I’ll try to keep this brief, and so will likely run roughshod over important points. I’m curious about what’s doing the work on our intuitions in so-called manipulation cases when people deploy them to theorize about responsibility. These are cases in which someone is one way, values-wise, and then her brain is manipulated by a team of neuroscientists/god to produce within her a new set of values (or subset of values), so that she now performs some action for which she is not responsible — or at least that’s what our intuitions are supposed to be.

But there are two features of such cases that aren’t prized apart, typically: (a) the fact of external alteration of values; and (b) the fact that the values have changed pretty radically. But which one is doing the work on our intuitions? We can imagine each without the other. Suppose that Jay is a pretty bad hombre, committing at least one hurtful action per day. Now suppose the neuroscientists destroy the values generally producing such actions and replace them with exactly similar values. Jay then commits a hurtful action. Is he responsible? Seems to me quite clear that he is (although perhaps your intuitions differ; if so, I’d like to hear from you). It’s thus not the external manipulation per se that’s doing the intuitive work.

Perhaps, then, it’s the value change? But value change in and of itself doesn’t typically cause us to change our intuitions about someone’s responsibility. Nomy Arpaly, Michael McKenna, and Jonathan Glover have all given real-life cases in which a person with one set of values transforms into someone with a different set of values (e.g., the confirmed kid-hater who comes to love his own child; the hard partier who becomes the industrious worker; the religious convert). And we don’t think these people aren’t responsible for what they thereby do.

I guess there’s a third possibility: the suddenness of the change. But why should that make a difference? In talking about related matters, Parfit once remarked that it’s the fact of the psychological change (and how large it was) that matters, not its rate.

Still, maybe the rate of change is derivatively important, as a sudden change in values may make it harder to attribute the new values to a person as her own? One might then stipulate that numerical identity can be preserved across manipulation (as Al Mele and Ish Haji explicitly do) even when the values with which one is identified have changed so dramatically as to alter one’s moral identity (or, more generally, one’s practical identity). I wonder about this, though. It seems there can be sudden, dramatic changes in some of one’s values in real life cases — in, e.g., a religious conversion — without a change in one’s responsibility status. (There is much more to say here, but I won’t.)

There’s an important related issue here as well: When and for what are the people in these cases being appraised? Suppose Clara-1 is a nasty person who does nasty actions but then is externally manipulated/brainwashed to have a different, kinder set of values for a 24-hour period (while preserving her numerical identity). Call this Clara-2. During this day she performs several kind actions. Then the next day she’s back to her old nasty self and performs more nasty actions (Clara-3). (This is modeled on a famous Mele case.) We might assess the actions performed on the kind day or the pre-kind day (or the post-kind day, but that’s less interesting). And we might, for each of those assessments of actions, appraise Clara-2 or Clara-3 (or Clara-1, but that’s less interesting). So we could appraise Clara-2 for the kind actions on the kind day OR we could appraise Clara-2 for the pre-kind-day’s nasty actions. Or we could assess Clara-3 for the kind actions on the kind day or for the pre-kind-day’s nasty actions.

My own intuitions are that Clara-2 is not responsible for the nasty actions of the pre-kind day but she IS responsible for the actions on the kind day. And it seems to me that Clara-3 is not responsible for the kind actions on the kind day but she IS responsible for the nasty actions on the pre-kind day. Then it’s not the manipulation as such that alters intuitions but rather that it seems a mistake to attribute actions to an agent whose values are in fundamental discord with the values that were manifested in those actions. And if they are in concordance, then one may attribute those actions to the agent. But then in real-life cases, this would imply that undergoing a religious conversion should get you off the hook after all for your Saul-like persecuting actions pre-conversion. And that seems wrong.

I know a lot has been written on this. References are welcome, but also just your thoughts on the matter. What’s the source of altered intuitions in manipulation cases, and can it be carried over coherently to real-life cases? (Apologies for the meanderings, but I’m trying to return us to the days of the occasional half-assed post.)

21 Replies to “Manipulation Cases and Responsibility: What’s doing the work?

  1. My idea is that when a change is radical ENOUGH, then what we have in effect is a new moral agent (and that might be the case even if they’re the same person on standard theories of personal identity).
    I think the change might be radical enough if the pre-change agent could not in a normal way make plans for what to do post-change and carry those plans out, because her pre-change will is simply not effective post-change.
    When there’s a break in agency, there’s a break in responsibility. Basically, my intuitions follow yours. I think Manuel Vargas suggests something similar in Building Better Beings.
    (Obvs this is not a perfectly worked out theory…)

  2. My first question is, why are so many discussions of these types of cases still laid out as if “values” were some discrete module in the brain/mind which can be replaced, even in theory, without any change to anything else? It seems to me that the most distinctive feature of normal human valuation is that it is reflective/recursive; we at least implicitly value, not just some first-order object (say, being nice to others, putting ourselves first, etc.) but /having responded to (some subset of) our entire situation with this particular first-order valuation/. If you change the first without the second, you’ll get someone who finds himself doing something he disapproves of on a second-order level. Now, in practice we don’t always have all our orders lined up coherently, and sometimes we do things at the 1st or 2nd order level which we do not approve of, and on reflection would disapprove of, at a higher-order level. But we are more likely to find such changes jarring if they regard, not mere changes in taste (strawberry vs. chocolate) but serious, central values like whether we respect other people.

    In the real-world cases of value conversions, I presume that either the higher-order valuations are also changed, or become active when before they were inactive; or (when either the change is for the worse, or when some of the higher-order values are themselves still mucked up, e.g. Huck Finn), they are repressed and reflection is intentionally avoided at some point. And then my assignment of responsibility tends to follow the extent to which I think the person is capable of being guided by some higher-order evaluations of her first-order values (whether or not they are *actually* so guided). But many hypothetical scenarios (not yours alone) under-describe the situation. If I don’t know how continuous the person’s higher-order values are (which could at least potentially lead to her being troubled or at least startled by her changed first-order valuations, or reevaluating her past behavior, etc.) I have no intuitive responses to the scenario. Of course, very significant and radically discontinuous alternation of all of these higher-order values raises the question of whether this is really the same person after all.

  3. The difference between the real life conversion cases and the value change cases you mention is that in the real life cases the agent can identify an event to attribute the change to (e.g. God’s intervention, a religious experience, etc.).

  4. Suppose that the neuroscientists had destroyed Jay’s values and replaced them with similar but not identical values, so that Jay would occasionally (but rarely) act in a way that he wouldn’t have acted had the neuroscientists not tinkered with his brain. Would you say that he is responsible for those “aberrant” actions?

  5. Sofia, thanks! I agree that, to the extent some appeal to identity is doing some work, it’s not numerical identity but something like moral or practical identity (you use the language of it being a different “moral agent”). That means we need to get clear on those identity conditions. Fortunately, there’s some awesome recent empirical work (by Nina Strohminger and Shaun Nichols et al.) that goes into detail about how folks view changes in moral characteristics as grounding a change in the agent’s “identity.”

    Scott: Exactly right. I’m trying to grant to those who engage in this sort of argumentation everything they want, which includes “values change,” the relevant intuitive responses, and underdescribed cases. The real world (and real value systems) are much more complicated than this literature has really allowed to this point (for the most part, with exceptions I’ve noted). I’m aiming to get it back to real world cases, and as I mentioned to Sofia, actual empirical research into what people find to be the most disturbing sorts of psychological changes for reappraisals of persons.

  6. Ben: Interesting. That may be true of some cases, but not all. Arpaly’s case of the former hard-partier who becomes an industrious worker may involve no such (identifiable) event. Further, suppose the team of neuroscientists looked you in the eye as they manipulated you (and talked you through the procedure). Now you’ve got an identifiable event to attribute the change to. Or do you mean there’s no identifiable event the agent can point to as what sparked *her own* subsequent change?

    Mike: Nice case. I don’t know, is the answer. Here’s the issue: When we’re talking about a whole “set” of values, a change to just one of the set may not change one’s pattern of behavior very much, and so we might be inclined to think, “Well, he’s changed very little.” But when it comes to individual (tiny) values that influence a very small number of actions, the change with respect to *them* could be viewed as quite significant. So perhaps when Jay acts in that aberrant way, my intuition is that it’s not his action but the scientists’, but when viewed through the prism of his *patterns* of behavior and aretaic assessments, he’s still the same moral agent. (I see that I’m letting my own views about multiple qualities of will shine through her, which I’d intended to avoid.)

  7. Dave,

    This is interesting stuff. In response to your question about what’s driving our intuitions in manipulation cases, it’s worth mentioning another relevant empirical project (sometimes also involving Nina Strohminger, along with Josh Knobe and and others). I’m sure you know of this. But it seems relevant to your question and so worth calling attention to.

    There is evidence that our intuitions about the “true self” track our value judgments. Roughly, the findings suggest that those inclined to think that x-ing is good will be more likely to attribute a desire to x to an agent’s true self than those inclined to think x-ing is bad. If this is, indeed, how we think about these matters, then our intuitions about, say, conversion cases are likely sensitive, among other things, to our assessments of whether or not the conversion was for the better. Thus, a conversion case in which someone goes from a jerk to a saint may elicit different intuitions than one in which someone goes from a saint to a jerk. We may be more likely to judge that, in the first case, the conversion does not undermine responsibility, but that it does in the second case. Our intuitions about cases involving external manipulation may, similarly, track our evaluations. We may be more likely to judge that external manipulations that render agents better people are less responsibility-undermining than ones that render them worse people. I’m not sure how this line of thinking would apply to your case, where the external influence leaves the agent’s values as just as bad as they were.

    Thoughts about any of this?

  8. If Arplay’s case is meant to be like real life case rather than a purely hypothetical one with dodgy stipulations (like the value change cases tend to be), then I think the hard-partier will find something to attribute her change to. If I’m just supposed to accept that she doesn’t have any identifiable event (and that this change occurred overnight), then I don’t see how it is any different from the value-change case.

    In the revised neuroscientists case (we might also just imagine the neuroscientists implant a pseudo-memory of an identifiable event), I’m inclined to think the value-changed agent *is* responsible. But of course recall that I take the same sort of line you mention in the post about value changed agents in my “Compatibilism and Personal Identity” (as does Andrew Khoury in his “Synchronic and Diachronic Responsibility”). What’s important to emphasise is that the issue is more about responsibility over time, i.e. the persistence of the property of being morally responsible for A-ing. Just because Clara-2 instantiates it doesn’t mean Clara-3 instantiates it.

    One thing that I think is ambiguous in your post is the part about Saul “getting off the hook”. I think that reasons/expectations/duties to apologise or compensate last longer than the fittingness of attitudes. So it’s possible for a person to no longer to be a fitting target of the negative reactive attitudes for an earlier wrongdoing and yet still have a reason/expectation/duty to apologise or to compensate others for that wrongdoing. So just because (let’s say) Saul is no longer a fitting target of the negative reactive attitudes for his misdeeds doesn’t mean he is completely off the hook. Does this mean I think Clara-3 has a reason/expectation/duty to apologise for Clara-2’s action? I have more to say here, but I’ll leave it there.

  9. Ben, yes, this is yet another complication. I’ve struggled with the empirical stuff on the “true self” for a while now (David Faraci and I have a new article on it in PPR, along with a new study on the Huck Finn/JoJo cases). I go back and forth between thinking that it’s a debunking argument against general appeals to intuitions in cases like this, and thinking that it’s just an excellent articulation of the complicated intuitions we *have* in cases like this (in which case theories have to take them into account). Do you have thoughts on this?

    I have no doubt that were X-phi work done on this variety of manipulation cases, where the direction of value modification (from good to bad or vice versa) were the relevant variable, people’s intuitions would very likely be variable as well. If so, then what’s doing *some* work in manipulation cases is assessor’s antecedent normative commitments! And that’s surely not what the defenders of the cases want!

  10. Ben: Can’t the process of maturation, for instance, take place gradually, with no defining event, but where the end points are dramatically different?

    I’m in total agreement with your general strategy, articulated in the last paragraph, in which the many different aspects of our responsibility practices are actually prized apart. We can see that *some* of them may be apt at later times while others aren’t. I think this is a terrifically promising and plausible strategy, and I’m glad that you and Andrew are taking it up. Again, I’ve tried to grant the manipulation-case theorist everything he wants, which includes dramatic and coarse-grained use of the term “responsibility,” as in, “Is Clara-2 responsible for her actions during the kind-day?” What I’ve been hinting at, and what you’ve made more dramatic, is that “responsible” includes a huge variety of attitudes and practices, some of which might be appropriate, some of which might not be, depending on what factors are relevant, when we assess the agent, and what we are assessing the agent for.

  11. Yeah, I agree there doesn’t need to be a defining event for incremental but overall radical changes. I just think it’s important for the sudden radical change cases.

    Cool! For what it’s worth, it’s inspired by a strategy you’ve taken with respect to personal identity and practical concerns.

  12. Dave,

    Thanks for the response and also for flagging your work with Faraci. It’s interesting! I’m not entirely sure what to think about this, but I am very sympathetic to the Good True Self view. The empirical evidence is suggestive, and I find it pre-theoretically attractive. I am taken by the idea that we are (often? always?) taking things in, examining them and making judgments from a standpoint that is not evaluatively neutral. This seems to me most plausible in the contexts we are talking about, namely, judgments about other people’s actions. The GTS adds to this picture the idea that we are predisposed to take people to have good-by-our-lights true selves. There’s something attractive about that.

    I’m not sure what to say about the debunking vs. complicating take. I suppose I tend towards the latter, but with the caveat that, as I understand things, the GTS view allows that what one takes to be good is subjective. So there is an air of debunking, because there is no claim about objective value. What one takes to be a good true self varies according to one’s values, and someone with abhorrent values would tend to attribute characteristics to an agent’s true self that many others would find repulsively bad. One attraction about this is that this allows the GTS to be a descriptive view, and it is compatible with various accounts of how we *should* judge matters. That is, we may appeal to the GTS in order to describe why someone is making the judgments she is, while at the same time appealing to other views in order to argue that she should be judging differently.

    Your second point has to do with what the defenders of manipulation cases want. Well, I agree that empirical study would likely uncover the patterns you suggest. But I’m not so sure that would be unwanted by those trotting out manipulation cases in order to impugn compatibilism, say. This is for the reason I just gave–there is room to argue about what the correct judgments should be, even while granting the GTS-type story about why they are what they are.

    One final thought, and then I’ll quit. In thinking a bit about your PPR paper with Faraci, I wonder if there really is a conflict between the GTS view and the Knobe effect. The latter has to do with judgments about intentional action, whereas the former has to do with judgments about the true self. But it seems clear to me that not all intentional action issues from the true self. Rather, the true self is implicated in (certain forms of) responsible action, as well as, I would think, self-governed action. But these are both plausibly much more robust forms of action that intentional action. So it seems to me that the different studies are getting at different things. What do you think?

  13. Thoughtful remarks, Ben. Two responses. First, here’s why I sometimes think of GTS conclusions as contributing to a debunking argument in some of the responsibility domains: How *else* might one argue about how people *should* view the deep self except by appeal (in some respect) to intuitions about perhaps more straightforward cases? But if actual intuitions about cases are variable depending on antecedent normative commitments, then on what basis might we argue for the incorrectness of those intuitions?

    Very interesting distinction between the GTS view and the Knobe effect! (For those at all interested, Faraci and I argue that the GTS view actually predicts the *opposite* of the results Knobe got about what counts as “intentional” in his famous chairman/environment cases.) Obviously, this could be tested empirically. But on its face, I doubt there’d be much difference, at least insofar as people attribute a wide range of psychological features/events to people. After all, in the GTS studies, some of the things being attributed to people’s true selves included desires, attractions, and beliefs, so I don’t see why intentions wouldn’t be among those things. To the extent these features are relevant to responsibility (and they all are, to some extent, I would think), they ought to be grouped together. But again, this is empirically disconfirmable, and it would be interesting to see such a study.

  14. Dave,

    Thanks for your thoughtful reply. I’ll take your comments in reverse order.

    I’m afraid I wasn’t very clear in articulating my thought about the lack of conflict between the GTS view and the the Knobe Effect study. Here’s another stab: Even if, as the GTS view has it, people’s attributions of psychological states to others is influenced by their own values, this need not suggest that people would be more likely to attribute an intention to the CEO in the help case than in the harm case (assuming that most people value helping the environment more than harming it). And this is because one might attribute an intention to an agent *without* thereby attributing it to his “true self.” After all, the GTS view, as I understand it, is compatible with attributing psychological states to individuals, even when these are not a part of their true selves. So one interpretation of the Knobe Effect study is that people are more likely to attribute an intention to the CEO in the harm case than in the help case. But this is not yet to say that they attribute this intention to his “true self.” Does this make sense? Seem plausible?

    On the topic of debunking, I’m less pessimistic, I guess. After all, there are familiar arguments (e.g., Socrates, Mill) about why some groups of individuals’ judgments are more reliable than others’. One might run a similar kind of argument, beginning with an appeal to the claim that certain evaluative standpoints are more correct than others and using this as a basis for the claim that the intuitions of individuals with the relevant values are to be privileged over others’. The result would be that we can still argue from claims about intuitions about cases to claims about the true self. And we needn’t be deterred by the fact that intuitions are variable. Of course, there are worries about privileging certain standpoints over others, especially when it comes to passing judgment on others’ true selves. And these worries should not be taken lightly. But there does seem to me to be interesting space for interaction between the type of descriptive view given by the GTS stuff and purely normative work in value theory. And this opens up the possibility that the GTS view is not simply a debunking argument against the use of intuitions in the relevant cases. Thoughts?

  15. Perhaps interestingly, I judge manipulation cases differently depending on the moral character of the resultant personality. Suppose your neuroscientists instead replace Jay’s wicked values with benevolent ones. At least after a week or so has passed (to allow a bit of habituation), I judge that Jay is praiseworthy in virtue of all the good deeds he does. But when we manipulate the brain of a good person to make him act in evil ways, then, even a month later, I’m inclined to judge that none of the bad things he is doing are in any way his fault.

  16. Ben, I got the point, so I must not have articulated my response very clearly. Responsibility/attributability, for Knobe, is all about the true self. To the extent, then, that intentions are among the things that are sufficient for responsible/attributable action, there’s no (prima facie) reason to think that they won’t be among the psychic elements that should be governed as well by the GTS results. But they aren’t, in the Knobe effect cases. But as I said, this issue could be settled empirically, and it would be interesting to see whether that’s the case. In other words, I agree with you that, in principle, people wouldn’t have to be attributing intentions to the chairman’s true self, but I’m highly dubious, given the prima facie place of intention in responsibility/attributability.

    You’re right that there are other resources available for identification of mistakes in attributions. People might have a correct normative theory but not issue in attributions that are consistent with it. Or they might have a clearly mistaken normative theory. Or the issuances might conflict with the TRUE attributability theory. I was most concerned with the last possibility, wondering how *that* sort of mistake might be cashed out, i.e., how one might determine the true attributability theory independently of appeal to intuitions.

    Todd: You have given an answer that is precisely grist for Knobe’s mill!

  17. Dave,

    Thanks for your patience. I think I see the issues more clearly now. I want to deny that responsibility/attributability is all about the true self. It seems to me that one can be responsible in the attributability sense for a wider range of things than just actions that issue from elements of one’s true self. On my view, the true self is more closely tied to autonomous action, which has stricter conditions than responsible/attributable action.

    Given this, it would make sense to suppose that people attribute intentions, even in responsibility/attributability contexts, without it being the case that they are attributing these intentions to the agent’s true self. It makes sense to think that subjects might attribute an intention to the CEO in the harm case without attributing it to his true self. What’s more, the GTS view would seem to put pressure on people to attribute the intention to the CEO in the harm case *without* thereby attributing it to his true self. He intentionally harmed the environment, but this intention is not a part of his true self. And on my view, this would roughly translate to something like: the CEO is responsible in the attributability sense for harming the environment, even though it was not an action he fully endorses.

    Perhaps I have an idiosyncratic view of things. But it strikes me as having the virtue of compatibility with both sets of empirical findings. But maybe I’m missing something else here.

  18. I suspect that there might be another element to explaining this intuition. In all of these cases the manipulation is done by some other agent/person (neuroscientists, God, etc.). The manipulated agent becomes something like a pawn in the manipulator’s game. Suppose A manipulates B into doing X. When I ask whether B is responsible for doing X, I am tempted along this line of thinking: “It’s not *really* B’s fault. A is the one to blame.” I think this line of thinking is misguided since it is possible for there to be plenty of blame and responsibility to go around—both A and B might be responsible for B’s doing X. But, when assigning blame, I am often tempted to be most angry with the person that is *ultimately* responsible for whatever happened. So I am suggesting that the intuition (that B is not responsible) might be partially due to letting the manipulator’s blameworthiness/responsibility distort our view about the manipulated agent’s blameworthiness/responsibility.

    Some evidence in favor of what I’m suggesting is the fact that the zygote argument is typically considered to be a version of the manipulation argument. But in the zygote argument there hasn’t been a sudden alteration of values. In fact, it appears that there has been no alteration of values at all (since there were no values there to begin with). And it doesn’t matter whether the zygote argument actually works. My point here is just that the tendency to consider the zygote argument as being akin to the other manipulation arguments suggests that the manipulator’s role is playing some role in our intuitions.

    Another quick note: In my first paragraph, I conflated responsibility and blame. But “being blameworthy/praiseworthy for X” is often thought to be distinct from “being responsible for X.” This is another issue that I think complicates our intuitions on these cases. I’m often conflicted about what to think in these cases because it is difficult to disambiguate my intuitions about the different forms of responsibility.

  19. Ah yes, Ben, interesting. I’ve laid out a distinction like this myself somewhere, between “thin” and “thick” attributability, although I didn’t connect it up to autonomy (I have no thoughts about autonomy). Yes, these senses are often conflated in the literature, but it could well be hard to prize them apart in empirical work. I think both senses pertain to responsibility, though, but the thin sense doesn’t necessarily ground any sorts of reactive attitudes on its own. The mere fact that I executed an intention in performing an action pins the action on me as *something I did*, but it doesn’t yet pin on me that the action *reflects on me* in any interesting way. That latter requires appeal to motives, attitudes, and/or character traits (i.e., quality of will).

    I agree, Tim, the fact that there’s another agent is in fact *supposed* to be what guides our intuitions, as this agent is somehow bypassing our autonomous control, say. But the point of my thought experiment of the agent who replaces my values with exactly the same values was that we have reason to think the manipulator is not what’s doing the heavy lifting after all. As for the blameworthy/responsible conflation, you are dead right. Faraci and I point out this sort of difficulty in the empirical work that’s been done thus far in our recent PPR paper.

  20. “But the point of my thought experiment of the agent who replaces my values with exactly the same values was that we have reason to think the manipulator is not what’s doing the heavy lifting after all.”

    This case is an interesting one. I think you’re right that the existence of the external manipulator per se is not doing the work. But the important difference between standard manipulation cases and your case (where the values are replaced with identical values) is that, in the former, the manipulator makes a difference in outcome. So if my suggestion is correct (that we [perhaps incorrectly] downplay the responsibility of the manipulated agent in our search for something like “ultimate blame”), then we might explain the intuition in your revised case as follows. We still hold the manipulated agent responsible because we screen off the manipulator as being irrelevant to explaining why the manipulated person acted as she did. Of course, the manipulator is relevant to explaining why the manipulated person acted as she did, but as a heuristic we tend to ignore facts that don’t make a counterfactual difference.

Comments are closed.