Blog Coordinator

Knobe's X-Phi Page

X-Phi Grad Programs

« Measuring Moral Judgment | Main | Vacancy for PhD students: Measuring Moral Relativism »



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

Josh May

I'm really glad to hear about these results, David! In a paper of my own, I complain about this very issue with Murray & Nahmias's article.

In fact, there is some evidence in their own data that people are reluctant to grant the ability to decide in a deterministic universe. As I say in my paper:

"After reading the crucial abstract case [from Nichols and Knobe] (which seems to yield strong incompatibilist intuitions), only 31% of Murray and Nahmias’s remaining subjects agreed that a person in that universe has "the ability to decide" to do something other than what they actually decide to do (and only 58% for two of the other four scenarios)."

(Sorry if you make this point too, but I haven't yet had a chance to read your paper.)

And all this suggests that one of their bypassing questions problematically prejudges this issue:

- "No Control: In Universe [A/C], a person has no control over what they do. (Bill has no control over what he does.)"

M&N require "competent" participants to disagree with this statement. Yet, as I say in my paper, if the ability to decide is required for control, then it isn't clear that disagreeing with this statement to some degree is a misinterpretation of determinism. And if incompatibilist intuitions in particular tend to generate the denial of abilities to decide in a deterministic universe, then M&N are selectively throwing out incompabilist responses as incompetent without warrant.

Here's the link to the draft of my paper if interested:

David Rose

Thanks for posting the comment and link to your paper Josh. And glad to hear that we're largely in agreement on the Murray and Nahmias results. Provided that Shaun and I can add an acknowledgment during the proofing stage, we'll definitely mention this paper of yours!

Eddy Nahmias

I find David and Shaun's paper interesting and their alternative interpretation (to Murray and Nahmias*) of people's intuitions is fascinating and plausible. However, I don't think it's the best explanation for the existing data or for most people's psychology for various reasons. Here, I will just put one objection starkly.

Their model suggests that most people who (1) read the Nichols and Knobe scenario (universe A), (2) interpret it to mean determinism, (3) take determinism to rule out the existence (or efficacy?) of human decisions, (4) disagree with FW/MR questions about whether agents have free will or responsibility (or are blameworthy), and (5) agree with "bypassing" statements about whether decisions, beliefs, and desires have no effect on what the agent ends up doing. They seem to suggest that step (3) is the main cause of the responses in (4) and (5), and that it's the no FW judgment in (4) that drives the bypassing judgments in (5). (I'm not clear on whether they are saying people make bypassing judgments only because they have made the 'decisions don't exist' judgment.)

My main concern is with steps (2) and (3). I do not think most people are interpreting the scenario in these experiments to mean (only or mainly) ‘determinism’—rather, I think they are instead simply reading Universe A to rule out human decisions or their efficacy (moving from (1) to ‘no decisions’). If so, then it is not clear that they are taking *determinism* to rule out human decisions, and then considering FW/MR and bypassing in terms of that inference.

Why do I think people are reading Universe A directly to rule out human decisions? Because (a) people read A along with a description of another universe B, “in which almost everything that happens is completely caused by whatever happened before it. The one exception is human decision making.” And (b) people read that in Universe A “given the past, each decision *has to happen* the way that it does. By contrast, in Universe B, decisions are not completely caused by the past, and each human decision does *not* have to happen the way that it does.”

Regarding (a), it is quite easy to read the scenario with the contrast it suggests such that “human decision making” occurs in B *but not* in A. So, regardless of the ‘complete causation’ description, people may simply take it that they are supposed to think human decisions happen in B, whereas in A there are no decisions or at least no human decisions.

Regarding (b), it also seems relatively easy to interpret these closing sentences to suggest that decisions *have to happen* the way they do in universe A. I believe that it is extremely folksy to think that when it comes to choices and decisions, it can’t be that they *have to happen* the way that they do. So, I agree that if people read a scenario this way they are likely to interpret it to mean decisions do not happen in that universe (I think the folk are not committed to choice and ability to do otherwise themselves being incompatible with determinism, but more on that later). And from there, I agree we’d see steps (4) and (5), the no FW and bypassing results.

I’ve pushed this line before, so I don’t want to rehash it again. But the Rose and Nichols’ mediation/modeling results here, as far as I can tell, are based on comparing responses from (A) participants who read A vs. B and then are asked about A, compared to (B) participants who read A vs. B and then are asked about B. Given that methodology, I’m not surprised if there are big differences in the two group’s interpretation of whether decisions exist (or are efficacious) and about whether, in turn, free will exists and bypassing occurs.

Furthermore, for those participants who really do read the case to involve determinism and then take that to mean decisions don't exist, it seems we can still interpret that as a mistake and hence another way to push the error theory for incompatibilist intuitions. It’s highly contentious to think determinism means that decisions are impossible.

There's more to say, but for now, I hope others will join the discussion.
* The paper by Dylan Murray and I is published as early view at PPR here:
(email me if you can't get access.)

Josh May

Hi Eddy,

I'm not sure your explanation of the Rose and Nichols data can fit with mine, however (in the paper linked in my comment above). I wasn't trying to directly test whether (in)compatibilism is more intuitive, but the results speak to this issue.

First, I provided participants with only one case each, so there was no Universe A/B contrast (deterministic vs. not). Yet, in the first experiment of the paper, participants in the relevant condition were less likely to attribute free will, even when the agent was otherwise described as having psychological control over her action (Condition 3). More precisely, the determinism variable alone (corresponding roughly to what I call "liberty") had a statistically significant impact on mean agreement with an attribution of acting freely. (I used a Likert scale to measure degree of agreement with the attribution, rather than the usual Yes/No measure common in this literature.)

Second, while this experiment involved the "had to happen" terminology, the final one replaced this with "will happen." Yet the results were pretty much the same. Indeterminism alone had a statistically significant impact on responses, independently of control (what I call "ensurance").

However, changing the wording did make for a smaller effect size of this indeterminism variable (well, "liberty"). So I think there is something to your hypothesis that the modal language can play a role. But I doubt that it can explain away all the relevant incompatibilist intuitions here, given that the effect remained even when your worries aren't applicable. And, in that case, I then tend to think that the Nichols & Rose data do pose a problem for you and Dylan.

Josh Shepherd

Hi David/Shaun,

Cool paper. I just wanted to chime in about one thing: in your 'Study 2' you use the phraseology 'has no effect.' In a paper I did on bypassing (one that touches a lot of the same issues of bypassing incompatibilism etc.: if you ever read it I'd be interested to get your take, paper's here I suggest one might read 'had no effect' in two ways. I say this:

"Three of the bypassing questions state that an agent’s desires, beliefs and decisions ‘had no effect’ on what the agent did. Nahmias and Murray seem to want these questions to be read literally. In other words, when participants agree that an agent’s desires ‘had no effect’ on what she decided to do, subjects are taken to judge that the desires play no real causal role. But it is possible that participants read ‘have no effect’ in a different way. When speaking of a game between American football teams Louisiana State and Arkansas, for example, it makes sense to say that Louisiana State’s defense ‘has no effect’ on Arkansas’ offense, even though Louisiana State players are making tackles. It also makes sense to say that a philosopher’s argument ‘had no effect’ on an audience, even though the argument caused much thought, many aggressive questions, and so on. Here what we are saying is something like ‘the argument did not convince anyone,’ or ‘the defense is powerless to stop the offense.’ If participants read ‘had no effect’ in this way, then their pro-bypassing responses could be consistent with the view that agents’ desires, beliefs, and decisions play causal roles. Instead of asserting that desires, beliefs and decisions play no causal roles, they could be asserting that desires, beliefs and decisions are powerless to change anything."

Do you think people might be switching readings of 'has no effect' from the Practical Reasoning to the Theoretical Reasoning statements?

Also, I'm interested in what you're finding about decisions, but this is a long enough post already. Maybe we can talk about it sometime, or later on this thread.

Jason Shepard

Hi David & Shaun,

One of the major take home messages of your paper seems to be that that the folk notion of free choice has, at its core, indeterministic assumptions (ms p. 21-22). But what exactly are those indeterministic assumptions that lay at the core of the folk notion of free choice? A plausible candidate might be the ability to do otherwise. Perhaps for the folk, free choice requires genuinely open possibilities. But the existing evidence suggests that the folk don't think the ability to do otherwise is required for choice (nor for free will). See: Shepard, J. & Reuter, S. (2012). Neuroscience, Choice, and the Free Will Debate. AJOB - Neuroscience, 3, 7-11.

If it's not the ability to do otherwise that lays at the core of the folk notion of choice, could you say more about exactly what indeterministic assumptions do lay at the core of the folk notion of choice.

David Rose

Thanks everyone for posting these very interesting comments.


Your hypothesis that people are engaging in some sort of "contrastive reasoning" when given the case is interesting. Of course, if it is true, it would need to be tested. Given Josh May's studies, it seems that there is some reason to doubt the hypothesis.

Josh S:

That's a really cool idea. I'd be pretty surprised though if this is what was going on in our study 2. Plus, it seems to me that the model we get in study 2, with nonexistence-->bypassing suggests against this.

Jason S:

Thanks for drawing your paper to our attention. It looks very interesting. I wasn’t able to access it though. Would you be able to post an accessible version? I’d be interested in taking a look at it.

If what you report on the folk is correct, then it looks like the ability to do otherwise might not lay at the core of the folk notion of choice. This is very speculative and crude but people might think it comes "from the inside" in ways similar to what source incompatibilists say.

Dylan Murray

Cool results! I need to read your paper, Josh, but one quick note: in n. 14 of the PPR paper, we mention that dropping the "No Control" question from the Bypassing composite score slightly lowers the Cronbach's alpha among the questions used to compute it. So I doubt our results would change much qualitatively if we did so.

I also want to second Eddy's worry that an immediate move from the description of determinism to the conclusion that decisions don't exist sounds *a lot* like confusing determinism with bypassing. Granted, it's not bypassing of the sort where the relevant states exist but get cut out of the causal chain, but instead a form where the relevant states just don't exist in the first place, and so automatically get "cut out" of the causal chain. Maybe that doesn't count as "bypassing" in quite the sense we had in mind, but (to my mind) it's a mistake nonetheless.

As Jason suggests, it seems like the real issue here concerns how "choice" and "decides" may already have compatibilist and incompatibilist readings. Along with his, the results of our "choice" and "ability to decide otherwise" questions suggest that it's the former that people actually employ and care about. But the issue could be tested directly - e.g., by asking whether people can make "genuine decisions/choices" in deterministic universes.

One last point: The only experiment that directly intervenes on Bypassing or MR/FW in any of our studies - and so the experiment that would seem to have pride of place in adjudicating causal arrows - is our Study 2, in which the vignettes simply stipulate that Bypassing doesn't occur. Compared to Study 1, where we made no such stipulation, MR/FW judgments were higher. David and Shaun make some interesting remarks about that on p. 18, but considering all four scenarios here (not just the N&K abstract case), it strikes me as *much* more plausible to think that the difference in Bypassing (built directly into the scenarios) is having a causal effect on MR/FW judgments, rather than vice-versa. The alternative seems to require that participants somehow just ignored the information about Bypassing when they first read the scenario and until after they'd made their MR/FW judgments. But especially since (most of) the questions are randomized, that seems unmotivated. So even though they don't rule out a causal effect in the other direction, these results still seem to provide direct evidence that Bypassing judgments do have a causal effect on MR/FW judgments.

David Rose

Hi Dylan,

Thanks so much for the comment. The stuff about your intervention study is really interesting. But, I’m guessing that you actually ended up changing other variables and not just bypassing.

So, in intervening on a variable X to see if it has an effect on some other variable y, the intervention must be perfectly selective in directly changing x, while leaving every other variable undisturbed. That is, it can't affect other variables in the system. Interventions that are not perfectly selective, but end up altering another variable(s), provide little information about the causal relation between x and y. Interventions that are not perfectly selective but end up also altering other candidate variables, are "fat hand" manipulations. Here's an example from Scheines (1997), "suppose we are attempting to experimentally test hypotheses concerning the causal relations between athletic performance and confidence. Suppose we intervene to inhibit athletic performance by administering a drug that blocks nutrient uptake in muscle cells, but that this drug also imitates the chemical structure of neurotransmitters that inhibit feelings of insecurity and anxiety, thus serving to directly increase anxiety and lower confidence. This intervention provides little help in trying to reason about the sort of causal relation that exists between athletic performance and confidence, because it directly alters both variables." Given the evidence from the models that Shaun and I produced, it is natural to think that your intervention was not perfectly selective, that it was a "fat hand" manipulation and that you altered other variables (e.g., choice, mr/fw) in the process of intervening.

Josh May


I'm glad you mentioned your interesting paper, which I hadn't come across yet. One potential limitation is that you only used categorical Yes/No questions. The effects of some of these factors (like the unconditional ability to choose otherwise) might be more easily detected on a scale, and perhaps with a larger sample size.

This is especially true if my hypothesis is right, which says roughly that application of the concept of free will is affected by both the unconditional ability and the compatibilist-friendly notion of control. So removing just one factor isn't enough for people to wholly reject the claim that the agent acted freely (hence no difference in your Yes/No data).

I got support for this by using a scale measuring degree of agreement with an attribution of acting freely. I found a distinct effect of something like the unconditional ability to do otherwise (which I tie to the having of options or "liberty"). So, even given your data, it may still be that the unconditional ability to do otherwise plays some (albeit partial) role in ordinary thinking about free will (and via some conception of choice or options).

Jason Shepard

Hi Josh,

Thanks for your kind comments.

Whether you use a scale or yes/no questions depends a lot on your research hypothesis (and the dialect in which your research finds itself). Sometimes the point can be made more powerfully with categorical responses (especially if what you are actually interested in are categorical responses!). I am very distrustful of converting scale responses into categorical responses. This is a common move, as people are actually interested in the categorical responses even though they use a scale to measure.

But just in case you think categorical responses have inherent and fatal shortcomings irrespective of your research question/hypothesis/etc. (hey, some people (reasonably) hold this view), I will say that I have obtained very similar results when I use scales. (I have since used the standard agree/disagree scales for follow-up projects that I am working on.)

On a more theoretical level, I should note that our paper is only trying to point out that the unconditional ability to do otherwise is not *necessary* for ordinary attributions of choice and free will. Whether considerations of the unconditional ability to do otherwise have *any* influence on people's attributions of free will is an open question ... and I think it is very plausible that it has *some* influence on people's attributions of free will, but notice that the modifiers *any* and *some* make for a relatively weak claim. What I am most interested in is what is essential to our ordinary understanding of free will and choice. Or to stay away from talk of 'essential', what is the core of our ordinary understanding of free will and choice.


I am not savvy enough to attach a paper to my comments, so I am emailing you the paper.

The comments to this entry are closed.

3QD Prize 2012: Wesley Buckwalter