Finking Frankfurt

Things have been pretty lively in the Garden of late. I hope this relatively technical post won’t put an end to that…

I’ve been thinking about what Frankfurt-style cases do and don’t show. I take it that they do show that alternative possibilities (APs) are not necessary for agents to be responsible for (token) actions. But I’ve been playing with a possible response on the part of those who think that alternative possibilities are necessary for MR more generally. Mightn’t they argue that though MR for token actions does not require APs, nevertheless the power to exercise dual control is necessary?  ("Dual control" is  Kane's term; I take it to be a synonym for what Fischer calls “regulative control”).

What started me down this road is noticing the similarity between the structure of Frankfurt-style cases and cases in which dispositions are finkish. A disposition is finkish (inter alia) if s, the stimulus which elicits it, also causes that very disposition to vanish. Consider, for instance, Mark Johnston’s ‘shy but powerfully intuitive chameleon’. Johnston’s chameleon is green, but, because it is so shy, it blushes bright red whenever it intuits that it is about to be put in a viewing condition. The chameleon is therefore disposed to look green, in normal viewing conditions, unless and until it is actually about to be in normal viewing conditions.

The shy chameleon is disposed to look green, though it never actually looks green. Its disposition is masked by the very stimulus that is necessary and sufficient for eliciting it (being placed in normal viewing conditions). Its failure actually to give response r does not count against its possession of the disposition to r. Now turn to agents in Frankfurt-style situations. Their apparent moral responsibility for their choices and actions, despite their inability to choose and to act otherwise than they do, is supposed to show us that APs are irrelevant to moral responsibility. But we might argue that these agents are appropriate targets for the reactive attitudes because they have the power to exercise dual control. Powers, like dispositions, can be masked without their possessors losing them (even temporarily).

Frankfurt-style cases might show that APs are not necessary for MR in individual cases. But they do not show that the power to choose alternatives is not a necessary condition of moral responsibility. Principles like “ought-implies can” might therefore better be understood as a claim about the powers of responsible agents: such agents are under moral obligations only if they can choose alternatives – that is, only if they have the power to exercise dual control over their actions. The demonstration that we can be morally responsible for something on a particular occasion, when our power to exercise dual control is masked, does not show that we need never possess that power.

Comments

[James Gibson]

To keep the Garden in its lively state...

Thanks for this post, Neil. I'll bite. Your argument, I take it, is supposed to show that FCs prove, perhaps on their best days, that alternate possibilities are not necessary for responsibility for action tokens. But they do not demonstrate that having dual-power in general is not a necessary property possessed by the agent, though masked in the Frankfurt scenarios. Accordingly, not having the power to do otherwise has not been demonstrated to be an unnecessary requirement for responsibility in general.

So, the thesis (T) is in part this:

(T) For some token action x at t, FCs truly show that S need not have the power at t to bring about an alternate course of action with respect to x (i.e. the complement of x) for S's being responsible for x at t.

OK. But this seems to entail:

(2) For at least one token action x at t, the power to bring about a different action at t with respect to x is not necessary for S's responsibility for x at t.

And (2) entails that

(3) For every action x, an agent is responsible for x only if the agent could bring about ~x

is false. (I am leaving aside the issue about whether FCs can deal with the intention to do otherwise).

You say that we may still have this power, but it is apparently masked in the Frankfurt case. So, our having the power to bring about alternate courses of action at least sometimes is not disproved to be a necessary condition for responsibility, even if the power is masked in the case. I think the Frankfurt defender should grant that we may still have this masked power as you suggest, but ask, "if (3) is false, then what role does the property *having dual-power on some occasions* play in assessing anyone's responsibility on any particular occassion?" So, I'm asking? :)

I tend to think that what your argument really shows is that should anyone choose to take that line of defense of dual power as a necessary condition for moral responsibility, even if masked or not exemplified in any particular action, they now have the burden of showing why our having that control is explanatorily relevant to responsibility at all. Imagine a world in which our entire lives are one big Frankfurt scenario and we never exemplify that property in any actual sequence, but only exemplify guidance control. Should we ditch agents being in fact responsible?

[Derk Pereboom]

I say a bit about this idea in Living Without Free Will, pp. 27-8. Michael McKenna has in the past argued for this type of view in "Alternative Possibilities and the Failure of he Counterexample Strategy," Journal of Social Philosophy 28 (1997), pp. 71-85. There he says that a necessary for moral responsibility is a power to be the author of one’s action or not. Scotus and Ockham took a similar position. A crucial fact about this sort of condition is that it could be satisfied even if the agent cannot currently activate this power, for the notion of retaining a power while the agent currently cannot activate it is coherent. Even when Maurice Greene (or the current champion) is asleep, it would seem true that he retains the power to run 100 meters in 10 seconds, despite the fact that his being asleep is currently an impediment to his activating this power. As this shows, there are cases in which it is natural to agree that an agent has a power even though it cannot currently be exercised. However, it would be implausible to maintain that no matter what the nature of the external impediment, the agent still retains the power. Suppose, for example, that a patient has a tumor that puts pressure on his brain so that he can no longer do cutting-edge mathematics. If the tumor were not putting pressure on the brain, he could do the mathematics. But imagine that it is causally impossible to remove the tumor, or for its existence to cease in any other way, without the patient dying. Then, it would seem, he lacks the power to do cutting-edge mathematics. Analogously, suppose that in a Frankfurt-style case – I use my Tax Evasion case – the neurophysiologist has implanted the device in the agent’s brain, which is triggered by the occurrence of the moral reason, but she has also made it causally impossible to remove or disable the device without killing him. Joe then permanently cannot choose not to evade taxes. Under these circumstances Joe would appear to lack the power not to evade taxes. But still, it is intuitive that he is morally responsible. So the proposed condition would seem mistaken.

[Neil]

Oh dear, reinventing the wheel again. Curse my memory! I agree, James, that much more needs to be said about what role dual power plays in underwriting MR. I don't have any argument to give at this point, but it seems to me very plausible to think that it does play a role. First of all (and this goes to Derk's response as well) all of us - compatibilists as well as libertarians - think that behavioral flexibility is a necessary condition of MR, right? That's why we assess MR by going to counterfactuals. Nefarious neurosurgeons ensure that our behavior is not flexible. Second, notice that epistemic openness seems a condition of MR. The agent must believe that how they act is open, or they are not MR (interesting thought experiment: you are in a FC, but you do not know what the neurosurgeon wants you to do. You X. Are you responsible for X-ing?)

Derk, I'm not convinced that the agent in your example lacks dual power. A lot depends on how we conceptualise ability. Think of Perry's way of thinking about it: I have the ability to X just in case X-ing, directly or by way of Y-ing, is among the action routines stored in a particular module in my brain. If that's right (if, that is, something like this tracks our intuitions about ability), then in your example the agent retains the relevant power.

[Derk Pereboom]

Neil -- that's a difficult issue, I think -- my intuition is that the agent does not retain the power, but I expect that reasonable people will have differing intuitions about examples of this sort. But backing up a little, isn't implausible that a power for an alternative possibility that I have, but that I cannot exercise at the time of my choice, can explain my moral responsibility for that choice? And wouldn't especially those who are attracted to the idea that having alternative possibilities explains moral responsibility resist the claim that I'm still responsible when I can't exercise my power for alternatives?

[John Fischer]

For whatever it is worth, some of my thoughts on these matters are in my paper (on the blog), "The Frankfurt Cases--the Moral of the Stories."

Also, I take it that Neil's reference to "Finking" is a reference to Martin's work on "finkish" properties in the philosophy of language. Is this correct? I have for some while wondered if there might be fruitful connections with the discussions in that area of philosophy of lanugage and our (obsessive?) ruminations about the Frankfurt-type cases.

[Neil]

I agree that the analysis of ability is difficult. I actually get no strong intuition one way or the other in these cases. In the case of a finkish disposition, the analogous move would be to claim that the entity loses the disposition just in case the masking agent cannot (necessarily? contingently?) be removed from the entity consist with its further existence and possession of the relevant property. You might argue that the shy but intuitive chamelon has the disposition to look green because you can easily imagine a chameleon therapist curing it of its shyness. But a mirror which does not shatter when hit with a hammer is not fragile, because the goblin who casts a spell to prevent the shattering iff it is hit with a hammer cannot be removed from it without its losing its fragility. This doesn't seem ad hoc to me. But I'm still not convinced... why shouldn't our ability to imagine the hobgoblin/brain tumor away suffice?

How is all this relevant to MR? I think it cuts across the issues in FCs to some extent. I'm conceding a lot to the adherents of FCs: token (if you like) dual control is not necessary for MR. But the power to exercise dual control gets you in the ballpark...

John, I'm aware that finkishness has a prehistory, and that Martin is the first link in the causal chain leading to its use today. But my knowledge of finkishness comes from the metaphysics literature, and especially Lewis's paper in Phil Q.

[Daniel Cohen]

It might be worth pointing out that Michael Smith has made a similar connection between finkish dispositions and the Frankfurt cases. (See his ‘A Theory of Freedom and Responsibility’ in Cullity and Gaut (eds.) “Ethics and Practical Reason” OUP 1997, and ‘Rational Capacities, or: How to Distinguish Reckless, Weakness, and Compulsion’ in Stroud and Tappolet (eds.) “Weakness of Will and Practical Irrationality” OUP 2003.)

Smith’s way of developing the point is somewhat different, however, to Neil’s. As Smith puts it, in order to hone onto the intrinsic properties of an agent that constitute their ability to do otherwise, we must abstract away from any other ‘masking’ properties of the agent or their environment (eg. the chameleon’s shyness, or Black) that might prevent the intrinsic property from manifesting itself. (Thus Smith, unlike Neil, holds that agents in Frankfurt cases *do* in fact possess the ability to do otherwise.)

My worry about this proposal is that it seems in danger of over-generalizing. It gets the right result in saying that agents are still responsible in Frankfurt cases, however it seems to imply that agents may be responsible in less obvious cases as well. Consider another case (discussed by Smith) where a woman fails to form an accurate belief (about whether she should have another drink) because she is subject to a kind of self-hatred “that she could neither acknowledge nor get rid of” (Smith 2003, p. 18.) Intuitively, this is a case of ‘epistemic compulsion’, where we should deny that the agent in question possesses the capacities necessary for moral responsibility. Smith’s analysis above, however, suggests that we should abstract away from the woman’s self-hatred, in order to determine whether she has the capacity to form the correct belief. But if we do this, it may turn out, contrary to our initial intuitions, that the woman does indeed possess the capacity to form the correct belief.

(Derk’s tax evasion case, above, would seem to show that the ability to do otherwise isn’t necessary for responsibility. This case seems to show that, at least on Smith’s analysis of the ability, it isn’t sufficient either.)

[V. Alan White]

Since this is my initial foray into discussions on this site, let me first say how delightful it is to have such a great net resource as this for free will research. I found this site through Brian Weatherson's, and my highest compliments for its format and content, drawing some of the best minds in the field. Simply terrific!

Now, for my two cents on Frankfurt examples, and why I see no prospect of significant rapprochement between advocates and naysayers.

From quite another direction, I was digesting William Lycan's thoughts on “The Gettier Problem Problem” (wonderful paper—Weatherson links to it) when it occurred to me that Gettier and Frankfurt counterexamples have a lot in common, at least at first glance. They both get traction from some kind of epistemological isolation of their protagonists from objective metaphysical conditions that might affect our broader judgment about what's really going on. For Gettier victims (if that's the word), they are unaware of objective conditions that rob them of real (“full” or correspondence) knowledge despite having justified true belief. For Frankfurt miscreants, they are unaware of objective but not-actually-active overdetermining conditions that otherwise would have required them to misbehave as they actually do. A disjuncture between what constitutes the content of the subjective states of people and the objective conditions under which those states are formed is the common basis for philosophical significance of the counterexamples, even though there are opposite effects of disconnection. In Gettier, knowledge fails because of the disconnection (a subject's lack of knowledge of interloping conditions); in Frankfurt, responsibility holds despite disconnection (a subject's lack of knowledge of the overdetermination).

Certainly the two sets of counterexamples are comparable in the rich literature they've produced as well. But as Lycan says about the Gettier case, we seem to be no closer to getting a consensus on its solution, and that may have implications for pessimism about analytical philosophical projects in general. And parallel remarks might well be applied to the Frankfurt case.

But if the Gettier case is a tough nut for analytical nutcrackers, Frankfurt's counterexamples are conceptual diamonds. The reason for this I think is relatively straightforward—and ironically, can be shown analytically.

If we encapsulate under entailment the justified-true-belief view (JTB) that Gettier challenges it would be:

(a) [(J & T) > K]

where J is the justification, T the true belief, and K knowledge. So Gettier claims:

(b) ~[(J & T) > K]

and backs his claim with plausible, non-equivocal interpretations of the components of these to show that the following occurs in Gettier cases:

(c) [(J & T) & ~K]

which shows (a) is false.

In short, Gettier counterexamples exhibit abject logical conflict with the JTB view.

Now similarly cast under entailment a view of moral responsibility amenable to PAP or something like it:

(a*) [R > (I & A)]

where R is moral responsibility, I is something like (appropriately reflective) evil intent, and A is some availability of alternatives. (Note: the form [(I & A) > R] cannot provide the basis for Frankfurt-style counterexamples since its denial leads to logically posing (I & A) against R.) Frankfurt's examples then behave something like (a)-(c) above in claiming:

(b*) ~[R > (I & A)]

by describing in presumably non-equivocal components the situation

(c*) [R & ~(I & A)]

where the negation conjunct is false precisely because A is and thus we have the same logical clash of (c*) with (a*) as with Gettier's (c) with (a).

However, the formal difference of (a) and (a*) is obvious. Because of that (c) in effect criticizes a conjunctive sufficient condition in (a) by showing its consistency with a false consequent; (c*) in effect criticizes a conjunctive necessary condition in (a*) by falsifying one particular conjunct.

In short, Gettier showed that a particular combination of something ain't enough to produce knowledge, even though it's left open what is enough (if anything). Frankfurt shows that a particular something is too much for responsibility, though it's left open what's enough for responsibility (beyond the suggestion that condition I might be).

Even shorter, the key topics of the conditionals (a) and (a*)—K and R—are in distinct formal places of controversy. There's no assumption that K is ever true in (c), because it is the conjoint condition of JTB that is under scrutiny. But in (c*) the truth of R must be assumed, and one condition of it dismissed. But nothing is learned about why R is true in the first place beyond the assertion that condition I is true as well.

By general form, Frankfurt counterexamples are more limited than Gettier's in logical force and what can be deduced from them even if they are successful in contradicting form (a*). They are just inherently more wide open than Gettier counterexamples with what they can claim to show with respect to moral responsibility as opposed to what Gettiers show about knowledge since responsibility is assumed as part of their description.

I know that in some sense all this is well-known. But the stark contrast with Gettier makes it clearer at least for me in a basic way why Frankfurt counterexamples will remain open to criticism.

[Neil]

Daniel,

Ignorance has been (at least since Aristotle) one of the traditional excuses. The debate here seems to concern necessary conditions, not sufficient conditions. So Smith's agent can get into the responsibility ballpark, on the suggested analysis, w/o that committing proponents to holding that she is responsible. You find it is counterintuitive to hold that she has the capacity in question. But perhaps your intuition is 'fed' by the better grounded intuition that (despite possessing the capacity) she is not responsible.

[Daniel Cohen]

Neil,

Smith treats both the practical and epistemic dimensions of responsibility on par – both require capacities: the capacity to desire differently than one actually does, and the capacity to believe differently than one actually does. Presumably, his full account of responsibility for action would make appeal to both of these capacities, something perhaps as follows: one is responsible for acting in a certain way if and only if (a) one has the capacity to form an accurate belief about what one ought to do, and (b) one has the capacity to desire to do what one believes one ought to do.

Given this, it cannot straightforwardly be the case, on Smith’s account, that ignorance is an excuse: one’s ignorance will not be excusable if one had the capacity not to be ignorant in that way.

Smith’s example about the woman who fails to form an accurate belief about what she ought to do is supposed to be an intuitive case of someone who lacks responsibility for forming a belief *because* they lack the capacity to believe differently. (The woman’s excuse is not simply that she is ignorant – this fact leaves it open whether or not she is legitimately to be held responsible. In fact, it is with regard to her ignorance that the very question of responsibility arises.)

The relevant intuition here is that the woman lacks responsibility for her ignorance because she lacks the capacity to believe correctly. This intuition will be problematic if I am right in my argument that, on Smith’s account of capacities, she should be attributed with the capacity to believe differently.

[Neil]

Here is an extract from a paper I've just finished writing, on doxastic responsibility. I think it addresses these concerns:

The compatibilist account of responsibility for belief advanced by Pettit and Smith (1996) also seems to issue in an asymmetric account of responsibility for belief. Pettit and Smith argue that we believe freely just in case our believing is governed by certain conversational norms; in that case we are responsible for our beliefs. There are two possible ways to interpret Pettit and Smith, but on both their view implies an asymmetrical account of responsibility for belief, like the one to which Heller (malgré lui) is committed. Pettit and Smith argue that responsible believers are “orthonomous”, which is to say they are governed by right reasons. Perhaps, then, an agent is a free and responsible believer just in so far as her beliefs are governed by the evidence: if I believe that p, and all the evidence available to me supports that p, then I am praiseworthy for my belief; otherwise I am excused responsibility. On the other possible interpretation of their account, an agent counts as a free and responsible believer just in case their belief-forming mechanisms are responsive to evidence. On this interpretation, however, if I believe that p, and my evidence support my belief (regardless of its truth), then I am praiseworthy for my belief (since I have done my epistemic duty), but if I believe that p in defiance of my evidence then my belief-forming mechanism is not properly responsive to evidence, and I am excused blame. I could only be blamed for my belief if I believed that p, despite knowing that my evidence supports ~p and while remaining rational. But rational agents cannot satisfy these requirements, as Pettit and Smith themselves recognize: since your beliefs give you ‘your sense of what is and your sense of what appears to be’, you are ‘denied an experience whose content is that you are believing such-and-such in defiance of the requirements of facts and evidence’ (1996, 448). Pettit and Smith deny they are committed to such an asymmetry, but it seems that they are wrong.