatives.

On the other hand, the problem with stories of the (ii) sort is that *Black* doesn’t deprive Jones of alternatives to the responsibility grounding event. The only arguments that he does involve modal or counterfactual fallacies (versions of the fatalist fallacy, or causal backtrackers).

And this is so whatever your account of responsibility or the responsibility grounding event.

In my 2000 paper, I was trying to show this in an austerely formal way. I labeled the (i) and (ii) kinds of cases "conditional" and "counterfactual" intervention and I used the ability of a coin to come up otherwise, and the fairness of the toss, as my surrogates for the human ability to do otherwise and the responsibility of an agent. I wanted to show that Frankfurt stories fail as a matter of logic independently of any particular theory of agency, ability, or responsibility. Apparently many people found this too abstract. Which is why, in my reply to Fischer, I made the arguments in terms of the homely example of the ability to ride a bicycle. This prompted some people to accuse me of “bait and switch” ! But at bottom the argument was the same. Maybe adapting your terminology will help more people to see it.

In particular, I think the above makes it clear why Fischer's way of defending FSE’s doesn't work. Fischer seems to think that if we have a Black of the (ii) sort who is an infallible predictor, but keep him hovering close by to do (i) sorts of interventions, we will end up with an FSE that performs as promised. But (ii) style additions to (i) style stories can't help unless they deprive Jones of alternatives to his responsibility grounding acts and they don't-- not because my theory of agency says they don't-- but as a matter of logic.

Even if you you don't agree with all this, thanks, Neil, for giving me some tools to make my case in another, and perhaps, more persuasive way.

Like Warfield, I think we can learn from Frankfurt's arguments even if they fail. But unlike Warfield I don't see how we can learn anything unless we are prepared to press the fundamental question of whether or not the arguments work and face up to the awful possibility, as Fischer put it, that thirty-five years of “vigorous” discussion by “sophisticated philosophers” rest on “a basic logical mistake”. Warfield found Tomkow's remarks rude. Maybe. But I found Warfield's reproach to Tomkow disturbingly anti-intellectual. How can it ever be "off topic" to wonder if a philosopher's arguments are valid?

[Eddy Nahmias]

OK, now that Kadri (using Neil's terminology) has put things this way, I think I'm getting clearer on what's going on. It seems like these complaints about FSCs are going back to Widerker's original complaint that they either require determinism to work or don't rob Jones of the alternatives relevant to ground moral responsibility (basically because we can trace Jones' responsibility to some action for which there were relevant alternatives). I admit I haven't read all the relevant papers and posts here, so if I'm missing something obvious, sorry, but if this is the basic issue, please explain:
1) why the Mele/Robb FSC which includes indeterminism fails.
2) why it is a logical fallacy to build a FSC with a global intervener who has perfect foreknowledge even in an indeterministic world, or if that doesn't work,
2) what's wrong with a FSC with a global intervener who intervenes when he sees a reliable prior sign of undesired choice, a sign which is a sufficient condition for that choice (so there's a "pocket of determinism"), but there is lots of indeterminism in the universe including in Jones' life (so the case does not rely on universal determinism).

More generally, do the "FSC-bashers" think that FSCs can be shown to be useless without relying on some sort of Beta-style (transfer of nonresponsibility) principle?

[Fritz]

Vihvelin says:

'unlike Warfield I don't see how we can learn anything unless we are prepared to press the fundamental question of whether or not the arguments work and face up to the awful possibility...'

I'm not quite sure why she thinks *I* hold the view that we can learn something here without being prepared to press the question of whether the Frankfurt arguments work. Like Vihvelin (and several others) I have pressed this question (and the negative answer to it) and think it's important to do so.

I also don't hold the view, and did not say here or anywhere else, that it is "off topic" to wonder whether a philosopher's arguments are valid. What I said was "mostly off topic" were remarks pressing on whether the examples work at all made after Neil had *already* in an earlier reply to the same person said that

"I'm content to limit the scope of the paper, so that it is not intended to show you anything interesting if you don't think that FSEs work *at all*."

To respond to this by saying (roughly) "but didn't Vihvelin show that the examples don't work at all?" struck me as "mostly off topic". But apparently at least Vihvelin disagrees.

[Neil]

Yet another flight, yet another late posting. I believe in civility in philosophical debate. But I've met Robert Allen, infamous for incivility, and I'm convinced from talking to him (from getting to know him a little, not from explicit arguments advanced to that effect) that (apparent) incivility can reflect nothing more than passionate involvement in the issues. If there's been any incivility in the exchanges, I interpret it as reflecting that kind of commitment. In any case, I do not hold anyone responsible, because I know that if the agents involved had not been uncivil on their own, a nefarious neuroscientist would have intervened.

Eddy, I take it that Kadri's dilemma defence is different from the Widerker/Kane dilemma. She does not require (or want) the assumption of determinism. It's a related charge, but not the same one. Instead, the claim is that either the signal for intervention is a responsibility-ground event or it isn't. If it is, than it's more than a mere flicker of freedom. If it isn't, then Black (himself) doesn't deprive Jones of alternative possibilities. He lacks them in any case. Take Black off the scene and in the actual sequence, in which Jones fails to think a thought that is a necessary condition of his doing the right thing, Jones can't do the right thing.

As I said, I'm somewhat sympathetic to the line of argument. The stumbling block is that it may be even more restrictive about responsibility than I want to be (and that's pretty restrictive). Let's revert to Pereboom's tax evasion case (so I don't make any stupid mistakes that will take the debate off in an irrelevant direction), but leave the counterfactual intervener out of it. So: Jones is contemplating cheating on his taxes . He is so psychologically constituted that he will only choose not to cheat for moral reasons, and that it is a necessary condition (but far from a sufficient - we may assume indeterminism here, if we want, but only to make it vivid that it is not a sufficient condition) of his so choosing that a certain moral consideration, M, occurs to him (involunarily). Recall that if there were a counterfactual intervener, Jones's thinking of M would be the signal for intervention. As a matter of fact, though, there is no counterfactual intervener (if you can imagine such a scenario). Jones fails to think of M, and therefore fails to choose not to cheat on his taxes.

As I understand the line of argument, Kadri and Terrance are committed to saying that Jones is *not* responsible for cheating on his taxes, because he lacked alternative possibilities. He did not control (by hypothesis) whether he thought of M, and since his thinking of M was a necessary condition of his having the (occurrent-sense) ability to refrain, he did not control whether he refrained. As I say, I'm somewhat tempted to agree (I think the epistemic conditions on responsibility are the least well-understood in the debate; I further believe - and have argued quite extensively, most recently in a forthcoming Synthese article - that attention to them leads to restrictivism). Notice, however, that if this line of argument works, there may be a lot less responsibility in the world than we think. Agents like Jones might be *very* common, and moral responsibility *extremely* rare.

[Kevin Timpe]

Neil,
I'm intrigued by what you say here. I agree with you that the epistemic condition required for free will warrents further discussion. For example, I think that Pereboom's suggested condition is too strict (which, if I remember correctly, appeals to an agent knowing and doesn't account for cases of culpable ignorance). Do you know when your Synthese paper is forthcoming? If it's not real soon, would you be willing to send me a copy of it via email? I think that this is an especially important given the connection between the epistemic condition and the tracing principle, which is discussed some above. And here I think there is some overlap between the issues you are discussing and Manuel is discussing in his paper--since for him and his 'semi-structualist' account, tracing becomes important when BASRA (what an AWFUL acronym, Manuel! I concur with your footnote 3) isn't fulfilled. But I also thinks that Manuel does a great job in his recent Midwest Studies paper of showing how problematic it becomes to satisfy the epistemic condition in those cases that ground the tracing.

[Terrance Tomkow]

It is depressing how a dialogue conducted over few postings can go so wildly off track.

Levy wrote a paper in which he said FSCs don't work for a certain sense of 'can' but they do for the "occurent sense" of can.

I argued, in some detail, that, no, they don't work for the "occurrent sense" either . I think our exchange has been fruitful though I'm still waiting for a definition of the "occurent" sense.

Warfield objected I was off topic because I think FSC's never work and Levy was addressing those who think they sometimes do. What, one wonders, does Warfield think "the topic" *is*?

On another matter, let me point out that I did not advocate the view Levy calls "restrictivism". It is not the basis of my objections to FSCs. I merely offered it as an example of a way one might argue that Jones is unable to do otherwise. In fact, I think that the "restrictivist" view (though not under that label) underlies many people’s incompatibilism. It is certainly worth arguing about. My point was that Black (and Frankfurt) are *irrelevant* to any debate over "restrictivism" as they are, I think, to most interesting debates in this area.

[Kadri Vihvelin]

Eddy.
You raise some reasonable questions, but I think you have misunderstood my argument. I am not an incompatibilist, and my objection to Frankfurt stories is not based on the charge that they require determinism. The dilemma is an austerely logical one. It’s not relevant whether an FSE assumes or denies determinism; it’s not relevant what you mean by “can” (so long as you don’t mean by “can” what the fatalist means: “compossible with all the facts”). My argument is that the FSE-based argument against PAP fails, period.

About your more specific questions:

1) Why the Mele/Robb argument fails: See p. 12 of my 2000 paper (posted online; see the first link in the Comments section for Fischer’s paper).

2) I don’t claim that it is a “logical fallacy to built a FSC with a global intervener who has perfect foreknowledge even in an indeterministic world”. I claim that the arguments don’t work even if we grant the assumption that this is possible. See pp. 13-22 of my 2000 paper and pp.11-12 of my online response to Fischer.

3) Re the problem with Frankfurt stories with “a global intervener who intervenes when he sees a reliable prior sign of undesired choice, a sign which is a sufficient condition for that choice…but there is lots of indeterminism in the universe including in Jones’ life”: See pp. 13-20 of my online response to Fischer (and make the corresponding adjustments so that there’s lots of indeterminism).

More generally, though, my claim is that the details of Frankfurt stories do not matter. Black is either what I called a ‘conditional intervener’ or he is a ‘counterfactual intervener’; as Neil so helpfully put it, he is either someone whose intevention is conditional on “the responsibility grounding event” or he is someone who intervention is conditional on some earlier event that is a reliable indicator of the responsibility grounding event. Insofar as he is a pure conditional intervener, he deprives Jones of some of his alternatives but not all. Insofar as he is a pure counterfactual intervener, he doesn’t deprive Jones of any alternatives.

[Fritz]

Actually Terrance, it was Neil who first objected that you were off topic -- as he said:

"I'm content to limit the scope of the paper, so that it is not intended to show you anything interesting if you don't think that FSEs work *at all*. That leaves me a large and respectable audience. So I don't think I need to reply to your line of argument."

But rather than rehearse what is right here in the comment thread, I'll go away.

[Neil]

Kevin,

copy of the Synthese paper coming your way. No idea when it'll come out - fairly soon, I hope. It'll be very clear to you that some of the ideas there are precursors of the line I take here. Luckily only some. I agree with Manuel totally: the tracing condition is difficult to fulfil. But whereas Manuel goes modus tollens on this, I go modus ponens. That is, Manuel thinks (I thiink he thinks) that if we reply upon the tracing condition there would be much less moral responsibility around then we believe; hence we ought not to reply upon the tracing condition. I think we have no choice but to rely upon the tracing condition, and for that reason (among others) there is a lot less moral responsibility around then we tend to believe.

Terrance, I'm not going to give a precise definition of occurrent right now. I agree that I owe you one. I even think that it wouldn't be hard. But my head is too stuffed with straw to trust myself today. Here's an imprecise pointer: an agent has the occurrent-sense ability to phi when the agent has the ability (=power sense) plus the opportunity to phi. Occurrent-sense ability concerns access to possible worlds at a time.

My point in introducing restrictivism was this: I did not take you to be advocating restrictivism; certainly not radical restrictivism. My claim is that if you think that FSEs don't work because they fail at the first hurdle (yout (1) above), in the manner I suggested and Kardi seconded (actually, given her historical priority, Kadri zeroed) then you (and her) are committed to a radical restrictivism. If agents are responsible only when they have (occurrent sense) alternative possibilities, and the agent in my case lacks alternative possibilities even in the absence of a counterfactual intervener, then we turn out not to be morally responsible far more often than we think. It would be appropriate for a defendant in a murder trial to plead not guilty, on the basis that she is psychologically constituted so that she could only refrain from murder if a certain reason involuntarily occurred to her; since it didn't, she isn't responsible (and if courts ignored this kind of plea, it would have to be for policy and epistemic reasons, not because it mightn't often be true).