The Manchurian Candidate

Eleonore Stump has criticized the Fischer/Ravizza approach to moral responsibility in various interesting articles. In some of these she invokes Raymond, the main character in the Manchurian Candidate. She first discusses Manchurian Candidate type manipulation in her piece, "Intellect, Will, and the Principle of Alternate Possibilities." reprinted in Fischer and Ravizza, eds., Perspectives on Moral Responsibility (Cornell 1993). She further develops this critique in her contribution to the Buss/Overton volume in honor of Harry Frankfurt. There Harry Frankfurt pronounces his agreement with Stump's critique.

Mark Ravizza and I are certainly aware of many difficulties with our theory, and I find the problem of manipulation particularly vexing. I do however make a stab at defending our view in my recent paper in Journal of Ethics (Kluwer), "Responsibility and Manipulation". (The paper is also posted online here at the Garden of Forking Paths.)

I was just wondering whether anyone had any thoughts about this debate, or, perhaps more interestingly, about how to handle manipulation cases in general. Of course, it would be much easier if I could take the "Dennett" way out and say that the cases are too crazy, or that they are not possible (one can't induce mental states electronically), and so forth. That would solve the problem, but I am not convinced by the move, and I (and my coauthor, Mark Ravizza) have at least tackled this nasty problem head-on.

Finally, have people seen the new version of the film? How would you compare it with the original film?

Comments

[James Gibson]

My good friend Chris Franklin recently wrote up a defense of the Fischer/Ravizza approach to moral responsibility, building off an epistemic priniciple that attempts to defend when it is appropriate, for instance, to take responsibility for an action or omission. To be very brief - since I can't find my copy of the paper and I'm not sure how much I can say about this since he wants to publish it - he suggests that the attitudes have to be formed in an appropriate way, something analogous to how the reliabilists talk about acquiring (true) beliefs that become knowledge. The thing I found appealing about his response was that, if correct, defending your theory against manipulation cases doesn't depend merely on the intuition that the non-manipulative mechanism in the actual sequence is different from a manipulated mechanism. Since you (John) have read his paper, why not hop on board with Chris's defense?

[Chris]

Here's an idea that I like: the brainwashing turns Raymond into a new person, or puts a new person inside of him. So it wouldn't be appropriate, say, to take money out of his Raymond's old bank account to pay fines incurred as a result of his assassinations. But it would be appropriate to punish the assassin. (Or rather, we might think of Raymond-the-assassin as like a bad person who has a good person attached to him. Suppose a killer kidnaps an innocent and super-glues himself to the innocent. OK to put both in prison? Not sure. But OK to hold the killer responsible, whatever that means.)

[Tamler]

Would it be appropriate to punish the "assassin" for non-consequentialist reasons? He was entirely created by those evil communists. The "assassin" himself isn't in any way responsible for being an assassin, nor for having the desire to kill people in the most efficient way possible (using his reason). It seems to me that while we would certainly want to lock him up (as well as the "real" Lawrence Harvey) for pragmatic reasons, the assassin himself is not DESERVING of any blame or punishment. No more so than than the real Lawrence Harvey. It's just bad luck that the assassin was created this way and now has to suffer the consequences. (Question begging? I guess. But it seems plausible to some degree anyhow...)

[robert allen]

I favor Al Mele’s approach to manipulation cases, which is based on the notion of “sheddability.” The determined but unmanipulated agent, according to Mele, has the ability to shed any one of her values. Thomas Kapitan objects that this definition is too broad, since an agent, who can shed values only to replace them with others intended for her by a manipulator, is still under the latter’s control. In response, I propose adding the condition ‘in favor of another not intended for her by those responsible for instilling it’. Such an agent is, thus, able to “subvert” the wills of others.

[Dan K.]

Is there any reason why manipulation should be treated differently from other cases of insanity, like that which might be brought about by serious mental illness or a brain tumor? Most of the ideas here seem to be equally applicable to both kinds of cases.

[Daniel Cohen]

Here’s a diagnosis of the problem of manipulation that construes it as emerging out of an inconstant triad of independently plausible claims. It follows that any resolution of the problem will necessarily involve biting some bullet or other.

This is the inconsistent triad:

(1). Manipulated agents are no less responsive to reasons than regular people.
(2). Responsiveness to reasons is sufficient for responsibility.
(3). Manipulated agents are not responsible (as opposed to regular people).

Giving up (1) is implausible because manipulation seems describable in such a way that while the manipulated agent’s values have been changed, their disposition to correct their values in light of the evidence remains unaffected. Indeed, cases may be described where an agent’s sensitivity to reason is improved as a result of manipulation.

While Fischer and Ravizza don’t accept (2), complications of this simple claim seem rather ad hoc. Proposals that involve adding further necessary conditions for responsibility seem to undermine the inherent plausibility of reasons-responsiveness as explicating the nature of responsibility.

(3) is hard to deny for familiar Strawsonian reasons. Our intuitions about responsibility derive from our intuitive distinction between those who are and those who aren’t appropriate objects of the reactive attitudes. Manipulated agents just don’t seem appropriate objects of the reactive attitudes.

[Eddy Nahmias]

Putting the problem in the form of this inconsistent triad is very helpful. Though I can't develop the idea fully here, one possibility is that any manipulated agent who is actually responsive to reasons in the way normal adults who count as responsible are (according to the criteria of 2, and hence they really satisfying 1) should in fact be considered responsible (contra 3). The reason 3 seems so intuitive is that no real world examples of manipulated agents are in fact appropriately reasons responsive. Raymond in the Manchurian Candidate is a good example: he is directed to kill the only love of his life in such a way that he does not stop as he sees the life drain out of her (he had as strong a reason imaginable to stop killing her and he did not). (By the way, the movie is pretty good, certainly well acted and very creepy, though there's something I feel it's missing from the original).

As often happens with intuition pumps about free will, they pump our intuitions by messing with the context or presenting things in a misleading way. Even us clever philosophers cannot overcome a lifetime of seeing brainwashed people act like zombies and thinking there's no way such people should count as responsible. I know I can't shake that intuition. It's hard to know what our intuitions would be (should be?) were we to find out that someone we've known and loved for a long time who seems entirely normal and responsible is in fact a manipulated agent (or a robot--I haven't seen Stepford Wives yet--anything interesting there?). We'd surely be creeped out but would we think the person (?) was no longer responsible? (Wittgenstein suggests we wouldn't know what to think if we found out we were all mechanical robots--similarly, a la van Inwagen's supposedly devastating response to the Paradigm Argument, what would we think if we found out that all humans had always been directly manipulated by Martians?)

One reason we may treat a manipulated human (or robot) different is that we may no longer think of them as *persons* at all but for some reason tangential to issues of responsibility. This possibility suggests another familiar response to Daniel's triad: 3 is false (manipulated agents aren't responsible) because manipulated agents are not persons at all. Perhaps to be a person you have to have a history of development in an appropriate social and physical environment (so Swampman is not a person). Or perhaps to be reasons-responsive 'for real' you have to have engaged with the features of the world that provide reasons rather than just having them artificially induced. These responses may be ad hoc but they are very intuitive (to me at least). OK, enough for now.

[Joe]

The above comments are all very helpful, especially Daniel Cohen's triad. Here are a few points that I have in response.

It is unlikely that manipulation cases can tell us much about the freedom relevant condition for moral responsibility, and for this reason this particular case does not pose a serious problem for the Fischer/Ravizza view.

Suppose, for instance, that one offered the case of Raymond from the Manchurian Candidate as a counterexample to the libertarian agency theory. After all, there is no reason to think that manipulated-Raymond is not the cause of his actions. It is doubtful that anyone would consider this to be a counterexample to the libertarian agency theory. The agency theory is just a theory about the freedom condition for moral responsibility, and the problem might have something to do with one of the other necessary conditions for moral responsibility. Fischer/Ravizza can make a similar response.

It is also important to note that manipulation cases are widely varied. Persons are often manipulated by false information. Perhaps Raymond was given the false belief that person S was Hitler together with the belief that it was morally permissible to kill Hitler whenever one had the chance. It seems that Raymond would not be responsible for killing S but the reason for this has little to do with whether or not he acted freely.

In a lot of manipulation cases we are not given enough information to explain exactly why the agent performed the action that he performed, and the explanation might vary widely depending on the precise mechanism involved. What about the fact that Raymond was ‘brainwashed’ tells us whether the brainwashing affects his level of freedom or his knowledge, whether it renders him insane or simply ‘not himself’?

One last point, against (3) in the Cohen's triad. It is not the case that manipulated agents are never responsible for their actions. Suppose that Clinton was set up and that Monica Lewinsky was placed in the White House by Republicans who knew that the President had a weakness for women with big hair. Clearly, Clinton was manipulated but that does not mean that he was not responsible for his actions.

It all comes down to the details: Precisely how was one manipulated? What was the mechanism involved? Why did he do the things that he did? In a lot of manipulation cases these details are left out and for that reason a single, clear answer to the question 'Why is he not responsible for his actions?' cannot be given.

[robert allen]

Daniel,

For those of us who do not accept F & R's insightful analysis of free will, abdicating 2 of your triad is not "biting a bullet." We consider manipulation cases to be straightforward counterexamples to their view. (Cf. Tomis Kapitan, "Autonomy and Manipulated Freedom," in Phil. Perspectives 14, 2000.) I would further suggest that such cases show that free will entails not just reason-responsiveness, but responsiveness to reasons that are sheddable in favor of motives unintended by one's influences. You are right, though, that were F & R to add such a condition it would be "ad hoc" and "undermine the inherent plausibility" of their account. Those of us who do not consider r-r to be sufficient for responsiblity, on the other hand, are not liable to this charge.

[Manuel]

I'm basically in agreement with Eddy and Joe about this thread. I think that different manipulation cases should be handled differently, and that we should be careful to think about how the manipulation cases work. I'm inclined to think that if you have a case with all the features of your favorite theory of moral responsibility (ignoring manipulation), whether the agent was manipulated or not should be irrelevant to determining whether you have a responsible agent. Similar things can be said for the narrower issue of free will.

One case that can be useful to think about is the case where you 'manipulate' an agent into having all the features required for freedom and responsibility. Some might be inclined to treat that as a counterexample to the above view, but I'm inclined to think child-rearing is an example of turning a non-responsible agent into a responsible agent through lots of cajoling, manipulation, etc.. Some ways of doing this are likely less and more morally desirable (shocking your kids with cattle-prods vs. versus rewarding good behavior) but if it all works out right, you get a responsible agent out of it.

(Warning: self promotional remarks follow!) I discuss manipulation cases in more detail in a forthcoming paper "On the Importance of History for Responsible Agency" which you can get off the papers on agency website. Also, Chris Grau has a paper there that is a response to a prior version of the paper, so you can see what may be wrong with my paper.

[Robert Allen]

Manuel,

Do you not discern a morally important difference between “Manchurian” conditioning (MC) and parenting, viz., that the former but not the latter would leaves its subjects controllable by others? I prefer to look at the matter this way:

1.If MC negates a free will (FW), then either there is a morally important difference between MC and parenting or the latter also negates a FW.
2.MC negates a FW.
3.Thus, either there is a morally important difference between MC and parenting or the latter also negates a FW.
4.Parenting does not negate a FW.
5.Thus, there is a morally important difference between MC and parenting, viz., children can subvert the wills of the parents, the subjects of MC can do nothing but what their conditioners intend.

If I have been "manipulated" into acting responsibly, then I simply cannot act irresponsibly, that is, against my manipulator's wishes. Thus, my conditioner is solely responsible for my "responsible" behavior. Why should I receive any credit for how I act? Granted, a parent deserves some credit for the fact that his/her childeren are responsible adults. But, since they are out of his/her control- they could act against his/her wishes- they should share in the praise. Bottom line: MC could not be used to create a responsible agent, if by that you mean someone deserving a measure of praise/blame for his/her conduct.

[Joe]

Robert,

(Apologies to Manuel and Robert for jumping in!)

I agree that there is a morally relevant difference between “Manchurian” conditioning (MC) and parenting. I indicated earlier that I'm not at all sure of the exact processes underlying MC, so I’m equally unsure about what constitutes the difference between the two cases. But I am sure that the difference is not that "children can subvert the wills of the parents [whereas] the subjects of MC can do nothing but what their conditioners intend." (I assume that ‘can’ here is meant in the usual all-in ability sense of the word.)

Suppose that agent X is conditioned to kill person Y unless X happens to raise his left hand at time t, and that X can raise his left hand at t. Otherwise, it is highly probable (though not deterministically necessary) that X will kill Y. Unfortunately X has no idea about the connection between the raising of his left hand and the (non-)killing of Y. Thus, X doesn’t raise his left hand at t and he does kill Y. It seems that X is still not responsible for killing Y.

On the other side, parents often condition their children by locking them in their rooms, or by severely limiting their choices in other ways. If I get my son to study for an exam by threatening to revoke his driving privileges, does that mean that I cannot praise him for getting an ‘A’ on the exam? I don’t think so. (My own son is only 5 but I imagine that I will have to resort to such tactics one day.)

It seems that there is more to differentiating MC from parenting than what you’ve suggested above.

[Robert Allen]

Joe,

The agent in your case cannot subvert the will of her conditioners, as she does not know what it takes, viz., raising his left-hand.

Threatening a child with a loss of driving priveleges is a far cry from MC. The fact is, we cannot MAKE our children do the things that we want them to do; we lack the power of Manchurian conditioners. (As my mother used to say, "I can't make you, I can only make you wish you had.") Were you to use such power to induce the studying, it would get done no matter what. The threat, on the other hand, would be effective because your son is rational, which is why he would be praiseworthy. However, given that praiseworthiness comes in degrees, an agent who does the right thing out of fear appears less praiseworthy, than someone who does it out of a sense of duty

[Robert Allen]

Joe,

I meant to say that "the agent ... cannot subvert the will of his conditioners, as he does not know what it takes ...."

[Joe]

Robert: I agree with everything that you're saying above. But then it follows that when differentiating MC from parenting we have to consider more than just what the agents can or cannot do, e.g. we need to take in to account epistemic considerations and issues about, say, origination.

By the way, I finally saw The (new) Manchurian Candidate last night. It was pretty good, though not as good as the original.

To back up some of the points that I made previously, the new version provides few details about the processes involved in MC. People undergo a variety of different physical and psychological treatments. They are ‘brainwashed’ and have ‘implants’ inserted into their brains and placed under their skin. They are given false memories. But the details are left out.

Raymond Shaw (Leiv Schreiber) does whatever anyone says so long as they first utter these words: “Sergeant Shaw. Sergeant Raymond Shaw. Raymond Prentiss Shaw.” But there is no information at all as to how or why Shaw is controlled by others. In particular, it is not clear whether he kills on the basis of false beliefs (in which case he might well be free yet still not responsible for his actions) or for some other reason.

Furthermore, Shaw’s behavior does not seem to result from a mechanism that is responsive to reasons. He is told to do certain things and he does them. It is not clear that his own reasons ever get to play a role. I don’t see how one could use this new film as the basis for an attack on the Fischer/Ravizza view. (This is not a direct comment on Stump, for I have not read her articles yet and, I assume, she does not use this new film as the basis for her criticisms.)

There are other parts of the film that are worth mentioning here but I’m reluctant to do so since it would give away some surprises. Do we have a policy about spoilers yet?

[Robert Allen]

Joe,

I prefer to think of the knowledge in question as a part of the ability I am identifying as free will. An agent cannot subvert a will unless she knows its contents. In granting Adam and Eve free wills, God made His will perfectly clear: "do not eat of the fruit ...." But suppose He had not made this (rather simple) request yet nevertheless indended for them to refrain from eating that fruit. Subversiveness would not have been possible. Again, to set someone free of you, you must let them know what it would take to become free of you.

Origination as described by Honderich, Strawson, and Pereboom is incoherent: a person could not have created herself; either she is eternal or was created by someone else. The idea of being free of one's creator, on the other hand, does make sense. Thus, the latter but not the former is a variety of free will worth wanting.

[Joe]

Robert,

I think that it is important to separate the many other distinct necessary conditions for moral responsibility from the freedom relevant condition. But that is beside the point. If you want to group the epistemic condition with the freedom relevant condition, then the (new) Manchurian Candidate provides a challenge for the Fischer/Ravizza view only if we can be certain that Shaw's actions are not the result of false beliefs or ignorance. I don't see how we can be sure of this.

I agree that the Honderich/Strawson/Pereboom concept of 'origination' is incoherent. I intended to use the term 'origination' in a more neutral way. I wanted to articulate a condition that would account for the fact that my son has the "power to induce [his own] studying" whereas I do not. At least I don't have this power unless I am capable of something like MC. That he has the power and I do not seems to be a necessary condition for his being responsible for getting the ‘A.’

I'd be interested to know whether you think that this, too, is part of the freedom relevant condition for moral responsibility.

As always, thanks for the comments, Robert!

[Robert Allen]

You're welcome, Joe.

So what you are saying is, if I make the epistemic condition part of the freedom relevant condition, then F & R are back in the ballgame, no longer sidelined by a case that my view (alone) can handle. But why should I think that someone can’t be r-r while entertaining falsehoods and half-truths?

Your definition of ‘origination’ is an eminently sensible alternative to the one that the nihilists would have us believe. (It reminds me of Reid's notion of our "active powers.") And, yes, I would incorporate it into my “freedom relevant condition for moral responsibility.” One would be incapable of subverting another's will were one to lack the power to induce an attempt at doing whatever that takes. Such would have been the case had God only allowed A & E to think about eating the forbidden fruit.

[Manuel]

Robert-

This response is a little late (I'm still in the process of moving), but sure, I'm willing to allow that there is a difference between Manchurian conditioning and many instances of parenting, though I remain unclear on what, exactly, Manchurian conditioning amounts to. That said, my point was a different one- I was merely suggesting that the fact of manipulation, by itself, shouldn't be enough to undermine responsibility. What matters (I think) is the impact of the manipulation. So, of course, if manipulation renders the agent incapable of reasoning, shedding beliefs, or whatever cluster of capacity you think are necessary for responsibility, then sure, manipulation undermines responsibility. But the reason for the undermining isn't the bare fact of manipulation, but rather, the impact of manipulation of (what I call) the basic agential structure of responsibility.

[Robert Allen]

Manuel,

So our dispute is merely a semantic one: I take manipulation to be anything that renders an agent incapable of shedding values, so as to be controllable by another. If something fails to have that effect on an agent, such as ordinary parenting, then it's not manipulation.