It's on! (Mele on Doris, Knobe, and Woolfolk)

Here's our first ever GFP Online Reading Group Event. Make it awesome.

Initial comments (below) by Al Mele on Doris, Knobe, and Woolfolk's paper "Variantism about Responsibility."

For this to work, you need to start commenting. Al's given us some great reflections below. Now you join in. Thanks in advance.
-------------------

From Al Mele:

DKW’s paper was a pleasure to read. I’ll get the discussion started with some scattered comments.

1. Frankfurt-style and Woolfolk-style stories

Many readers will have noticed a difference between the Woolfolk et al. stories about Bill and Frankfurt-style stories. In F-style stories, the potential controller doesn’t cause the agent to decide to A and doesn’t cause the agent to A. He is prepared to do this if necessary; but the agent does it all “on his own” and the potential controller stays on the sidelines. In Woolfolk’s story, the “drug makes [Bill] unable to resist the demands of powerful authorities,” and they order Bill to shoot Frank in the head. So it certainly seems that they cause Bill to shoot Frank. Even so, “subjects judged the high identification actor more responsible . . . than the low identification actor” (p. 13).

Why might that be? (I’m assuming that “more responsible” here is short for “more responsible for killing Frank” and that the kind of responsibility at issue is moral responsibility.) Consider the following suggestions about part of what the subjects might be thinking about Bill in the high-identification condition:

A. (1) the bad guys caused Bill to kill Frank, and (2) Bill’s desire to kill Frank (or Bill himself, as in agent causation) also caused his killing Frank. (This is consistent with causal overdetermination and with joint causation.)

B. A1 is true, A2 is false, and Bill has some responsibility for killing Frank because he would have killed him even if the bad guys hadn’t made him do that.

(These suggestions aren’t meant to be exhaustive, of course.)

Possibly, suggestion B has led some of you to wonder whether some subjects would say that Bill has some responsibility for killing Frank in a strange story in which it is clear that Bill does not kill Frank. Wait! Am I nuts? Well, consider a story along the following lines. Bill makes his discovery about Frank and is extremely upset. Bad guys paralyze his arms and hands, put a gun in his right hand, raise his right arm (by electronically stimulating his arm muscles) so that the gun is pointed at Frank’s head, and (by further electronic stimulation) cause Bill’s paralyzed right index finger to depress the trigger. “Bill was certain about his feelings. He wanted to kill Frank.” What’s more, Bill thought that he was acting: he did not realize that his arm and hand were paralyzed. And he felt no reluctance about blowing “his friend’s brains out.”

Suppose that “subjects judged” this Bill “more responsible [for killing Frank] . . . than the low identification” Bill. What should we make of that? Well, given that these rational subjects know that Bill did not kill Frank (even though Frank was killed), they cannot be expressing the belief that he has some moral responsibility *for killing* him. They are doing something else with their words. This certainly is possible, and it makes one wonder whether some of the respondents to Woolfolk’s own high-identification story are doing something similar with their words. Running the story I sketched would produce some evidence about this. (Suppose it turned out that the responsibility rating is not significantly higher in Woolfolk’s high identification case than in my high identification case.)

Lots more below the fold.

2. The “determinism” studies by Nahmias et al. (in *Phil Psych* 2005) and Nichols & Knobe (NOUS, forthcoming)

The idea for the supercomputer story was hatched here in Tallahassee at the Pitaria – a restaurant some of you know from conference lunches. Eddy Nahmias told the lunch group what he wanted to test, I suggested a supercomputer story that would entail that the world in which it is set is deterministic but would not use the word “determinism,” Eddy seemed a little reluctant to try such a story, and Joshua Knobe (who was here visiting for a few days and had a problem with his voice) quietly persuaded Eddy that it was a good idea. My thought about not using the word “determinism” was that, when people learn this word outside of a philosophy class, they “learn” that part of its *meaning* is “something that precludes free will and moral responsibility.” And even if one defines it for them in a standard philosophical way and encourages them to ignore what they used to think about its meaning, old linguistic habits die hard.

Now, Nichols & Knobe (NK) don’t use the word “determinism” in the text they give their subjects. But they do something related. They say that in universe A, “if everything . . . was exactly the same up until [etc.] it *had to happen* that . . .” and that “given the past, each decision *has to happen* the way that it does.” How is this related? Well, traditional compatibilists contend that agents sometimes “could have done otherwise” in deterministic worlds (in a sense of the quoted expression that they say is relevant to moral responsibility) while granting that the entailment from laws + past to future events holds; and NK’s description of universe A steers subjects away from a traditional compatibilist view of things. (The assertion that it had to happen that S X-ed is naturally read as entailing that S could not have done otherwise than X.)

I’m not suggesting that the abstract/concrete difference makes no difference in subjects’ responses. (Indeed, I would reject this suggestion. On concrete vs. abstract questions about intentional action see my 2001 paper in the Malle, Moses & Baldwin MIT volume [pp. 27-35 passim].) But the *had to happen* element in the description of universe A may be making part of the difference, and I think it would be useful to run a test of the abstract/concrete effect that dispenses with that element and focuses on capturing the entailment business that defines determinism. This can be done in non-technical terms with a description of universe A that features a supercomputer that predicts all future events with absolute certainty based entirely on its complete information about the past. (This will trouble Humeans about laws of nature, but the story should be kept pretty simple.)

Here’s another point. As DKW say, “The results were dramatic.” But might a big part of the effect be explained by context switching? Many subjects may understand “fully morally responsible” in the question about the concrete story in an every-day sense corresponding, in the case of bad actions, roughly to “guilty and fully punishable according to law,” and they may understand the same words in the abstract question in a much more metaphysical or theoretical way. (For relevant discussion, see DKW, pp. 19-20. Some readers may find pp. 71-72 of my *Free Will and Luck* interesting in this connection.)

3. Knobe’s chairman studies

I was very impressed by these studies when I saw them several years ago, and I still am. More recently, I was impressed by a finding reported in a draft by Shaun Nichols and Joseph Ulatowski (NU). In one study, they asked subjects why they answered the question about the chairman as they did, and in another, all subjects were asked about both chairmen (to test for an order effect). Subjects who gave the majority response in the “Harm” scenario typically appealed to what the chairman *knew* in explaining their answers, and those giving the majority response in the “Help” scenario typically appealed to the chairman’s lacking an *intention* or *motive* to help the environment. When other subjects were asked about both scenarios, a very interesting pattern emerged. Of the 44 subjects, 16 said that both the harming and the helping were intentional, 14 said that neither was intentional, and 14 said that the harming but not the helping was intentional. (This is the finding that impressed me.)

The combination of these two new findings suggests a hypothesis: There are at least three different concepts (or conceptions) of intentional action in our communities, and in two of them knowledge plays some roles that it doesn’t play in the third. (NU suggest a related hypothesis.) The hypothesis can be made more precise and then tested. (Fiery Cushman and I are thinking now about tests.) In any case, partly in light of NU’s finding, I do wonder whether Knobe’s studies get at something that we should think of as “the folk concept” of intentional action or whether there are several different concepts of intentional action out there that collectively account for the data. (BTW, there is also evidence of at least two concepts of *intention* in our communities, in one of which knowledge plays a role that it doesn’t play in the other.)

4. “The overwhelming emotion asymmetry”

I wonder whether part of what is at work here is an “in-character” / “out-of-character” asymmetry. Perhaps quite a few people are tacitly thinking that good actions done “because of . . . overwhelming and uncontrollable sympathy” are done by good, sympathetic people whereas bad actions done “because of . . . overwhelming and uncontrollable anger” typically are out of character for the agent. One way to try to get evidence about this is to tweak the first scenario along the following lines:

Jack had never had even a streak of kindness in him. He had always been a selfish and entirely unsympathetic man. But one day, overwhelming and uncontrollable sympathy suddenly came over him, and because of that Jack impulsively gave a homeless man his only jacket even though it was freezing outside. Jack never did anything remotely like that again for the rest of his life.

Of course, there’s no reason for DKW to be unhappy with an “in-character” / “out-of-character” asymmetry, but it would be good to have more evidence about what lies behind “the overwhelming emotion asymmetry.”

5. “The negligence asymmetry”

DKW write: “responsibility attributions are often associated with determinations of restitution; since more severe accidents may call for greater restitution, it arguably makes good sense for people to be less stringent in their standards for responsibility in cases where the harm is severe than they would be in cases where the harm is mild.” Wouldn’t it make more sense to have the same standards for responsibility and to make restitution a function of a combination of degree of responsibility and degree of harm?

6. Causal judgments

In section 3, I suggested that there may be more than one concept of intentional action (and intention) in our communities. The same may be true of causation. This is testable.

The pen story is great. But what were subjects asked? Did they have the option of saying that both caused the problem and the option of saying that neither caused it? If so, how large were the minorities: in particular, those who say that both caused it and those who say that neither caused it? (Someone trying to screw up the experiment might have given a third minority response: that only the assistant caused the problem.)

I have raised a lot of questions – more than I would have raised in settings in which I had the burden of offering some answers! But my loose approach seemed to me appropriate for starting the discussion. I hope you found “Variantism about Responsibility” as stimulating as I did.

Cheers,
Al Mele

Comments

[Neil]

There are two ways to respond (critically) to papers like this. One is Al's way: to look for faults in the design and the interpretation of particular experiments. Al's suggestions are certainly fruitful, and deserve discussion. But I'm going to suggest a different way of criticisng the paper. Rather than focus on the details of the experiments, I'm going to step back and examine the methodology more generally. I apologise in advance for the length of this comment.

The general idea is to try to defend reflective equilibrium (RE), with all its intuition-pumping, against the attack presented here. Most of the debate over RE has occurred in political philosophy, where there is a standard accusation levelled at it: because it is rooted in folk responses, it is inevitably conservative. There is also a standard response: that though RE must begin from folk intuitions, it has the resources to explain and to justify radical revision of these responses. Rebuilding the ship at sea does not commit us to following its original design. How can RE motivate radical revision? By way of tensions and conflicts within folk morality. If there are such conflicts, then we have the resources to move far beyond our starting point.

RE, between intuitions and theories that systematize them, also characterizes the method of cases that Doris, Knobe and Woolfolk are concerned to criticize. They hold that RE commits traditionalists to conservativism about folk judgments. They are themselves conservatives, but they hold that the commitment to conservativism conflicts with invariantism, which is a common assumption these philosophers make. Folk morality is itself variantist; we therefore cannot preserve its major features while maintaining the invariantist commitment.

Now, there are two responses we can make to Doris, et al's attack upon RE. The first, less radical, response is to point out that it has the resources to motivate revisionism. That is, there are conflicts in folk morality which can be used to motivate a more-or-less radical revision in responsibility ascriptions. I suspect that Doris et al. will accept that claim. Nevertheless, they will deny that this kind of revision is attractive. Why? Actually, I don’t have any clear idea why they think this. They say plenty of things that seem to address the issue, but – so far as I can tell – they seem to content with arguing that we don’t have any decisive reason to prefer revisionary invariantism to conservative variantism.

But there is a more radical response available: you just can’t have conservative variantism. Why? Because RE (here) does not merely provide us with the resources for radical revision, it requires it. Folk morality does not only contain tensions and conflicts, it also contains a fairly deep commitment to the resolution of these conflicts. And the evidence gathered together in this paper actually manifests that commitment.

It does not manifest the commitment directly. Instead, it is built into the study designs. The studies all use a between-subjects design. That is, similar vignettes are never give to the same subjects; instead, vignettes which differ only (say) in the moral valence of the effect, or the moral seriousness of the result, are given to different groups of subjects. This is a standard procedure in psychology: we look for statistically significant differences between groups, control for various explanatory factors and reveal the contours of folk belief. Nothing wrong with the approach, either in general, or as applied here. It really does show that folk morality is (superficially) variantist. But notice the justification for using the between-subjects design: were we to give the vignettes to the same subjects, they would have revised their judgments to make them internally consistent.

Now, there is nothing abstruse about noticing a conflict in our judgments, nor in having such conflicts pointed out by others. Those political philosophers (like me) who have argued that reflective equilibrium can motivate and justify radical revisions of political beliefs hold that this is precisely how moral progress occurs: a Martin Luther King, or a Gandhi, or the leaders of second-wave feminism, come along and point out that the practices, policies and judgments that characterise a society are inconsistent. They argue for a way of resolving the conflict. They remain within folk morality when they do so.

Doris, et al. cannot have conservative variantism because they can maintain the variantism only so long as they very carefully avoid drawing the folk’s attention to it. In other words, they can’t conserve the full range of folk beliefs and commitments; they can’t even accommodate one of the most deeply felt commitments, the commitment to consistency. If that's right, then folk ascriptionists of responsibility are not deeply variantist, and we must opt for revisionism, rather than conservatism.

[Joshua Knobe]

Al,

Unfortunately, it simply wouldn't be possible for me to respond to all of the interesting comments and suggestions you offer here, but I did want to ask a quick question about one aspect of your view.

At one point, you suggest that Shaun Nichols and Joe Ulatowski might be on the right track in their superb recent paper on intentional action (now available at http://www.unc.edu/~knobe/Nichols-Ulatowski.pdf). Specifically, you suggest that there might actually be three different concepts of intentional action.

I wanted to ask what implications you think this view might have for the study of moral responsibility. The ordinary view would be that people always use the concept of intentional action when they are trying to determine whether or not someone is morally responsible. So would you suggest that people are actually using three different concepts here, depending on the nature of the case?

Neil,

I appreciate your concerns here, and I can see how they might apply to the claims we make about determinism (where we suggest that people literally have logically inconsistent intuitions). But I wonder if you would say that this same problem arises for all aspects of our view.

There is no logical inconsistency involved in the side effect asymmetry or the overwhelming emotion asymmetry, and both of these effect arise within-subject. Similarly for the asymmetries in attributions of causation and intentional action.

So I wonder if you might agree to a compromise position -- revisionism for the logically inconsistent intuitions about determinism, conservativism for the remaining asymmetries.

[Neil]

Hi Josh,

I've actually sketched a conservative account of some folk intuitions in a forthcoming Phil Psych paper. So I certainly don't want to reject your suggestion out of hand. Here's how I want to respond. When we do reflective equilibrium, we should bring to bear *all* relevant intuitions (where the test of relevance is: actually makes a difference to the results) and *all* relevant beliefs. So, for instance, we bring to bear all considerations from the heuristics and biases literature, and all relevant statistical knowledge, and so on, and then ask our subjects whether, once they have fully absorbed all this information, they are still willing to hang on to the judgments.

For instance, there's no obvious contradiction in judging according to the conjunction fallacy. But once we explain to subjects that conjunctions can't be more probable than their conjuncts, the subjects will retract the judgment (or we need to get ourselves a new set of subjects!) So we can't just limit ourselves to contradictions, in doing reflective equilibrium. In short, reflective equilibrium needs to be as wide as possible.

Of course, the wider we make it, the less empirically tractable it gets.

[Kip Werking]

Some comments:

1. I agree with Neil that there are two styles/approaches available here. I found his comment to be very helpful and insightful.

2. Of course, I approached this paper from a more non-realist perspective (about mr/fw/autonomy). And, with that in mind, I want to emphasize that the distinction between conservativism and revisionism does not seem so clean and neat (at least to me). Non-realism about free will or moral responsibility is supposed to be a revisionist view. But it is conservative at least in the sense that it *based* upon higher-order intuitions which just happen to conflict with lower-order intuitions. Indeed, the folk belief that "most people, most of the time, are morally responsible for their decisions" is the *only* folk belief that, to my knowledge, hard determinism, etc., revise. Similarly, orthodox or realist views seem *revisionist* to the extent that they ignore these other intuitions (e.g. incompatibilism is a notorious intuition that some allege the folk to have). In other words, it is not as if hard determinism, etc., sprang out of nowhere, completely unhinged from folk beliefs. So to say that "[conservativism is] the view that folk belief is a constraint on philosophical theorizing" (DKW, p. 4) but that hard determinism, etc., is not conservative (perhaps not at all) seems not quite right.

Let me give one example to drive the point home: as best as I can tell Frankfurt, Fischer, Dennett, Mele, and Watson all believe that if God (or whoever) set up a deterministic universe such that it would unfold exactly according to God's (or whoever's) plan, including his detailed plan for one's individual life, then most humans in that universe, most of the time, will still have free will, or at least be morally responsible for their actions. Now, this is a bullet to bite and it strikes me that if the philosophical notion of free will was supposed to be incompatible with anything, it would be incompatible with *that*---that is why Watson calls views which accommodate this possibility "hard" compatibilist views (and so all prominent compatibilists, to my knowledge, are "hard" compatibilists, although they do not call themselves that).

So it is not as if orthodox views (or at least compatibilist views) preserve all of the folk beliefs that non-realist views do---they do not. For example, it is not the case that realist views preserve N folk beliefs and non-realists views preserve the same set of N beliefs minus one (or some lesser number). Rather, they may both be revisionist in different ways, even if the orthodox revisionism is more subtle. Now the question becomes: in this battle between inconsistent intuitions, where do our loyalties lie?

I think it's very important to consider that, if both sorts of views are revisionist, but in different ways, *how* do they differ? First, let's label some intuitions:

Non-realists revise the folk belief MR ("most people are most of the time MR").

Orthodox views (or at least compatibilist views) revise the folk belief GD ("people cannot have freedom or responsibility if God designed the entire universe such that their life would unfold according to his plan")

Of course, whether the folk actually believe GD is controversial (I know of no data on this, but I would expect this "hard" position to receive little support and I cannot imagine a better question to ask the folk). But one might substitute TNR, or some less controversial premise. There should be at least *some* controversial premise that orthodox (or at least compatibilist views) revise, or else they will seem to owe us an error theory for why so many intelligent people made mistakes about what should be obvious.

Then one difference that seems important to me is this: if we are setting out, philosophically, to determine whether most people are, most of the time, morally responsible (or have free will, or are autonomous, or whatever), then MR seems much more like a *conclusion*, and GD (or TNR, etc.) seems much more like a *premise*---even if it itself is based upon underlying premises. Let's call this idea, that MR and GF (or whatever) are different in this way, PR.

I'm not sure if PR is true, although I suspect it is, or something like it is. Then what I would want to ask is this: is there nothing to be said for the notion that, when evaluating certain conclusions, we should give more weight to less conclusory premises and less weight to more conclusory ones? In other words, might we expect the chance that the folk get the concepts wrong to be *less* than the chance that they get conclusions-based-on-those-concepts wrong?

Let me give an example (considering that, as Tamler has noted, many philosophers who have given up on God a long time ago still cling to belief in fw and mr): might we expect the chances that the folk get the concept of God right to be *greater* than the chances that the folk get the conclusion "God exists" to be right? Might we have more confidence that, if the folk say God is a bunch of attributes (omniscient, omnipotent, omnipresent, worthy of worship, omni-benevolent, etc.) they are likely to be right (in the sense that the term "god" actually has that common usage); but if the folk say that "God exists" (and so an omniscient, omnipotent, omnipresent, worthy of worship, omni-benevolent being actually exists), then they may very well be wrong---or at least our confidence in this folk belief should be less than our confidence that the concept is right, and less because the folk are talking more about conclusions than premises?

[It is not obvious that GD is more closely related to concepts that MR is, and so it might be better to just ask the folk “what is free will and what does it involve?”]

Let's call this premise, that we ought to give more weight to less conclusory folk beliefs, WEIGHT. Then:

PR + WEIGHT would favor non-realist revision over orthodox (or at least compatibilist) views.

But, of course, whether either PR or WEIGHT (or both) are true remains to be seen.

I began by arguing that the distinction between revisionist and conservative views might not be so clean and neat. DKW seem to acknowledge as much in at least the following two sentences:

"On such models, folk morality is not, of course, inviolate; just as reflection on folk morality may constrain philosophical innovation, theoretical innovation may compel revision of folk morality. Nevertheless, it is very often asserted that an ethical theory which is sharply at odds with folk morality is at a competitive disadvantage when compared with a theory that is not."

This led me to argue that non-realist views are not so revisionist as some might suppose. But suppose that they are largely revisionist. Would that be so bad?

Note that DKW have not worded the last sentence of that quote in terms of endorsement. The sentence says that *others* regard revisionism as a competitive disadvantage and not that the authors do. Indeed, the authors seem to limit themselves to arguing that invariantism and conservatism are inconsistent. In either case, if one is considering revisionism about certain folk conclusions, as opposed to concepts or premises, then isn't this disdain for revisionism an assumption in favor of human rationality? This assumption should be controversial considering the ongoing dispute about human rationality, and the wide body of literature showing the many ways in which human beings are systematically irrational.

Consider the following passage from Pereboom’s Living Without Free Will:

“At this point, the compatibilist might argue that according to ordinary intuitions, Plum in Case 4 is morally responsible, and that these intuitions should be provided more weight than we have given them. But in the incompatibilist view, one consequence of determinism is that ordinary intuitions about moral responsibility in specific cases are based on a mistake.”

[But one might also consider that Thomas Jefferson wrote: “State a moral case to a ploughman and a professor. The former will decide it as well, and often better than the latter, because he has not been led astray by artificial rules.”]

The disdain for revisionism seems to leave little room for the possibility of views like Pereboom’s. But certainly it *is* possible that the folk could make the sort of mistake Pereboom describes. I don’t see why this possibility should be disfavored a priori, especially when it has been contested so hotly, and the evidence in favor of systematic human irrationality, at least in certain ways, is robust. If someone suggests a revisionist view, as opposed to a more conservative one, then let’s examine the arguments and the evidence. Just the fact that a view is revisionist, without consideration of the arguments and evidence, should not already count against a view. Views like Pereboom’s should at least have a fighting chance.

In summary, I’m not sure that non-realism is as revisionary, in contrast to orthodox (or at least compatibilist) views, as some might otherwise think. If one focuses upon how both non-realist and compatibilist views are revisionary, then we can sort out the battle-between-two-revisionisms and decide where our loyalties lie with respect to competing intuitions. But even if it is more revisionary, I’m not sure that counts against it.

[Al Mele]

Joshua asked about a hypothesis I mentioned: “There are at least three different concepts (or conceptions) of intentional action in our communities, and in two of them knowledge plays some roles that it doesn’t play in the third.” Here’s an example of what I have in mind. Suppose studies reveal that the following three groups exist in our communities:

1. Those (40% of the population, say) who consistently make judgments about intentional action as someone who accepts the following principle would:

P1. Whenever an agent who A-s is not seeking to A, he does not A intentionally.

2. Those (30%, say) who consistently make judgments about intentional action as someone who accepts the following principle would:

P2. Whenever an agent knows that if he A-s, he will, as a consequence, B, then if he A-s intentionally, he B-s intentionally.

3. Those (25%, say) who do not fall into group 2 and who instead consistently make judgments about intentional action as someone who accepts the following principle would:

P3. Whenever an agent knows that if he A-s, he will, as a consequence bring about some morally bad effect B, then if he A-s intentionally and is indifferent to bringing about B, he brings about B intentionally.

Suppose (to keep things simple) that the same studies reveal that when side-effects are not at issue, all three of these groups agree.

We can say that the folk concept of intentional action can be read off from the majority responses of this population. Of course, we’d get the asymmetry for the chairman questions; and we can say that the folk concept of intentional action accounts for the observed asymmetry. Or we can say that groups 1 and 2 have conceptions of intentional action that yield no such asymmetry, whereas group 3's conception does yield this asymmetry. (Group 1 would count both harming and helping the environment as “not intentional”; group 2 would count both harming and helping the environment as “intentional”; and group 3 would count the harming but not the helping as intentional.)

BTW, Fiery Cushman and I conducted a pilot study (70 subjects; each was given the same set of 13 stories, including the two chairman stories) in which things go pretty much this way. We located the subjects who gave the “not intentional” + “not intentional” response, the “intentional” + “intentional” response, and the “intentional” + “not intentional” response to the chairman questions, and we found that the first group was group 1 on the other side-effect stories (all of which were morally “neutralish”), that the second group was group 2 on the other side-effect stories, and that the third group agreed with group 1 about the other side-effect stories. We also found that all three groups agreed about all but one of our non-side-effect stories. So far, it looks like the three conceptions pretty much share a sufficient condition for a non-side-effect action’s being intentional and differ about side-effect actions in the way I described. I’m guessing that further tests of non-side-effect stories will turn up some modest differences among the groups.

[doris]

Thanks to everyone for their interest! Special thanks to Manuel for organizing, and Al for his challenging comments.

Two posts of (some) substance to follow shortly; hopefully,more to fllow next week.

jmd

[doris]

Al raises sensitive and important points about the experiments DKW is built around. Since Woolfolk, Doris and Darley is closest to my heart, I’ll respond to some of Al’s remarks on it. Al’s points in quotes, my ruminations following.

“B. A1 is true, A2 is false, and Bill has some responsibility for killing Frank because he would have killed him even if the bad guys hadn’t made him do that.”

I wish we had included a probe regarding the counterfactual; I agree that the subjects may have been thinking along these lines. Suppose they were: i.e., the presence of identification indicates to subjects that the counterfactual is true, and the judgment about the counterfactual is what drives the responsibility attribution. This still seems to me a striking result. Note, in particular, that it suggests that subjects might be happy enough attributing moral responsibility when they do not believe that the actor was causally implicated in the relevant outcome. Thus, I think the answer to the following question of Al’s --

“Possibly, suggestion B has led some of you to wonder whether some subjects would say that Bill has some responsibility for killing Frank in a strange story in which it is clear that Bill does not kill Frank. Wait! Am I nuts?”

-- is “No, you’re not nuts. That’s very likely what some subjects would say.”

But a bit later, Al and I diverge:

“Well, given that these rational subjects know that Bill did not kill Frank (even though Frank was killed), they cannot be expressing the belief that he has some moral responsibility *for killing* him. They are doing something else with their words.”

I’m not sure that the folk view of moral responsibility in every instance presupposes causal responsibility. I’m also not sure – although I’m pretty tentative here – that it is mistaken (or as Al might say, a violation of some rational requirement) to decouple moral and causal responsibility. Needless to say, I owe a huge-ish argument here; I guess I’d begin by thinking about collective responsibility. But first, as a conservative, I’d like to get clearer on the folk view. I predict that subjects can be induced to attribute moral responsibility in the *clear* abscence of causal responsibility. Any takers?

A final note: WDD have consistent results (to date, not written up), with the prosocial behavior of kidney dontaion and the anti-social behavior of a My Lai-type war crime. These findings are in a sense less “philosophically interesting” than that reported in DKW, since they involve coercion that may not be thought to reach the “could not have done otherwise” threshold, but they do indicate that identification effects can be readily obtained.

jmd

[doris]

As Neil notes, one might critique a piece like DKW experiment by experiment (as Al does), or by forwarding general methodological or theoretical arguments (as Neil does). I’ve said on numerous occasions (e. g., 2002: Ch. 6; Doris and Stich 2005; Doris and Plakias forthcoming) that I’m generally suspicious of general methodological/theoretical arguments, but Neil’s concerns are serious ones, so I’ll try to make something of an answer.

I should say that I did not understand DKW to make, or require, an argument against RE, which is in fact a methodology I tend to endorse. But not much turns on this here.

The rub comes when Neil says,

“Folk morality does not only contain tensions and conflicts, it also contains a fairly deep commitment to the resolution of these conflicts.”

This is, of course, an empirical claim. I’m not sure what evidence there is there regarding this, although I might start by looking to the material on East/West cultural variation regarding “dialectical thinking” (which would likely complicate any story we tell). My sense – unencumbered by evidence – is that folk morality (at least around here) might not be all that conflict averse. Oftentimes, people seem more committed to their fairly particular commitments – say “pro life” and pro bombing civilian areas – than they are to any general ethical principle prohibiting the co-maintenance of the commitments. (Innact, my sense is that charges of inconsistency often earn the response of a shrug.) Thus, I’m inclined to think the following move in Neil’s argument is undersupported:

“Doris, et al. cannot have conservative variantism because they can maintain the variantism only so long as they very carefully avoid drawing the folk’s attention to it. In other words, they can’t conserve the full range of folk beliefs and commitments; they can’t even accommodate one of the most deeply felt commitments, the commitment to consistency.”

Now it may be that folk morality is intolerant of flagrant contradictions – ones fairly transparently of the A&-A variety. But as we try to say in the paper, and my estimable co-author JK has already said here, many of the variations we identify are obviously not of this sort. Indeed, it is not obvious that they are tensions of the less flagrant sort, such as those manifested by the “pro-life” strategic bomber.

Then I don’t see that our conservative variantism is subject to the general difficulty Neil adduces. So far, I see no reason to go more revisionist than the compromise solution JK suggests here. But exactly how revisionist we go – and here’s *my* general methodological commitment – can only be decided by detailed consideration of the many and various particular cases.

jmd

[Al Mele]

John writes: “But a bit later, Al and I diverge: ‘Well, given that these rational subjects know that Bill did not kill Frank (even though Frank was killed), they cannot be expressing the belief that he has some moral responsibility *for killing* him. They are doing something else with their words’ [The embedded quotation is from me, Al. Now John again.] I’m not sure that the folk view of moral responsibility in every instance presupposes causal responsibility. I’m also not sure – although I’m pretty tentative here – that it is mistaken (or as Al might say, a violation of some rational requirement) to decouple moral and causal responsibility.”

Two issues need to be separated here.

1. Can anyone (rationally) believe the following: Joe is morally responsible for the fire in his garage and there was no fire in his garage? No. What’s the general principle at work? This. No one can (rationally) believe the following: S is morally responsible for X’s happening and X did not happen. In "a strange story in which it is clear that Bill does not kill Frank," Bill’s killing Frank did not happen. And no one can (rationally) believe the following: Bill is morally responsible for Bill’s killing Frank and Bill’s killing Frank did not happen (i.e., Bill is morally responsible for killing Frank and Bill did not kill Frank).

2. Can anyone (rationally) believe the following: Joe is morally responsible for the fire in his garage and Joe is not at all causally responsible for the fire in his garage? (This time there actually was a fire in his garage.) Maybe so. Maybe someone who says this about Joe (rationally) believes that although Joe did not cause the fire, he intentionally omitted to put it out and therefore has some moral responsibility for it. This person may (rationally) believe that omissions in general are not causes.

Puns about not following me are welcome.

[Neil]

John points out that my claim about folk morality is empirical. He further claims that I don't have any evidence for it. But there's plenty of evidence. Consider, for instance, Hauser, et al's forthcoming *Mind and Language* paper: only 5.8% of subjects generated inconsistent responses when a within-subjects design was used. Or consider Haidt's moral dumbfounding: the fact that people are discomfited by conflict between their principles and their judgments is empirical evidence for my claim.

Suppose people weren't discomfited by this kind of conflict. This would just be more evidence for one of my beliefs: the folk are really confused. As it happens, they're not as confused as all that.

[doris]

Neil:

What I said was:

"This is, of course, an empirical claim. I’m not sure what evidence there is there regarding this . . ."

I don't think this imples anything as confident, or harsh, as:

"He further claims that I don't have any evidence for it."

I was expressing the suspicion that the evidence re felt consistency pressures is limited, a suspicion I'm still inclined to cling to, despite the evidence referenced in your latest post. I don't know the Hauser piece, but doesn't the dumbfounding stuff admit of a reading friendy to my view? I'd be inclined to take it as evidence of a willingness to *resist* consistency pressures and cling to particular judgments, much the sort of situation I had in mind in my previous post.

In any event, although the empirical questions regards consistecy pressures are important, I doubt they are especially telling re our variantism, without strong evidence to the effect that the variations we adduce typicaly run afoul of a consistency constraint -- something I, with my coauthors, continue to doubt.

In the end, my hunch is that our deep disagreement is not about consistency, but about the attractions of conservativism.

jmd

[Tamler Sommers]

I agree that the dumbfounding cases might support John's view since the subjects refuse to change their mind even after all their arguments are shot down. They'd prefer if their moral beliefs were consistent but seem willing to allow for inconsistency (or at least more willing than they are to say it's OK to sleep with your sister.)

On the other hand, some experiments by David Pizarro might support Neil's side. Pizarro designs some trolley cases (and other dilemmas) with outcomes that are more or less desirable depending on the subject's political orientation. He gets some cool results showing how inconsistent we can be when appealing consequentialist or deontological principles. For example, liberals are far more inclined to push Chip Ellsworth III from the footbridge to save the Harlem Jazz Orchestra than they are to push Tyrone Jackson to save the New York Philharmonic. (As an aside, that's far and away my favorite use trolley cases, without a close second.) And they justify whatever they choose with the ethical theory that fits. I think Pizarro made the analogy of a 'toolbox' of moral principles in one of his talks.

However, when subjects are given both scenarios, one following the other, they force themselves to be consistent even if it means attaining the undesirable outcome. So if you give liberals the 'Chip Ellsworth' scenario first, and they say 'push the guy,' they'll say the same for Tyrone Jackson--even though it is unlikely that they would have pushed Tyrone if they hadn't first heard about Chip. So the subjects do seem resistant to being blatently inconsistent in their use of moral principles.

Hope this is relevant. In any case, you gotta love the Chip Ellsworth III vs. Tyrone Jackson manipulation

[Neil]

For what it's worth, Haidt himself seems to take my side in the consistency wars... or so I read his latest paper.

http://faculty.virginia.edu/haidtlab/articles/haidt.bjorklund.social-intuitionists-reason.doc

[Joe]

I'm pretty busy and HAVE NOT been following any of this debate closely BUT something that Al said (Aug 24, 2006 12:59:19 PM) in response to John caught my attention. First, though, I want to make it clear that I had nothing to do with that fire in the garage!

In his discussion of responsibility, H.L.A. Hart develops an interesting concept he calls 'role responsibility.' For example, Hart claims that the captain of a ship might be responsible for the fact that some cargo is thrown overboard (I think this is the example) simply because he is the captain of the ship.

If the concept of role responsibility makes sense, then this MIGHT favor John's view over Al's (provided that I understand them both correctly). For instance, slightly altering the above example, we might say the Captain Joe IS morally responsible for the fire on the ship AND Captain Joe is not at all causally responsible for the fire on the ship. This strikes me as a different kind of case than the one that Al was discussing.

I do not like this consequence and tend to be on (what I take to be) Al's side of the debate. But I also think that there is something intuitive about the concept of role responsibility. What to do? I have my own suggestions but I want to know what the rest of you think before I'll reveal them!

This is a great idea by the way! Some of the best discussions in the Garden in a while!

[Neil]

When I'm not doing free will (of my own volition) I am compelled to work in applied ethics. Applied ethicists LOVE role responsibility. I've always been a bit sceptical. Obviously, responsibility attribution has to be sensitive to features of an agent's context. What did the agent have control over? What did he take himself to have control over? Those are, very roughly, the control and epistemic conditions of moral responsibility. Role responsibility seems either to be an inflation of this trivial truth, or just plain wrong (when we say an agent is responsibility for X, because a good Y has relevant control over X-ing, and even though this agent did not satisfy one or both the relevant control or epistemic conditions).

[Al Mele]

In the post of mine to which Joe is referring, I had no quarrel with role responsibility. Suppose, regarding the weird case I was discussing, that Fred (a rational respondent who is not going out of his way to be deceptive in responding to an item on a questionnaire) gives a “yes” answer to the question “Is Bill morally responsible for killing Frank.” Now, Fred knows that Bill did not kill Frank. He knows that the following has happened (from my first post): “Bad guys paralyze [Bill’s] arms and hands, put a gun in his right hand, raise his right arm (by electronically stimulating his arm muscles) so that the gun is pointed at Frank’s head, and (by further electronic stimulation) cause Bill’s paralyzed right index finger to depress the trigger.” Fred also knows (from my first post) that “‘Bill was certain about his feelings. He wanted to kill Frank.’ What’s more, Bill thought that he was acting: he did not realize that his arm and hand were paralyzed. And he felt no reluctance about blowing ‘his friend’s brains out’.”

OK. So what does Fred mean to communicate? Not literally that “Bill is morally responsible *for killing Frank*.” For he knows that Bill did not kill Frank.

And now we’re pretty close to the punch line. Recall (from my original post) that “In Woolfolk’s story, the ‘drug makes [Bill] unable to resist the demands of powerful authorities,’ and they order Bill to shoot Frank in the head. So it certainly seems that they cause Bill to shoot Frank. Even so, ‘subjects judged the high identification actor more responsible . . . than the low identification actor’ (p. 13).”

This paragraph is the punch line. We’d get some evidence about what some subjects who answer “yes” to the question “Is Bill morally responsible for killing Frank?” in Woolfolk’s story might mean by running my modified version of his story. If a significant percentage of people answer the question about the modified story as Fred answers it, then some people who answer “yes” to the question about Woolfolk’s story might mean something pretty similar to whatever it is that Fred means. And Fred does *not* mean that Bill is morally responsible *for killing Frank*. (What Fred might mean is left as an exercise for the reader.)

(For anyone who might think that I’m suggesting that Bill did not kill Frank in *Woolfolk’s* story, I add the following: Not so.)

[Dan Speak]

Excellent paper, gentlemen, and terrific discussion too! I'm learning a lot from the interaction. That I have more to learn is going to become clear shortly.

I'm wondering if there is conceptual space to maintain conservative invariantism in the following way. Suppose that the folk view is two-tiered. The bottom tier involves a commitment by the folk to the preservation of our blaming/praising/accountability practices. The top tier expresses what I'll call the “normative depth” of these practices. It is from the top tier that we are able to draw out various incompatibilist responses from people. The bottom tier generates the compatibilist responses. Strictly speaking, the two tiers are not representing fundamentally inconsistent positions. This is because the very same practices can be understood to depend on moral concepts of greater and lesser depth. Now it would be natural for the folk to confuse the applicability of a certain moral concept with its semantic and moral value. It would be the philosopher's job to pull these apart and clarify the issues. But the philosopher could, thereby, be demonstrating that the folk view is neither deeply confused nor variant. Couldn't she?

[Joshua Knobe]

I am truly amazed at the depth and helpfulness of all these comments! Manuel certainly deserves a standing ovation.

Unfortunately, it just isn't possible to address all of the interesting ideas people have brought up, but I do want to say a few words about two of the ongoing discussions.

Neil,

I don't think John meant to suggest that people aren't at all interested in consistency. Rather, his point was that there simply isn't any inconsistency in the particular phenomena we are discussing here.

Take the overwhelming emotion asymmetry. This is a tendency for people to regard overwhelming emotion as diminishing blame for bad behaviors but not diminishing praise for good behaviors. The key question is whether this phenomenon involves any sort of inconsistency at all. It does involve treating judgments of blame differently from judgments of praise, but one could simply argue that blame truly is different from praise and that the differing criteria are therefore justified. So if there is some kind of inconsistency here, it would certainly have to be a very subtle, complex sort of inconsistency -- one that is not immediately apparent on considering the intuitions themselves.

Al,

That experiment you did with Fiery is fantastic! I think it really illustrates the promise of experimental philosophy. In particular, it shows how much we can learn by looking not only at the intuitions of the majority but also at the intuitions of the minority.

The one question I wanted to ask was whether you see this experiment as providing evidence for your view that there are three different concepts involved here. I mean, suppose that someone looked at the preivously available evidence and ended up opting for one of the many other theories (say, one of the theories according to which there is only one concept of intentional action). Do you think that your new results would provide some reason for such a person to switch over to the view that there are actually three concepts? If so, how exactly would the argument work?

[Neil]

Josh, I agree that there's no obvious conflict involved in the overwhelming emotion asymmetry judgments. I think that there are other other conflicts in other cases, though, and that therefore you can't have conservativism variantism with regards to other factors. Moreover, the kind of variantism you can have will be mild, because in the absence of conflict variantism isn't exciting.

Question: suppose that a within-subjects design made the overwhelming emotion asymmetry vanish. What then? Why should we prefer any variantism (however mild) over my revisionism?

[Al Mele]

Joshua,

One thing that makes the issue difficult, as you know, is that there is considerable disagreement both about what concepts are (or what “concept” means) and about what conceptual competence is. But here’s a question. When we find different patterns of the kind I mentioned involving three sizeable groups in a respondent population (groups summing to about 96% of the population), if we say that we get at the folk concept by averaging across the whole population of respondents, what are we likely to mean by “the folk concept”?

In a 2003 paper in *Phil Psych*, I wrote: “By ‘the folk concept of intentional action’ here I mean a concept of intentional action that I will suppose to be reflected in the judgments a substantial majority of Knobe’s respondents make about stories they are given – judgments that certain actions described in those stories are intentional, unintentional, or not intentional.” This was a device for side-stepping some technical questions and moving the paper along. But if I had seen data about individuals’ responses to a battery of stories that displayed the patterns that Fiery and I found, I might have said something like this: “By ‘a folk concept’ of intentional action here I mean a concept of intentional action that I will suppose to be reflected in the judgments a sizeable group of respondents makes about stories. . . .” Which would be a more productive way to proceed? And how should we go about trying to answer the preceding question?

[Paul Torek]

Thanks for bringing us this forum, Manuel and DKW. Now, how did the discussion get this far without mention of this gem:

' Instead of the normative status of the behavior or outcome determining the appropriateness of praise or blame subsequent to an attribution of responsibility based on evaluatively neutral causal and psychological attribution, responsibility attribution is from the outset deeply infused with normative considerations: it is, one might say, normative “all the way down.” '

To this passage I have nothing to add - except my enthusiastic agreement! But I wonder, doesn't this point sound, not so much "variantist", but, well, unifying? Which brings me close to my main worry, whether "variantism" is actually a proposition.

Moral "theories" come in a wide variety of degrees of complexity. Oversimplifying a little, they form a continuous spectrum (oversimplifying less, a well-ordered continuum might not quite fit, but let's not go there). What sense, then, in drawing an arbitrary line, even a fuzzy one, and saying, over there, invariant, over here, variant?

Suppose that we have to admit a fundamental asymmetry between responsibility for good vs. bad actions. So, our account of morality gets a little more complicated. But, to quote a famous philosopher, "A ... theory should be as simple as possible, but not simpler". It seems rather obvious to me that if you are willing to approximate moral accounts as lying on a continuum as I have proposed, there is little point drawing any particular lines in this sand. Therefore, I suspect that the authors must have some other perspective in mind, from which to separate variantist from invariantist accounts. But what? And isn't mine a better perspective from which to view this (I mean aside from the excellence that it is, after all, mine)?

[Joshua Knobe]

Paul,

This is an excellent point, and we wholeheartedly agree. In fact, my esteemed co-author John has often emphasized this point in talks he's given on the topic.

Neil,

I feel like we may gradually be reaching a kind of convergence here, perhaps arriving at a view that involves some revision and some conservation.

A variantist theory is just a theory that applies different criteria in different cases (e.g., one criterion for morally good actions, another for morally bad actions). There's nothing in the very idea of variantism that involves a commitment to embracing inconsistent views.

Of course, it might turn out that closer inspection of the intuitions we've described will reveal certain subtle inconsistencies, and that sort of discovery would definitely provide good reason for the revisionism you suggest. But the problem would have to be discussed on a case-by-case basis; it's not as though there is some general fact about variantism itself that makes it always yield inconsistent views. So after all of the inconsistencies have been eliminated, we may still be left with a view that shows a fair amount of variantism.

Al,

It seems as though people's *concepts* of intentional action are just one of the many factors that affect their responses on these tasks. So if we observe that people give different responses, there is always an open question as to whether they have different concepts or whether they only differ in some other respect.

For example, the Nichols-Ulatowski theory is that people all have the same underlying concepts but that they differ in the interpretation they assign to the English word 'intentionally' -- with some people being more inclined to map this word onto one concept, some being more inclined to map it onto another.

In a variety of other ways, I suspect that Alicke, Adams & Steadman, Nadelhoffer and others would also say that people's underlying concepts here are the same but that some additional factor leads to the observed differences in responses.

Now, I do think it is plausible to suggest (as you have) that the differences in responses actually are due to differences in the underlying concepts. But I would love to hear a little bit more about what reason we have to accept this view rather than one of the others. Is there some specific reason to suspect that people's concepts actually differ here?