Blog Coordinator

Knobe's X-Phi Page

X-Phi Grad Programs

« Knowledge In and Out of Contrast | Main | New Papers on Moral Cognition »



Feed You can follow this conversation by subscribing to the comment feed for this post.

Eddy Nahmias

Geoff, this looks like a fascinating result. I don't know about others, but I typically cannot assess (or understand) the import of an x-phi study unless I see the materials (scenarios and questions). Can you provide more details? And if you feel like it, a bit more on what DRC is would help to (is that your own term or some existing term?)
(Anyway, that's, like, my opinion, man.)

Geoff Holtzman

Hubris begets typos. Eddy, "DRC" is what I get for making up an acronym on the spot- I've edited the post so it now reads "DRP" (thanks), and yes, it's my own term. The idea is that "moral responsibility" is ambiguous between two concepts, one of which is dependent on, and one of which is independent of, moral content, and that people recognize which of these uses they tend to apply.

First, I think the question of competence is independent of the question of factual accuracy. Most people who say, "I hate prunes because I don't like dried fruit" aren't committing a performance error; they've either committed a malapropism, or they've committed no error at all. That debate is irrelevant here- either way, there's no need to invoke affect to describe their decision process. Google Goggles probably makes the same mistake when it comes to fresh prunes.

Second, I think malapropisms with "moral responsibility" are certainly possible, but that neither the moral-content-dependent the or moral-content-independent approaches reflect a competence error. Say I (a man) have sex with Mrs. Smith's son, and Mrs. Smith becomes irate that I would do such a thing. Am I morally responsible for having sex with her son? On the reading in which moral content is essential to moral responsibility (you can't be morally responsible for something entirely amoral), I'm not any more responsible for an act like this than I am for wearing headphones right now. But on another reading, I am morally responsible: should sleeping with her son be a betrayal of a close friend, or should her son or I already be in a committed monogamous relationship, or should I have have a transmissible illness that I'm aware of, accountability would fall to me. This way of using "moral responsibility" to describe a kind of liability does not seem right to me, but I know people who are convinced it is right.

Either way, I saw fresh fruit labeled "prunes" about an hour ago, and hopped on wikipedia. I wouldn't have done that if I didn't have a clear set of (apparently inaccurate) criteria for prunedom in my head, and it wasn't a very emotional experience for me.

Geoff Holtzman

Pardon the format:

Side 1 – 4 questions

Imagine a world in which scientists figure out the exact state the universe was in at the time of the big bang, and figure out all the laws of physics as well. They put this information into a supercomputer, and the computer perfectly predicts everything that has ever happened and ever will happen. In other words, these scientists prove that everything that happens has to happen exactly that way because of the laws of physics and everything that's come before.

1. Do you believe that in the real world, everything that happens has to happen exactly that
way because of the laws of physics and everything that's come before?

No Yes

Imagine that in a world where everything that happens has to happen exactly that way because of the laws of physics and everything that’s come before, someone commits a crime.

2. Was this person free to choose whether or not to commit this crime?

No Yes

3. How morally responsible would such a person be for committing this crime?
Not at all A little Somewhat Responsible Very Extremely Absolutely responsible responsible Responsible responsible responsible responsible

4. How much would this person’s level of moral responsibility depend on the nature of the crime?
Not at all A little Somewhat Dependent Very Extremely Absolutely dependent dependent dependent dependent dependent dependent

Thomas Nadelhoffer


Here are a number of off the cuff remarks:

First, you describe your paradigm as involving an “abstract compatibilist” vignette. But that’s not quite right. For starters, a vignette can’t be either compatibilist or incompatibilist, only people’s responses can be compatibilist or incompatibilist. Vignettes are neutral in this way. They simply present certain threats—e.g., in your case, you chose a case that made both prediction and determinism salient to participants (much as Eddy and I did with our own supercomputer case)—and then you present Ps with questions. So, while cases can be abstract or concrete, they can’t technically be compatibilist or incompatibilist. Or so it seems to me…

Second, it strikes me that the attempt to show that affect doesn't play a role in these cases—given how pervasive affect seems to be in general (think somatic markers, etc.)—is a bit misguided. At best, what you should be trying to do is show that affect explains *less* of the variance than researchers previously assumed. After all, your own DRP explains .5, but that still leaves lots of room for other things…like affect. Or are you really wanting to say the entire intuitive process is affect-free?

Third, I am bit concerned about the ambiguity of the phrases and words you used as measures. For instance, rather than using “free will” language (which is vague and ambiguous enough as it is!), you adopted instead “free to chose” language. I think this is problematic. For while I would deny that this agent had free will, since I am an incompatibilist, I would nevetheless be inclined to say he was “free to chose” lest I lose the ability to distinguish his case from a case involving necessity or duress. But if this is so, then you artificially inflate the number of compatibilist-looking answers since you’ve allowed people to focus on one of two aspects of the situation—namely, whether determinism is true and whether the agent acted voluntarily. I have the same worry about “morally responsible,” since it allows far too many meanings to be an especially useful measure. I think you have to be careful to distinguish forward looking and backward looking considerations on this front—which is a lesson I have learned during the past few years working on the free will scale. But more on that later…

Fourth, and most importantly, I am bit confused about your surprise! After all, it is widely recognized that one’s level of responsibility for a crime depends on the nature of the crime (which is true regardless of whether one is a compatibilist or incompatibilist). Consider, for instance, the crime of murder. It’s badness is built into its definition. If you didn’t voluntarily kill someone with foresight or intent, you didn’t murder them (which is true regardless of whether determinism or indeterminism is true). So, of course, the “nature of the crime” is relevant to a criminal’s responsibility—it’s built right into the definition of the crime itself. If the criminal had merely been guilty of manslaughter instead, he would be less responsible. But to the extent this is true, your attempt to get explanatory mileage out of the DRP is in trouble. Too much information is built into that simple phrase for it to be discriminating. That’s why DRP “explains” so much of the variance.

Just some passing thoughts, though. I look forward to seeing the paper!

Geoff Holtzman

Thomas, thanks for the feedback!

First: You’re right, “XX cases” probably most reasonably reads as “cases where XX has been stipulated” (e.g., “determinst cases”). I intended “compatibilist cases” to be shorthand for “cases that required participants to make implicit judgments either for or against compaibilism,” but this was loose language on my part (a performance error!), and I will make sure to be clear about that in the paper. Thanks for pointing that out, I would have kept missing it.

Second: My thought is not that there is an affect-free, completely cognitive process at work here. The degree to which a person recognizes a dependence of moral responsibility on the nature of an act might mediate the effect of (affect-laden) value judgments on judgments of moral responsibility. What I mean is something like this: The inference from moral value to moral responsibility in moral cases may often be like the inference from economic value to financial responsibility, and we normally don’t think of financial responsibility as an affective competence. Consider those who do poorly on the Iowa gambling task: their “reasoning skills” seem intact, it’s just that they are unable to formulate the somatic markers necessary to affectively encode the economic prospects of their “hits” and “misses” in the first place. SoI think affect does play a large role in judgments of moral responsibility, but that this effect is largely indirect, and often mediated by judgments of moral value (and most of the research so far is on value judgments).

That said, this still leaves room for other non-cognitive processes to intervene directly on judgments of moral responsibility, and I certainly think they do. I’m a big fan of Adam and Edward’s work, and I actually measured BFI neuroticism and gender for this study. These two plus the DRP together account for more like 75% of the variance—and there’s a significant portion of the variance these factors account for but the DRP does not. It’s not meant to be a catchall, but rather to show that there are important, principled cognitive differences that underwrite concrete/abstract effects to a large (r > .7 for just DRP) extent.

Third: On free will vs. free choice, fair point. It would be interesting to the influence of word choice here directly-- has anyone done that? Although, I was careful in my post not to mention free will—I have a whole other spiel of where I think that comes into the picture. On moral responsibility for murders and crimes: I agree that the distinction between the forward-looking “capacity” and backward looking “accountability” definitions is extremely important, and shouldn’t be overlooked. I also think that it largely is overlooked by philosophers, and that this is a large part of why the compatibilism debate has become so entrenched. It seems to me that compatibilist views very often assume a backward-looking definition, whereas incompatibilists take that concept to be irrelevant and only care about a forward looking approach. As an incompatibilist (turned metacompatibilist, though incompatibilism still comes most naturally to me), I find Dennett’s claim that “free will is whatever makes moral responsibility possible” to sound awful ad hoc, a lot like metaphysical data mining, whereas Dennett wouldn’t find my (formerly committed) definition of free will “worth wanting.”

Fourth: I’m not surprised by the significance of the results at all - just that the effect was so large! A lot of people have taken after Shaun and Josh’s apparent surprise toward the very exists abstract/concrete effects: “What on earth could have caused these differences?” Gee, maybe individual differences in the interpretation of an ambiguous, ill-defined, and highly confusing philosophical concept? This, ultimately, gets at the philosophical thesis (as opposed to psychological processes) I’m interested in here: the term “moral responsibility” conflates two concepts (or maybe a whole spectrum of concepts), and inasmuch as arguments for and against compatibilism ignore the the paramaters of DRM that define their debate, nobody is going to budge because nobody cares about the concept the other is talking about. Each side is engaging in its own debate, and developing arguments incommensurable with those on the other side, though neither side is necessarily being unreasonable, given the rules of the game each is playing.

Too strong? This certainly doesn’t follow from the data; rather, the data are only meant to illustrate the idea, which I think warrants follow-up studies. The experiment itself was run mainly to test other, stronger hypotheses (e.g., that gender and neuroticism predict compatibilism only for abstract but not concrete cases… story for another time).

Geoff Holtzman

Thomas, I actually think you get at the core of my views by essentially denying the DRP when you say, "After all, it is widely recognized that one’s level of responsibility for a crime depends on the nature of the crime (which is true regardless of whether one is a compatibilist or incompatibilist)." I disagree with this on two accounts, which is what I like about these data:

(1) The part outside the parentheses is (according to my descriptive statistics) a false premise that leads to a lot of confusion in x-phi; or rather, that assumption is false, as is my own former assumption that it is widely recognized that moral responsibility dependent only on events before, but independent of events constituting or occurring after, an act. It turns out that there is a distribution of levels of recognition of this principle; you recognize it, I don't (at least not intuitively), and the people I tested are all across the board.

(2) The part in parentheses is false, and this is the point of my inferential statistics. DRP isn't accepted as true "regardless of whether one is a compatibilist or incompatibilist": its acceptance was correlated >.7 with judgments in favor of compatibilism here! And this prediction of compatibilist judgments is exactly why I think studying the principle might be fruitful.

Thomas Nadelhoffer

I was simply pointing out that "nature of the crime" seems far too ambiguous to do any explanatory work, or so it seems to me (without seeing either your data or your paper/argument). There are a number of things that are part of the "nature of the crime" when it comes to acts that are malum in se such as murder and rape that are relevant to compatibilists and incompatibilists alike (just as there are features of the crime that are relevant to retributivists and deterrence theorists, alike).

So, if I your participants had an expansive understanding of "nature of "crime,” they went one way. If they had a more restrictive understanding, they went another. What I was trying to point out earlier is that it’s unclear why I should follow you in thinking this has anything to do with compatibilism, proper.

Here are some features of a crime that I predict nearly no one would deny is relevant to responsibility ***if presented as such***--e.g., whether the perpetrator intended the results, whether the victim was himself guilty of a related crime (e.g., an assault and battery in exchange for a armed robbery), whether the perpetrator was having delusions, whether the perpetrator had control of his impulses, etc.

These are all “features of the crime”--and hence, at least on one reasonable picture, part of the "nature of the crime." I predict that had you presented your participants with questions specifically such as these, you would have gotten affirmative answers from nearly everyone—even the ones you’re now claiming don’t think responsibility is partly based on the “nature of the crime.”

My suggestion is not that there are not individual differences. My suggestion is that whatever difference you’re detecting need not have anything to do with compatibilism, determinism, incompatibilism, etc.

To have something to say about that, I think you have to ask different questions. Or, you could use some of the scales people have developed to measure free will beliefs. But as things stand, at least from what I can tell based on your post, you haven’t convinced me that you’ve isolated a compatibilist intuition!

As I said before, if you presented me with the deterministic and predictive scenario and asked me whether in the concrete case the person "chose" and is "responsible" you allow me to build too may different interpretations into my responses.

You have to use contra-causal free will and Galen Strawson-style ultimate responsibility and Pereboom-style hard determinism because those are the rival stories that are all consistent in various ways with your data. Unless and until you use language that is sharper than “free choice” and “morally responsible,” you’ve not done what you claim to have done.

But here again, once I see the paper, perhaps I will be convinced. That’s how this works, no? But it's not like I am just making it up when I talk about the role that "the nature of the crime" plays in people's thoughts about responsibility and punishment. Indeed, I wrote a half-baked dissertation about precisely this issue. And Mikhail has done more than I ever could have to show that in some important sense "nature of crime" is too broad for the purposes of getting at the intuitions I take you to be after.

Thomas Nadelhoffer


Since you asked for feedback, here are some additional clarifications of some of the worries I raised yesterday along with some additional concerns I now have upon further reflection:

First, I want to revisit and sharpen the main worry I raised—namely, that you used ambiguous language that invites even incompatibilists to give compatibilist-looking answers. If you want to be able to distinguish incompatibilists from compatibilists, you have to use more precise and discriminating language. For instance, instead of asking just one “free choice” and one “morally responsible” question, you could and should have used two (or more) questions/items for each concept you’re exploring so that you can go on to make the distinctions you want to be able to make. Instead you chose language that makes it maximally likely you get compatibilist-looking responses. But if this is just an artifact of your design rather than a real insight into how people think about these issues, then you have a serious problem. Minimally, it complicates your attempt to shed light specifically on the folk incompatibilism debate—even if you have shed light on folk attributions of responsibility more generally. For what it’s worth, I think this same problem plagues nearly all of the work that’s been done on this front. But that’s another story…

Second, I suggested that you shouldn’t be in the business of trying to show that the processes undergirding P’s responses are affect-free. Here is an additional observation that highlights my concern. Much of what Ps build into the “nature of the crime” is likely affectively loaded. For instance, imagine a gruesome murder involving torture, etc. The more details you read me about the nature of the crime, the more affect will presumably be generated. Indeed, the worse the crime, the more negative my affective response. That’s why reading about some crimes can literally make us nauseated with disgust and indignation. That’s also precisely why so many modern systems of criminal law tie the severity of punishment at least loosely to the severity of the crime. The more horrendous the crime, the worse we think the punishment ought to be (i.e. the more responsible we take the violator to be). Notice that this stance is consistent with both retributive and consequentialist models of punishment (and it’s also consistent with compatibilist and incompatibilist accounts of FW and MR). In that sense, it’s theory-neutral. It’s a basic principle of punishment and justice that the severity of the punishment ought to at least loosely match the severity of the crime. But to the extent that this might be all that you’re studies are picking up, I am not sure the story you’re wanting to tell with the data makes as much sense. After all, plenty of affective content is presumably being built into the “nature of the crime.” If so, then you can’t use your results to challenge the work by Josh, Shaun, and others. It’s only if you show that your P’s understanding of “nature of the crime” is itself affect-free that you can go on to challenge the models of the data that depend on affect.

Third, I am curious why you selected the particular case you did since it’s both complicated and atypical—especially since one of the issues was the defendant’s intent. It would be preferable to use a real world case that is clear cut in terms of the mental states of the agent.

Finally, I want to hear a bit more about how you're thinking about the alleged incompatibilists who deny that responsibility depends on the nature of the crime. After all, on your view, these are people who claim that the severity of the crime, the details of the crime, etc. aren’t relevant to the responsibility of the agent. You’ve assumed these respondents are mostly incompatibilists (which is an inference I have already questioned). Let’s assume that’s right. Then what do you think these people think responsibility depends upon, precisely? After all, it’s unlikely all of these people are FW and MR skeptics. So, if they don’t think that responsibility is at least partly tied to the nature of the crime, then what? In my eyes, it would be helpful if you asked some additional questions that would enable you to see what people are building into the content of “nature of the crime”—especially since it plays such a crucial role in your study. After all, if DRP is really just picking up on the fact that as a general rule, we (a) build mental states into the definition of certain crimes, and (b) we grade punishments at least partly on the nature of the crime (esp. the mental states that are built in), then it’s no surprise it explains so much of the variance of the results.

Thomas Nadelhoffer

p.s. Earlier I claimed that "it is widely recognized that one’s level of responsibility for a crime depends on the nature of the crime." You claimed to have falsified this claim. I disagree. The evidence for my claim is that most of the systems of criminal law in the modern world are (at least partly) predicated on the assumption that moral blameworthiness tracks the severity of crime. So, if you run a study with 115 (or 230) people that suggests otherwise, I am more inclined to side with the cultural tradition--especially when your study utilized wording I find problematic.

So, contrary to what you suggested, your descriptive statistic did not falsify my claim. Falsifying this particular claim would require more than social psychology. It would also require some comparative legal studies, etc. The worry about making sure that the "punishment fits the crime" has animated nearly every modern legal system. If we assume that legal systems can fruitfully be viewed as repositories for folk intuitions and beliefs about responsibility and punishment--which I think is a good assumption--then, you're going to need a lot more evidence than you've provided to falsify my earlier claim. Or so it seems to me. Either way, I look forward to the results!

Geoff Holtzman

Thomas, I’ll agree that “nature of the crime” is ambiguous. What I would like to separate are the events preceding and circumstances surrounding the act from the exact behavior of a person in the moment an act is committed. Wording this well for vignettes will be a challenge—but do would you agree that these are the two questions I need to ask/separate (and in that order, but worded better) to disambiguate in order to make stronger claims about just what the data show? That, and saying something neutral like “act,” rather than something weighted like “crime.”

On your second point though, I disagree that “the nature of the crime” will be equally relevant to compatibilists and incompatibilists, even if we define it in the broader way you do. If a person thinks some factor is compatible with moral responsibility, that person certainly needs to assess facts about the nature of the crime. But if a person accepts in principle that some fact of a case—- physical determinism, insanity, intent, or anything else—- is incompatible with the possibility of moral responsibility, isn’t she essentially done her reasoning process? Once some factor is identified as making it impossible for that person to be responsible, who cares about other factors that might limit responsibility?

If you disagree here, it might be that you and I are using “incompatibile” differently. Here’s how I use the word: A-->~B. Suppose Jones has never left the U.S., and Smith has been assaulted in New Guinea. I take it that never leaving the U.S. is incompatible with assaulting someone in New Guinea. This is enough to entail that Jones did not assault Smith in New Guinea. This means that Jones’s guilt is independent of the nature of the crime (he's not guilty), however we might construe the term. I don’t care if Jones had intent, or was delusional, or had his DNA all over Smith’s body. If I want to weight any further facts regarding the nature of the crime against Jones (other than his and the victim’s locale), then I would either have to relinquish my “geographic incompatibilism”, or deny the premise. If by incompatibilism, we mean the view that determinism is incompatible with moral responsibility in the way that being in the U.S. is incapable with assaulting someone in (being in) New Guinea, I’m not sure why a determinist would care about the nature of a crime. Or do we take incompatibilism about moral responsibility to mean different things? I just take it to be a thesis about the general incompatibility of determinism and moral responsibility.

However, all this does bring up another shortcoming in my design, which I think you initially pointed out. The strongest evidence would show that those who previously asserted the independence of MR and the nature of an act do not go back on this assertion in concrete cases. My data is for the contrapositive claim, and I’m not sure how to do the strong data in a way that wouldn’t confound the experiment by nudging participants to make judgments that adhere to their earlier assertions about dependence. Thoughts on that?

The comments to this entry are closed.

3QD Prize 2012: Wesley Buckwalter