Taking Responsibility for Luck

The Garden has been mighty quiet lately. Of course, winters aren't really the best time of year for gardens in general, so perhaps it's not too surprising. Still, I'd like to see us get back into some philosophy, so here's some food for thought.

Luck poses at least the following two problems for libertarianism: (1) It hurts, and (2) It doesn't help. That is, according to some objections, the indeterminism required for libertarianism actually diminishes control, whereas according to other objections, the indeterminism might not hurt, but it at least doesn't give agents any more control than they could have had without it. Ignore the second worry for now and focus instead on the first.

Consider a libertarian who accepts most of the details of Fischer and Ravizza's theory of moral responsibility, except that this theorist thinks that moral responsibility requires an ownership condition that requires the falsity of determinism. In particular, this theorist agrees with F&R that an agent need not know the details about the operation of her mechanisms in order to take responsibility for them. Now, if libertarianism is true, then some of our mechanisms operate indeterministically. But it doesn't seem like this fact would preclude an agent from taking responsibility for the mechanism. So why should we think that indeterminism would diminish control, then?

Maybe I'm missing some good reason to think that indeterminism diminishes control, but it seems to me that as long as you have taken responsibility for your indeterministic mechanism, you're good to go (at least with respect to the first threat posed by indeterminism).

Comments

[Kevin Timpe]

Neal,

Are you thinking that the process of taking responsibility for the indeterministic mechanism could itself be a function of luck? If so, then that suggests to me that the luck involved could be of the control-undermining sort. It strikes me that there may be a parallel here between what you are suggesting and compatibilist accounts of MR which introduce an ownership condition to avoid problems arising from manipulation, only to be criticized that one could be manipulated into satisfying the ownership condition.

[Cihan Baran]

Neal,

I am having trouble understanding your point, so I am hoping you can clarify.

You write that:

In particular, this theorist agrees with F&R that an agent need not know the details about the operation of her mechanisms in order to take responsibility for them. Now, if libertarianism is true, then some of our mechanisms operate indeterministically. But it doesn't seem like this fact would preclude an agent from taking responsibility for the mechanism. So why should we think that indeterminism would diminish control, then?

I take you to mean that (1) an agent might have an indeterministic action-producing mechanisms, (2) the agent may not know about the indeterminacy (or other details) of these mechanisms and (3) the agent can still take responsibility for the actions these mechanisms produce.

I'll grant you the truth of (1)-(3) but even then, how does this help with indeterminacy "hurting" freedom? Maybe, this has to do with my lack of knowledge of F&R's theory of responsibility and that's why I think I am missing something.

After all, imagine an agent X whose mental states are affected by the result of the lottery through some fantastical and bizarre connection. (Suppose the lottery machine's signals affect his brain.) As a result of the change in the agent's mental states, suppose the agent commits some crime. Suppose further that the agent takes responsibility for the crime and is unaware of the connection between his mental states and the result of the lottery.

These satisfy (1)-(3) but it still doesn't follow that the indeterminacy of the lottery machine doesn't undermine the agent's freedom.

Now that I think about it again I think I understand your point. The agent not only satisfies (1)-(3) but also conditions set by F&R. Now the question is, is the agent responsible, supposing F&R's theory is right?

Still, my intuition is to say no. The randomness involved sort of voids the other conditions. But I should think about it a little more.

[Q the Enchanter]

I don't see how such indeterminacy would undermine the normative expectations of the moral community. It seems to me that those expectations are grounded on the probability that a member of the community will behave morally well. But that probability judgment is based on observable behavioral characteristics that are far too gross, and fraught with far too much epistemic indeterminacy, to be materially influenced by the fact that the underlying behavioral mechanisms themselves are (ontologically) indeterminate.

[Mark Smeltzer]

Neal,

The main problem I see to the approach you've outlined is one that the F&R theory seems to address rather directly. In the case of mechanisms that behave deterministically, there is a strong sense in which the output of those mechanisms can be identified with the mechanisms: we can get a sense of the mechanisms involved by looking at the results.

In the F&R theory, an agent's MMR mechanism is one that produces similar results in similar situations, and moreover, is responsive to reasons. This gives the agent (and observers) a sense of who the agent is, in virtue of the mechanism's responsiveness.

The purpose of the agent "taking responsibility" for his MRR mechanism is to complete the epistemic gap. It is similar to Frankfurt's identification: it is part recognition, part appraisal, and part volition. If the agent believes that the actions produced by his MRR mechanism are his own and they are his own, we have both belief and warrant yielding knowledge of self and, ultimately, can be extended to account for knowledge of others.

In the case of an MRR mechanism that contains ontologically indeterministic elements, the F&R theory becomes rather murky: it is possible that the agent believes the actions from the mechanism are his own, but how can it be said that they really are his own if they are affected by indetermism internal to the mechanism itself?

It seems to me that an agent's mechanism is in jeopardy of not being MRR whenever substantive indeterminism happens within the mechanism.

[Q the Enchanter]

"It seems to me that an agent's mechanism is in jeopardy of not being MRR whenever substantive indeterminism happens within the mechanism."

There are at least two plausible lines of response here. First, it's doubtful that quantum-level indeterminacy in the brain cashes out as substantial indeterminacy at the level of intention and action: There is a welter of quantum indeterminacy within all the atoms of a baseball, after all, but that indeterminacy doesn't make a pop fly any harder to field.

Second, even if the indeterminacy in a suite of responsivity mechanisms is substantial at this level, we might still be able to "own" its output: If I put a cat in Schroedinger's chamber for a spell and the cat is irradiated during his stay, the cat's death is in a strong sense attributable to me, even though the Geiger trigger was calibrated at a .5 probability.

[Chris F.]

"First, it's doubtful that quantum-level indeterminacy in the brain cashes out as substantial indeterminacy at the level of intention and action: There is a welter of quantum indeterminacy within all the atoms of a baseball, after all, but that indeterminacy doesn't make a pop fly any harder to field"

I know many people make this point about quantum indeterminacy. But why is it "doubtful" that it cashes out as "substantial" indeterminacy?

[Q the Enchanter]

Chris, the material substrates of the components of the computer you are reading this on are shot through with quantum indeterminacy; but your computer's output is nonetheless highly determinate. Why suppose QI is any more "substantial" relevant to the output of F&R's MRR mechanisms?

[Chris F.]

Neal,

Here is a really schematic response. Whatever it is that makes people think indeterminism is incompatible with, or at the very least diminishes, control is just not the type of thing that someone can avoid by taking responsibility for her action. There are features of the process leading to my action that may undermine my control and there is no attitude I can take towards this process that could help.

I suppose that was probably not the answer you were looking for since you were wondering why someone who *accepts* most of F&R's theory should be worried. I think what Kevin said above is on the right track.

Q the Enchanter,
It still seems like a jump. Sure some things we observe are highly determinate (a baseball does not strike me as the best example; the computer’s output is better). But some things do at least *appear* to be highly indeterminate, e.g. human behavior. And I would imagine it is precisely this observation that makes people wonder if QI is more substantial in the brain than in other things. I guess it seems to me that we just don’t understand Q.M. to know how (in)substantial it is at the macro-level.

[Q the Enchanter]

"But some things do at least *appear* to be highly indeterminate, e.g. human behavior."

Two things. First, there's an equivocation lurking here with the term 'indeterminate.' The output of a computer is "indeterminate" with respect to our subjective reactions to them. But this indeterminacy is epistemic, not ontological. (Along these lines, it's not hard to think of contexts in which human behavior is more predictable than the behavior of computers.) It stands to reason (assuming our behavior is mechanistic) that the behavioral "indeterminacy" we observe in human decision is at least roughly of the same character as the behavioral indeterminacy of computer programs.

Second, it's at the level of *appearance* that human instincts and practices with regard to interpersonal accountability evolved. So appearances, however indeterminate, seem at least to have been determinate enough. It's not clear how the fact that "we just don’t understand Q.M. [enough] to know how (in)substantial it is at the macro-level" argues against an externalist justification of those instincts; if we don't know, then we don't know.

These arguments aren't meant to be dispositive, but burden shifting. If someone wants to say that QI modulates our MRR in some relevant way, then, I think they need to give a positive account of why that would be; at the very least, such a claim seems to run counter to the examples of mechanism that tend to come to mind.

[Chris F.]

Hi Q,

I am not sure I followed what was going on in the second paragraph. Would you mind elaborating?

I agree that we need to be careful about distinguishing ontological from epistemic indeterminacy. By using *appearance* I just meant to suggest that the justification one derives from observing the (although as you correctly note not always) indetermancy involved in human behavoir is defeasible since appearances may be deceiving.

Your last paragraph suggests we might be talking past each other. I was wondering what justifies the claim that it is doubtul that QI in the brain cashes out as substantial indeterminacy. I agree with you that if someone wants to argue that QI does indeed affect our brain in substantial ways that he has his work cut out.

[Q the Enchanter]

Hey Chris,

Take 2 on that second paragraph: The normative expectations that ground F&R's externalist account of responsibility evolved in a cultural milieu where appearances were all members of the community had to go on. However "indeterminate" human behavior appeared to them, then, it can't have been to a degree sufficient to impede their reliance on those normative expectations -- for if it had, the relevant expectations presumably never would have materialized. (One doesn't form expectations about the unduly unpredictable.)

As to "what justifies the claim that it is doubtful that QI in the brain cashes out as substantial indeterminacy," I've pointed to baseball and computers as analogies because they seem to exemplify a general feature of mechanisms and mechanistic processes, viz., that though they are riven with indeterminacy at the quantum level, they are nonetheless determinate at the macro level. That being so, we should probably presume that QI in the brain is (similarly) unlikely to figure in any substantial way in any "indeterminacy" we find at the macro level.

[Mark Smeltzer]

Q,

You surely realize that computers are designed in a way that actively preempts the propagation of irregularities to achieve reliable, deterministic output.

So, I'm wondering how you are able to mitigate this distinction in the analogy with QI in the brain? Do you suggest that the brain is similarly designed, and if so, how is that design accounted for?

Moreover, regarding QI and MRR, I merely said that, on the F&R line of reasoning, if something internal to the mechanism produces ontologically random output, then it poses a measured threat to the mechanism's MRR status. If the effect is sufficient to reduce the agent's mechanism to WRR status (weakly reasons responsive), then the F&R line yields the easy answer of "not responsible" -- the agent's endorsement of his own actions doesn't change that. Agreed?

[Q the Enchanter]

Hi Mark,

Your original argument seemed to concern what I'll call the "ownership problem" generated by the fact that the "MRR mechanism...contains ontologically indeterministic elements," since we can wonder whether its outputs "really are [the agent's] own if they are affected by indetermism internal to the mechanism itself." I think that concern is distinct from your subsequent claim that "if something internal to the mechanism produces ontologically random output, then it poses a measured threat to the mechanism's MRR status." Though I'm not sure what "ontologically random output" would mean ("output" is something I think of as determinate, even if it's random, but maybe I'm mistaken about that, or maybe you meant something else), I agree that if the output of a responsivity mechanism is random, it is only weakly responsive, and therefore inapt to ground the normative expectations of other members of the moral community. But then this point, it seems to me, has nothing to do with QI per se, but with the observed behavior of the agent. (If the output is random, it really doesn't matter what the source of the randomness is.)

I agree that "computers are designed in a way that actively preempts the propagation of irregularities to achieve reliable, deterministic output," but it's not clear how this is relevant. The issue is whether we should presume that QI inherent in the material substrates of brains is -- any more than in the case of QI in the material substrates of computers -- a source of macro-level indeterminacy in mechanical operation. Moreover, while computers are designed to overcome many sources of indeterminacy, QI does not seem to be one of these sources. Besides which, many of the brain's operations are demonstrably nonrandom (perception would be the prime example), and so QI in itself can't be a reason to think the brain's moral responsivity is bound to be random.

[Mark Smeltzer]

Q,

I can see how my use of the word "random" threw you off. I had revised that sentence from an earlier version, yet forgot to replace "random" with "indeterminate".

In my first post, which was written spur of the moment and not very well thought out, I tried to raise two worries that are relevant to Neal's original question.

The first is to raise a problem, stemming from issues pertaining to the ownership condition, where an action is brought about only by adding a bit of indeterminacy within the agent's mechanism. Presumably, if that indeterminacy were subtracted from the situation, there wouldn't be enough left to bring about the action in question. This should lead us to doubt identifying the action with agent's MMR mechanism (at this point we're still assuming it is MMR somehow). The strength of this doubt will be proportional to the degree which the indeterminism in the equation overpowers the otherwise deterministic characteristics of the MRR mechanism.

The second worry potentially undercuts the first in that I am suspicious whether a mechanism could be classified as MRR if a source of indeterminism exists within the mechanism robust enough to alter the event outcomes of the mechanism.

In so far as we share agreement on either point, I don't see that we have much to debate. I'm not suggesting that QI does have macro-effects. I am merely attempting to formulate a response to Neal's original questions.

For instance, Neal could take a line similar to Kane and say that indeterminism is actually important and necessary on a purely event causal level. Neal asks whether a libertarian could take F&R's account and still claim robust ownership over one's actions. It seems clear that Kane's style of libertarianism is almost just that: if it weren't for the fact that he accepts the consequence argument, there is nothing in his account that would prevent him from completely endorsing the F&R account. In Kane's account he tries to show, in his own way, how the agent's mechanism could be MRR and yet still sensitive to and shaped by indeterministic forces: in situations where the agent is in a moral dilemma and his reasons for and against an action are in a deadlock then, in virtue of this deadlock, the determining factor could be QI that builds up in the mechanism until it reaches sufficient levels to cause the result to go one way or the other and resolve the deadlock. In Kane's view, this would have lasting impact on the agent's mechanism as such experiences are important to the formation of the agent's character over time.

However, there are obvious questions to consider: do such deadlocks occur in practice? No real reason to think so. Is it QI that resolves these deadlocks when/if they do happen? No reason to think so (determinism could resolve the deadlock in the same manner just as efficiently). What separates this from a compatibilist account? Seems to be only the prior commitment to incompatibilism... which means Kane's account really just boils down to a debate about the consequence argument.

Of course, there are other libertarians we could pick on, but Kane's variety seems the most similiar to what Neal has in mind.

[Q the Enchanter]

Mark, your criticism of Kane seems able. [Rim shot.]

Anyway, I can see now that this whole argument I introduced about whether quantum indeterminacy makes a difference at the level of mechanism is beside the point, since Neal's question presupposes that the response mechanism "behave[s] indeterministically." Sorry about that...

To redirect, then, the question is: Can I claim ownership over moral response mechanisms that at least to some degree behave indeterministically? I think so, and I think so even if you conservatively assume (pace Kane, for example) that such indeterminism weakens rather than enhances the degree of relevant control.

Here's the argument. Per F-R, a mechanism is moderate reasons-responsive if it is responsive to a rich array of coherent or sane reasons for doing something-or-other. But F-R's account does not require that the MRR mechanism be perfect. It only requires that it be sensitive to patterns of reasoning within an appropriate range. In principle, then, there is no reason to suppose that just any arbitrary degree of indeterminacy added to an MRR mechanism would cause its operation to fall outside that range of appropriate responsiveness. But then if an MRR can withstand at least some degree of indeterminacy, such indeterminacy appears not to be a bar to ownership.*

_________________________

NOTES
*On F-R's account, an agent comes to "own" an MRR mechanism if (1) he subjectively takes his conduct to be in some sense "his own," (2) he accepts as legitimate and action-guiding the moral expectations placed on him by other members of the community, and (3) he came by these convictions in the right sort of way -- i.e., they weren't the result of coercion or manipulation. It doesn't seem to me that any of these conditions would be defeated were the MRR mechanism indeterminate.

[Tamler Sommers]

Neal,

You write that indeterminism poses two problems for the libertarian: (1) the indeterminism required for libertarianism actually diminishes control, and (2) the indeterminism might not hurt, but it at least doesn't give agents any more control than they could have had without it.

If you can demonstrate (2), isn't that sufficient to refute the libertarian? The libertarian is committed to the view that (a) determinism is incompatible with moral responsibility, and (b) indeterminism can provide the additional control necessary for MR. So for libertarianism to be vindicated, indeterminism has to help.

My point, I guess, is that once (2) is established, arguing for (1) is just an attempt to run up the score. (Something we know about in New England.) It's true that there would be a kind of irony if indeterminism actually undermined an agent's MR. But is there anything more to this question than that?

[Dan Speak]

Hi Tamler,

It doesn’t seem to me that demonstrating (2) would be enough to refute libertarianism. This is because the libertarian might be invoking indeterminism not to establish further control but to provide for some other condition she takes to be necessary for moral responsibility. Randy Clarke thinks, for example, that indeterminism might allow agents to achieve a special kind of difference making or it might provide agents with genuine alternative possibilities. So even if indeterminism doesn’t translate into further control, it might still be a necessary condition for moral responsibility.

That said, it’s not clear to me even that demonstrating (1) would “refute” libertarianism. The same Clarkeian points would apply. And it isn’t totally implausible to imagine a libertarian who thinks that the cost of indeterminism, in terms of loss of control, is worth the goods (difference-making or alternative possibilities, say). Isn’t this just Robert Kane’s view?

So maybe you need to run up the score. Put on your hoodie and find the mismatches to exploit.

[Kip Werking]

I believe both Al Mele and Laura Ekstrom have defended, at least in part, libertarian views in which indeterminism does not enhance control. Instead, it might, for example, provide a certain "independence from the past".

But like Tamler, I'm strongly inclined to not even call these views libertarian.

[Tamler Sommers]

Fair enough, but how about if I don't link the MR explicitly with control and rephrase the question like this. Neal says that there are two challenges for the libertarian: (1) indeterminism hurts (the case for MR) and (2) indeterminism doesn't help one gain MR. All one would have to do is demonstrate (2) to refute libertarianism, right? Arguing further for (1) would be piling on.

Still, I guess I understand the issue now. If indeterminism diminished control then the advantages (if there are any) provided by the other goods--alternate possibilities or difference making--might not be sufficient. In other words, indeterminism might allow for one necessary condition at the expense of another. Is that the idea?

[David Hodgson]

This is a bit reminiscent of the debate on this site about my TLS article in July last year.

Indeterminism in the sense of randomness hurts the case for MR, but if (as I argue) there is indeterminism in rational decision-making, then this indeterminism has the advantage to MR of freeing the agent from total control by past circumstances and laws of nature, that is, by pure luck. And I think it has no disadvantags for MR.

I've tried to argue this in detail in papers on my 'Reasonable free will' website (linked to this site), and also in an article published in Ratio in September last year called 'Making our own luck'.