Survey Says...

There have been 107 responses to yesterday's survey, but since I'm too cheap to "upgrade" my Survey Monkey account, I get to see stats on only the first 100.  So don't bother voting any more.

At any rate, there was a significant majority of respondents who voted "No."  Specifically, 62%, with 38% responding "Yes."  I suppose I could have added a "Not sure" option, or a "degrees of confidence" meter, but I didn't.

From the few comments that came in as well, perhaps some important details from the prompt were missing.  I threw in the "drunk" comment only as a (very minor) joke, not anything that should be thought to undermine Max's reliability at replacing intentions (which he can do easily and always successfully and by whatever means you're happy with); it was only a way to explain his being mistaken over the identification of the relevant intention.

Here was the point: manipulation cases work on our intuitions typically by having some external manipulator replace one intention with a very different sort of intention.  I was wondering if people thought, then, that it was the manipulation itself that undermined attributability, rather than the content of that intention.  What was especially interesting about my case, I thought, was that Max was guided by precisely the same sort of considerations by which I was guided, namely, the reasons there were in favor of punching (although this might not have come out clearly enough in the prompt).  If what guided me in the formation of some intention also guided Max in the intention he implants in me, and those intentions have identical content, then why wouldn't the subsequent intention be attributable to me?  Apparently, though, many people think otherwise: the manipulation alone is sufficient to undermine attributability.  I guess I am a bit surprised by this result, although again, it may also be a product of insufficient clarity in the prompt.

Thanks to all who participated.

Comments

[Lance]

I am reasonably confident that the question was sufficiently clear that I did not misunderstand it in any substantive way. But perhaps I overlooked something.

I'm assuming a single-question survey like this could be used to gather some ideas on what to develop for a more substantive survey; what I would be curious to see is a range of questions intended to pump the same intuitions and therefore yield similar results, and to present a set of these questions to different people.

I'd be curious to see if the questions would elicit consistent responses to subjects (although it'd be hard to tell if inconsistencies were due to the wording rather than something substantively different about the questions) as well as to see how people respond to attributility in general.

[Eddy Nahmias]

I'm not sure what's pushing my intuitions in this case (that's why I like to do x-phi by creating variations of scenarios to see what seems to push my intuitions around and then test whether it does the same for most people). But I suspect it has something to do with what I think drives people's reactions to manipulation cases--the sense that the manipulated agent loses "teleological control" over their actions. Their own intentions and actions are not responsive to their own goals, character, reasons, plans, but instead to those of the manipulator. Had you wanted to hug the guy instead of hit him, what would Max have done? Given his current state, it's hard to tell. But what if you describe the case much as you did, but Max really only replaces your intentions if they conflict with ones you form in conflict with reasons you would endorse (so, for instance, he helps you avoid weak-willed action)--and he's not drunk. Then, I'm not sure what my (or others') intuitions would be in a case (1) where we asked whether you act freely and responsibly when Max has to replace one of your compulsive or impulsive intentions with one you'd rather act on (if we asked whether the intention was attributable to you, I suspect the answer is 'no' because, as in the case you ran, the immediate and effective cause came from outside of you); and (2) where we asked whether you act freely and responsibly when Max tests his powers, as he sometimes does, by replacing one of your endorsed intentions with the same intention (if he does this enough, he starts to seem more like a Frankfurt intervener than a manipulator, and perhaps his intervention looks more like an overdetermining cause such that your replaced intention is attributable to you?).

[David Shoemaker]

Thanks Lance and Eddy. Eddy, I have in mind an overdetermining cause, in a way (with only a single determiner, though: Max). I would have acted on the intention I had, except that it was replaced at the last second with an exactly similar one.

I'd be curious too, Lance.

[Tamler Sommers]

Wow, I'm surprised that 62% have such obviously wrong intuitions. Probably those Pea Soup people mostly...

[Brent]

There should've been an option for attributability being impossible in the first place! No one else seems to come down hard on these manipulation cases for assuming attributability is plausible. Or is this one just for the compatibilists, in which case I’ll go wait quietly on the sideline.

[David Shoemaker]

There was such an option, Brent. It was "No."

That was the real point of the survey, Tamler, to see just how many people would get it wrong.

[V. Alan White]

David, my luck comment in the previous thread was motivated by the fact that the reasons for the would-be Max intention and the actual manipulator intention are the same: as luck had it, Max acted on his actual reasons by lucky means that accorded with and reflected those same reasons, even though those same reasons flowed from a different causal track (the luck was reflected in the manipulator's fallibility about Max's intention, and not the means of influence of such; she might have drunkenly made other mistakes about Max's intentions, such as mistakenly thinking that Max had no intention to act at all when in fact he had). The manipulator did not replace Max's reasons (as a type), but only instantiated them in a different intent (as a token). I assume here that the manipulator did not respond on the basis of assessing Max's reasons (by the cited fact of her misreading only Max's intention), and so she only luckily reflected them. An analogy for my perspective on this might be a case of occasionalism of action where a third party like God is a necessary condition of instantiating a physical action expressive of and arising from a particular person's will. As you can surmise, I voted against the majority on this one.

[David Shoemaker]

Some ambiguities in the original prompt have come out in the discussion on PEA Soup. I'm wondering, then, if the following change to the prompt would cause those who'd originally voted "no" to change their minds.

Suppose Max is an ensuring manipulator: he is caused by my judgments to give me an intention to do whatever I judge there to be decisive reason to do, in cases where there's a screw-up on my part. So when I'm weak-willed, say, Max replaces my weak intention with the one based on what I've already judged I have a decisive reason to intend. In the current case, though, he's wrong: he thinks I've formed an intention to hug, when I've actually formed an intention to punch. He then replaces my intention to punch with an exactly similar intention to punch. Is it now attributable to me?

[V. Alan White]

Whoops! My comment reversed the labels--Max is not the manipulator in them. Sorry for the confusion--I should have double-checked my memory of the example. But I still stand by the points.

[James Laird]

The fundamental problem with the survey is that an impossibility is assumed to be possible in the scenario (i.e., an intention may be cloned and/or replaced). What if we advance Max’s capability to the next *slightly* higher level, where he's able to clone an entire physical human being at one instant of time? Then our survey could ask the reader if the intentions of the clone are attributable to the original human. Generally speaking, I’m thinking that survey results for impossible scenarios don’t provide significant value.

Manipulators don’t replace intentions; instead, they exert forces that affect intentions. A person’s intentions are attributable to many factors, one of which is the “person” who has the intention. (The definition of “person” is the next complexity that enters the conversation...)

The scenario was fun to think about!

[Neil Levy]

Since my no vote was driven by concerns over reliability, I think I would change it ( not so much on the basis of the reformulation as the earlier information I was not supposed to take the drunkenness seriously). But I still can't come to a firm view of the case without information about reliability. Alan's luck comment is right: it depends on the degree to which the match I'm intention is chancy.

[Chandra Sripada]

Cool case Dave.

Here is my take. I think the case is non-standard in important ways that make it hard to interpret. Standard cases of non-coercive manipulation leave it unambiguous that the agent does in fact *produce the action*. We are then asked to assess a further question: Does the action (which the agent clearly produced) belong to the agent in the distinctive way that attributability requires? This amounts to assessing whether the action reflects, in the right sort of way, the agent's deep self.

Put another way, the first question is asks "Is the action Dave's (as opposed to someone else's)?" The further question says "Given that Dave did it, does the action *really* belong to Dave?"

In your case, I am unable to answer the first question affirmatively with any confidence. Now, I am not ready to supply a full account of when an action is or isn't produced by you as opposed to someone else (it probably involves causal connectedness, the operation of "normal" sorts of mechanisms, and the like...) But this question is not really about attributability itself, but rather about whether key *preconditions for attributability* are met.

So I think it would be a mistake to read intuitions in this case as informing us about the nature of attributability.

[David Shoemaker]

Glad you got involved, Chandra, as I was wondering what a view like yours might say about the case. Nothing yet, it seems.

I like to distinguish between shallow and deep attributability. Shallow refers to whether there's the right dependence relation between an action and a will/intention; deep refers to whether there's the right dependence relation between a will/intention and a deep self. I think as long as some will/intention is internal to my psychic domain (regardless of its source), the action is shallowly attributable to me. So in this case I would think the action is shallowly attributable to me. (This is true, e.g., of the unwilling addict too: the action is shallowly attributable to him.)

I don't think we're disagreeing too much here; what you're calling the preconditions for attributability, I'm thinking of as one kind of attributability, viz., the shallow version.

I guess I'm surprised, though, that you're having trouble assessing shallow attributability here. Do you have the same worries about the unwilling addict (substitute the addiction for Max as providing the "alien" source of the intention)?

[Chandra Sripada]

Hi Dave,

I like the shallow/deep attributability distinction, I think (I need to think through the implications more).

I guess what I am saying is that the intention implanted by Max is not after all part of *your* psychic domain. That is, it is not shallow-attributable to you.

Why? Several red flags are present. It lacks the right kind of causal-cum-psychological connectedness with your prior states. The mechanism by which it arose is not "normal". There is a chancy aspect to its origin (others have mentioned this as well). The implanted intention (unlike your original intention) is not causally connected in the right sort of way to a motivating reason (a belief desire pair according to the leading theory) that rationalizes it. Some of the preceding points are no doubt overlapping. But the point is that all these red flags go up. And none of these red flags are about deep attributability. So I think the case is really about shallow attributability. That may be your point all along.

As for the unwilling addict, I don't think these red flags go up. There are no questions about causal-cum-psychological connectedness, etc... The craving clearly is part of his psychic domain. But it starkly opposes his deep self. The craving is shallow- but not deep-attributable to him.

[Neil Levy]

Actual human beings are set up roughly as follows: the anterior cingulate cortex has the job of detecting conflicts between impulses towards actions and the person's values and commitments. These conflicts may be caused by the formation of an impulse resulting from a temptation or a mere response to an affordances (knives are for stabbing). All going well, when the ACC detects a conflict between the person's values and an impulse, it triggers the dorsolateral prefrontal cortex which overrides the offending impulse.

With human beings living so much longer in the 23rd century, brain parts routinely wear out and are replaced. Scientists developed MACCS - the Mental Attitudes Conflict and Control System - to replace the ACC/DLPFC complex. MACCS, or Max as it came to be called (Max utilizes AI, and can pass the Turing test), can be set so that people become ideally self-controlled, always acting in accordance with their values and commitments. But many people prefer Max to be set to replicate their biological ACC/DLPFC complex. Max is available as an app running on people's iDevices. It's a friend, it's a companion, a personal shopper and - in the background - a self-control system.

When Max detects a conflict between a person's values and impulses, he (often) overrides the impulses and forms an intention in line with the person's actual values. In this, he mirrors biological self-control resources exactly (if he is set up to). Due to a quirk in the AI, sometimes Max is triggered when there is no conflict. When the agent forms an intention that actual conforms with their own values, sometimes Max thinks that they have formed a conflicting intention, and replaces that intention with a new one which has exactly identical content. This quirk in Max's programming is well-known. It would be cumbersome to eliminate it and would result in a slowing of performance. Since the quirk does not alter the input/output relations - since Max as he is is functionally equivalent to Max as he would be were the quirk eliminated (except that he performs slightly more efficently than he would were it eliminated), the scientist leave the quirk untouched.

Are they making a mistake? Are they undermining the attributability of millions of actions performed by hundreds of thousands of people? I don't see why we should say that.

[David Shoemaker]

Neil: Brilliant, I love it. Let me know when the whole novella is complete.

Chandra: We should talk about this some more in person. I wonder how that could possibly be arranged? ;-)

Nevertheless, I was actually concerned in the original post about deep attributability, although I can see how you might be worried about the shallow version. We might think of that as a term of art, though, the articulation of a necessary condition for deep attributability: at the very least, the action must depend on a will internal to my psychic domain. This is how we get shallow attributability for the unwilling addict, depressed patients, Tourettic individuals, OCD patients, and others. Their wills are internal to their psychic domain, even though those wills might have an alien (internal) source. But I don't think shallow attributability is at all sufficient for moral responsibility of any kind; instead, it's sufficient only to answer the question "Did you *do* that?" Deep attributability, on the other hand, may well be sufficient for responsibility insofar as it answers the question "Did *you* do that?"

Thanks to everyone for their participation. I'll be more or less off-line for the next five days, so this will have to be my last word for now.