Causation vs. Prediction: An Experimental Study

Luke Misenheimer, a student at UNC, sent the following contribution.  Thanks Luke!

I recently became interested in experimental contributions to the free will debate, and I decided to run a study to investigate a possible difference between the results of some earlier studies. Using questionnaires, I told subjects about an imaginary (more or less) deterministic universe, and then I asked whether a particular action in that universe could have been performed with a free will.

My questionnaires were based on those used in earlier studies, and in constructing them I wanted to make sure that I really drove home the determinism to subjects. In addition to describing the imaginary deterministic universe, my questionnaires did several things to make sure that subjects really got the deterministic message: - they italicized key deterministic phrases; - they reiterated points about the determinism of the universe both generally and with reference to a particular action; - they asked subjects whether an example action in the universe was determined, and I ignored the other responses of subjects who answered in the negative; and - they had subjects compare the imaginary universe to our own universe, which most subjects took to be indeterministic.

My study had two kinds of questionnaires, both of which did all these things. Both questionnaires described a universe in which things that happen around the time an agent is born determine all of that agent's actions. The only intended difference between the questionnaires was this: according to one questionnaire, these things completely cause the actions; according to the other questionnaire, these things can be used to perfectly prediction the actions. The full questionnaires can be found here.

I was worried that all the mechanisms designed to hammer determinism into subjects' heads would cause subjects given both questionnaires to give equally incompatibilist responses (or to give slightly different but still incompatibilist responses). Instead, subjects given the causation questionnaires tended to give *incompatibilist* responses, and subjects given the predictability questionnaires tended to give *compatibilist* responses! So: what, if anything, do you GFPers think this difference between causation and predictability means about the prephilosophical concept of free will?

Comments

[Andrew Lee]

It sounds like the results support the view that folk psychology is a theory we employ to predict others' behavior by placing them in the "logical space of reasons". When your subjects heard that the actions were "predicted by" the prior state of affairs of the world, they could still find a way to place them in reasons-space (and thus attribute moral agency to them); but when they heard that the actions were unambiguously "caused by" prior states of affairs, this was enough to boot them out of reasons-space.

[Kip Werking]

I find the results so counter-intuitive that I would want to run the study again and try to replicate the findings. It also makes me wonder whether I am using "free will" in an idiosyncratic way. As I understand free will, it is violently opposed to perfect prediction: whatever free will is, it's incompatible without someone designing my entire life. So prediction would be very problematic and causation would not necessarily be any problem.

As much as I hate to admit it, these results seem to support the compatibilists' error theory: people confuse causation with compulsion. When you just say that one could predict another's actions, perhaps the folk don't take this to mean "compulsion" so much.

It is interesting that the study uses a morally salient event: embezzling money. The Nichols & Knobe finding suggests that, at least when the stimulus and question are framed in a certain way, such emotional and immediate situations trigger compatibilist responses. It would be nice to perform the experiment again with a morally neutral action (e.g. riding a bike) and see if the compatibilism holds.

It's good to see more data (on top of Nichols' recent research) supporting the folks' weak intuition that the world and human choice are indeterministic.

[Mark]

I think this survey points to the difference between these two cases:

1) At time T agent A began to exist as a bundle of prior causes and effects and the actions A performed after T were dictated by relations between that bundle of prior causes and effects and the situation at hand.

2) At time T agent A began to exist as the result of prior causes and effects and the actions A performed after T were dictated by relations between who A is and the situation at hand.

Perfect prediction is possible of the agents in both cases (by stipulation), but there is a sense in which the mechanistic causal account in (1) is insufficient to fully describe the agent-events after T in (2): agent-events in (2) issue from the conjunction of the agent and a situation, whereas in (1) agents themselves are reducible to (mere) situations and thus agent-events in (1) are also reducible to (mere) situations.

In case (1) the relationship between the past and the future suggest an analytic read, but in case (2) the relationship between the past and future suggest a synthetic read. This distinction is what I believe gives rise to the notion that there is room for perfect prediction and freedom, but not mechanistic causation and freedom.

So the real question is whether determinism entails the analytic read suggested by case (1) or the synthetic read suggested by case (2).

[Eddy Nahmias]

Kip, nothing in Luke's "prediction" scenario suggest the scientists are manipulating the Nuhams or that your intuitive fear of "someone designing my entire life" is being fulfilled. Prediction and manipulation may be related in practice but are not the same thing.

Some folk may not worry about prediction so much because at least they may have reconciled their belief in free will with the idea of an all-knowing God who knows what we will do in the future (though he does not manipulate what we do). Looking at people's religious views on these surveys would be helpful. On our current round of surveys (describing a world of Ertans rather than Nuhams!), we get info on religious affiliation, and we are about to run some scenarios with neutral actions (mowing the lawn) and some positive actions (giving money to an orphanage).

Luke, do you mind giving some more specific info on your data? (If you are waiting for publication, that's understandable.)

[Kip Werking]

Eddy,

That's a good point. I was confusing this survey with the Zygote Argument.

[Corey Tomsons]

Another online study about causation can be found at:

http://www.unc.edu/%7Eknobe/causation/index.htm

At least a couple GFPers will be please to know the Leiter Reports picked up on this. How long will the study be active?

As an aside, today CBC Radio had a discussion of neuroscience and free will. I've collected links to the broadcast here .

[Charlie Huenemann]

I don't mean this to sound in any way mean-spirited - I'm just curious -- but could someone direct me to a discussion of the relevance/importance of "experimental philosophy"? To be blunt, why should philosophers care about unrelfective opinions?

[Luke Misenheimer]

Andrew: I believe that folk psychology is something we primarily use for prediction, and that we don't normally think of folk psychological objects interacting causally. I think that, in believing this, I am agreeing with what you posted, but I'm not sure whether you're saying something stronger by talking about "reasons-space". Are you talking about reasons only because they're the folk-psychological objects necessary for moral agency (and thus free will), or is there some deeper reason?

Kip: My intention was to use a morally salient event that would not generate strong emotions in subjects. I figured that my embezzling money would be emotionally equivalent to Nichols & Knobe's cheating on one's taxes (which got mostly incompatibilist responses). I guess the harm from embezzling isn't usually distributed across as wide a population as that from cheating on one's taxes; does that (or something else) make embezzling seem more emotional to you?

Eddy: Responses to the question about free will were scored on the standard scale of 1 ("strongly disagree" with the possibility of a free will) to 7 ("strongly agree"). The mean response in the predictability condition was 4.56, with 63% of responses on the compatibilist side of neutral. The mean response in the causation condition was 3.38, with 30% of responses on the compatibilist side of neutral.

Charlie: I take the philosophical importance of this experiment to be its illumination of the prephilosophical concept of free will. Philosophers who work with free will might have reason to worry that their concept of free will has been influenced by their philosophizing; I'm pretty sure my concept has become significantly theory-laden. When I ask myself whether I have free will, I am interested in whether the ordinary everyday concept of free will applies to me -- not in whether the revised philosophical concept applies.

[Charles Edwards]

It's begging the question to assume what is to be shown. Once you've assumed time to be linear and causation to be intuitive, you'll certainly able to show that time is linear and causation is intuitive. If the putative intuition of the meaning of causation corresponds to the current notions of the scientists who follow a materialist's belief, then personal identity can never cause anything - they've already assumed personal identity "doesn't exist" therefore everything is caused by something other than the identity who is evaluating.

[Charles Edwards]

Wow! Look, there's tacit acknowledgment that something can "begin to exist". I wonder if it's possible to have "free will" begin to exist. If so, can we predict what it would be, at which point would it deterministically exist? It also appears contradictory to assume the past exists as a cause, while claiming it doesn't exist in the present. If it doesn't exist in the present, then how can it cause anything? If it does exist in the present, then why are we calling it the past?

[Andrew Lee]

Luke: no, I didn't mean anything stronger in particular by that locution. I think the act of holding something in the universe morally responsible just *is* the act of imagining it having acted according to a reason. What I think surveys like this test are simply the willingness of unreflective people in certain very controlled circumstances to exercise that imaginative faculty.

This brings me to your response to Charlie. You write, "When I ask myself whether I have free will, I am interested in whether the ordinary everyday concept of free will applies to me -- not in whether the revised philosophical concept applies." I suppose it is a matter of taste what questions one finds interesting, but isn't it really the case that what you're interested in is whether you have moral responsibility, regardless of what the hoi polloi think? People's "ordinary, everyday, unrevised" concepts tell us that heavier objects fall faster than light objects, and that your chances of winning the Monty Hall Problem don't increase by switching doors. I can understand being interested in why people think such thoughts, but it seems bizarre to be pegging your moral self-conception to the imaginations of others, just as it seems bizarre to be doing physics or probability by polling.

[Luke Misenheimer]

If I knew how to test evaluations of moral responsibility as rigorously and surely as I can test the predictions of physics or probability theory, I'd love to do it and leave polling behind. But the only way I know to evaluate moral responsibility is to use the moral theories that I'm given, which often involve free will as a requirement for moral responsibility. If the theory tells me to test for free will (and doesn't further specify "free will"), the concept of free will becomes interesting to me. If, though, a theory of moral responsibility doesn't mention free will, this kind of polling is indeed a bizarre way to test for it.

And maybe I have the wrong concept of concepts, but if folk physics predicts that heavy objects fall faster, I'm inclined to say that there are problems with the theory, rather than that the folk don't know which objects are heavy and which aren't. I would guess that the ordinary concept 'heaviness' is something more like 'the force with which something pushes down on my hand when I hold it up' than 'the speed with which something falls when not held up'. Engineers can have useful conversations about heavy and light things with the folk because they share the concepts, even if they don't share the theory.

[Paul Torek]

I think Andrew Lee and Mark are barking up the right tree. My variation on their theme is to suspect that your subjects implicitly see causation as more exclusive than it actually is. In its worst form, this error has it that if something is caused by X, then it's not caused by Y, for any X and Y. I'd like to see what the subjects say to the questions: was the action dictated in part by who the agent is (Mark's version)? Was the action done for the reasons that the agent took to be her reasons (Andrew's)?

[Charles Edwards]

The meaning of "cause" is of course problematic. However, most agree, that if X causes Y, then Y does not cause X. Determinists theorize the past causes the present. Yet, there is also wide agreement that for the past to exist it must have existed first as the present. In other words, first the present exists, then it becomes the past. Of course, some pasts occur before or after other pasts. But there is only One present. We can however claim of any past that it was once the present. It is strange thinking to suppose the thing that doesn't exist is the cause of what does exist. For example: What one writes causes a sentence. It is not the case that the sentence causes what one writes. If anything can have an origin in the present, then we can begin to understand that the origin of something can be in the present. Isn't it true that the origin of any past was it’s once (but no more) relevant present? Don’t we all experience the present as the foremost and necessary criteria to have any experience at all? In that we exist in the present, we have a greater potential to be the cause of a past. Those things which do not exist in the present have no chance at all.

[Luke Misenheimer]

Paul: I definitely would expect subjects to see causation as more exclusive than prediction (or causes as more exclusive than predictors). The causal story I tell subjects presents a complete chain of causes back to pre-birth events, and I agree that a likely explanation of subjects' incompatibilist responses is their inability to fit something like agency into that causal chain. My guess is that subjects think of causation as only properly applied in the physical domain, at least when it comes to causal chains of events, and so subjects (most of whom I'd expect to be dualists) can't fit anything mental at all into the causal chain I tell them about. So I'd predict that subjects would respond in the negative to both of the questions you propose, and also even to questions like "Did the agent's choices or beliefs or desires play any role in producing the action?" (Of course, asking these questions without changing the way subjects think about the story is tricky.)