Blog Coordinator

« Manuel "You Da Man" Vargas wins APA Book Award!! | Main | CfA: Gothenburg Responsibility Project »



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

Hi Joshua Knobe,

Do you have a suggestion about what explanation people could have for regarding those things as precluded by determinism? It seems to me that a compatibilist could argue that those things *can* happen even if determinism is true (for example, in the case of willpower, there might be a deterministic "competition" between two systems in a single brain that just happen to include two stages). The compatibilist could then reject that belief in indeterminism is *fundamental* and even propose that it could be corrected through a better understanding of determinism. Although I'm inclined to think that determinism indeed precludes some aspects of our practical lives, I'm curious about what you would say in reply to such a compatibilist proposal.

Hi Joshua

My impression is that these results also might undermine at least some of the arguments intended to show that most people have an indeterministic conception of free will and/or moral responsibility, for the following reason:

Even if a person judges that in a scenario involving a deterministic universe, people do not act of their own free will, it may be that her judgment is prompted by some of her assessments about what a deterministic universe would look like, but she would not make the same assessment if she were thinking of a deterministic universe that contains, say, people who suddenly change their minds, who think twice against doing something immoral, or people who do many of the other things your study did not ask about but many (or even most) people may well think would not happen in a deterministic universe.

This is so even if the result also supports the conclusion that many or even most people believe that indeterminism is a fundamental aspect of human action: They may believe it's fundamental, but that does not entail that their concept of free will/moral responsibility is such that under the assumption of determinism + the assumption that human mental states are like the ones we have in the real world (and so, their fundamental belief is mistaken), there would be no FW/MR.


What you say here is exactly right. Just as you suggest, one could argue that people are simply mistaken in these judgments. This actually strikes me as a highly plausible position, and I did not mean to be rejecting it when I used the word "fundamental."

All I meant was that indeterminism is not just some little extra thing added on top of everything else people believe about agency. Rather, people see indeterminism as fundamental to agency in the sense that they think certain aspects of our agency could not exist without it.


Nice point! Maybe one way to bring out the structure of this argument is to say that people's intuitions about moral responsibility and determinism are a product of two factors:

(a) their understanding of what a deterministic universe would be like

(b) their criteria for moral responsibility

If we conclude that people are wrong about (a), then it may be that their views about (b) combined with a correct view about (a) together yield a compatiblist conclusion.

Hi Joshua--

A concern for me is that these questions test first and foremost an ability to parse particular questions against compatibility with strict causality but in two different ways: maybe deterministic robots (say) can love, crave ice cream, construct arguments, but maybe not "suddenly" change their minds or "resist" an urge if so challenged. Note that the first questions are general, the second ones specific. So one parameter is generality versus specificity of context of evaluating compatibility with causality. This is an impression of your post; I have not read your paper, so I might be way off track.

Hi Alan,

I feel like you are getting at something really important here, but it is a little bit difficult to follow the whole idea just from this very brief comment. Do you think you might be able to see just a tiny bit more?

Thanks Joshua--

I read your abstract and squinted really hard at the fuzzy preview version but that's all I have of your article so far. So again, my remarks are spitballing at best--more motivated by your OP than anything.

Your rankings above suggested my impressions in the earlier post.


6.06 — Fall in love
5.81 — Have cravings for ice cream that cause them to buy this food
5.13 — Construct complex legal arguments
4.00 — Perform actions that go against the values that their parents taught them
3.69 — Think twice about doing something immoral
2.93 — Suddenly change their mind about what they were planning to do
2.86 — Have the willpower to resist an urge to eat a steak in front of them

The first four are types of situations as described. So it occurred to me that people would evaluate those situations with respect to an understanding of determinism in type-generalized ways. I put it above as having them think of deterministic robots who do these sorts of things, and in a general way. One factor I suspect is that people are influenced by popular culture here--David as presented in Spielberg's AI is an automaton but apparently can learn to love but also is portrayed as having desired goals to pursue to accomplish that goal (finding the blue fairy to make him a boy); Commander Data of ST--Next Gen is a machine but defends himself legally in claiming to be more than just a machine in a famous episode; any number of films depict robots and androids that defy their parent-programmers, as in Blade Runner. Scenarios like that might cause people to be easily swayed by the plentiful fictional plausibility of programmed machines that can engage such familiar types of behavior.

But the latter three scenarios seem less susceptible to type-related analogies with so-called robotic behavior, even though they too can be worked out compatibly with deterministic-style programming with much more detailed explanation. Thinking twice about a decision might involve a program sub-loop for kinds of decisions that invoke a time-delay to receive additional data to redirect conditional procedures, or suddenly changing a machine-mind by involving a pseudorandom sequence in some instances (and even using them pseudorandomly in metaprogramming), or using a combo of these previous two to reproduce occasional resistance to programmed urges. But who besides AI people would even try to work out such much more specific kinds of behavior compatibly with deterministic-style programming? Given that they do not have lots of ready-made Sci-Fi examples to refer to for explaining such specific and much more mundane behaviors, then people do not have the resources to imagine how that compatibility would obtain, so their credences about compatibility with determinism drop.

That's my guess anyway. More general deterministic-robot-style behaviors have been described in popular culture that accord with the general behaviors of humans. More specific and chaotic human behaviors are less commonly presented as machine-consistent (though Ex Machina does a pretty good job for portraying the whole mess of human conduct as machine-compatible, good and bad). After all fiction usually paints in broad strokes about good and evil behaviors, whether machine or human--and those messages stick. But the particular fact (e.g.) that Rutger Hauer's android finally decides not to kill Harrison Ford's Blade Runner character--an example of changed mind--is just one episodic part of a much larger narrative where the plausible compatibility of that one decision with determinism of android composition is too easily lost.

Don't know how good any of this is. Just some (random?) thoughts. Thanks for listening.

Great work, Joshua!

In my dissertation (tentatively titled "Free Will and Meaning in Life) I suggest that our metaphysical view of the world has direct bearing on most of the meaningful evaluations we have about most people in our lives. Put differently, I think folks assume a metaphysical world view that's more in line with libertarians rather than a deterministic metaphysical picture and this work seems to be in line with that. If folks think things would be drastically different, that means they think it's a certain way now - this suggests a libertarian notion of free will. It shouldn't be a surprise then that the practices we engage in and the evaluations we draw about others on a daily basis are predicated on agents having certain abilities.

Anyway, great stuff!

Alan, I think you bring up a great point.

Re: 5.81 - sure, folks think the desire "causes" them to get ice cream. But believing that the desire is necessary but not sufficient is consistent with saying the desire is *a cause* but not *the cause*. Maybe most folks are pluralists when it comes to causation. They'd be right ;)

Re: 4.00 - This would seem to suggest free will for sure.

ETc. ETc.

I guess what I'm suggesting, and I think Alan was gesturing at (though maybe I'm way off), is that there are available interpretations of the data that would suggest that your thesis

"In other words, people do think that many of the things we do in our ordinary lives would also be possible in a deterministic universe (constructing arguments, falling in love), but they also think that some of the things we do actually would not be possible if the universe were deterministic (resisting urges, changing our minds)"

is false.

Sorry I'm jet lagged and this was a bit all over but I wanted to chime in as I think this study is important. It's important because a lot of folks are writing about the connection between free will and moral behavior. Pereboom, Caruso, Waller, Sommers, STrawson, and many others are making claims about what the world would be like in the wake of FW skepticism. Studies that focus on our moral psychology as it pertains to causation are helpful for those of us thinking hard about this very difficult subject.

Thanks Joshua,

I'm tempted to pay my way through the paywall, but I'll try to visit the library instead. I hope that more research in this vein will throw up a pattern - something that pinpoints some key feature(s) that people think to be incompatible with determinism, and which then explains the particular judgments. I have my own hypothesis. Short form: people mistake what David Velleman calls epistemic freedom for indeterminism. (I think calling it "epistemic" understates it, but never mind.) Changing your mind and resisting urges are cases that highlight our "epistemic" freedom.

Just finished reading the paper, nice, systematic work! What I find interesting is that most behaviors actually got above 4 rating of Compatibility, with median maybe 5? (Sometimes these ratings do not seem to be stable, like the item "Leave their long-term spouse unexpectedly to pursue a new love interest" scores 3.69 in Exp2 but 5.75 in Exp3.) So it seems people are mostly either indifferent about the compatibility with determinism or are even "behavior-compatibilists" to a degree.

This would be a bit in line with the cited literature that people might not understand determinism and its implications (or even care?). I wonder how the results would look like in a different group of subjects who would receive a more detailed explanation of determinism and how our minds work (e.g. something like the passage suggested by Dennett in "Some observations on the psychology of thinking about free will" in Baer et al 2008).

And perhaps it is only if people are forced to answer such an abstract question that they "translate" it (unconsciously?) to something like a question of Effort or Spontaneity. Which is of course very interesting.

I am really sorry that this paper was trapped behind a paywall. It is now available for free at:

Alan and Justin,

These are all very helpful suggestions. Actually, we pursue in the paper the idea that the only behaviors people can perform in a deterministic universe are the ones that are seen in some sense as 'robotic' or 'mechanical.' Specifically, we asked participants the following question:

Sometimes people’s actions seem to proceed mechanically like clockwork. Other times, people seize control and take things in a different direction.

For each of the following behaviors, we want to know whether you agree that people can only do it by seizing control in this way.

Just as I think you might have suspected, the behaviors that were picked out as being more mechanical according to this criterion were precisely the ones that people saw as most compatible with determinism.


Very nice points. You are completely right to say that even though we are getting a lot of variance from one item to the next, the means in general tend to be relatively high. I also like your suggestions about how to go after this issue more fully.

In any case, the differences observed between the results in the different studies presumably just reflect the fact that we used a between-subject design with a relatively small number of participants rating each individual behavior. This gives us enough power to determine whether or not certain other items are predicting the variance across behaviors, but there is still a pretty high standard error for the mean we report on each behavior considered separately.

This result is really interesting; I suppose that the natural extension for me (and perhaps also someone like Justin) is to consider how this might influence judgments about certain moral concepts that are generally considered impacted by determinism. Consider whether someone can "be punished" in a deterministic universe. I'm not sure how this interacts with the active/passive distinction, but it looks passive to me. However, it seems like it's necessarily conditioned on an active behavior (i.e. 'behaving badly/morally impermissibly/etc.'). If the distinction develops for non-moral concepts in a pretty straightforward way, it may be somewhat less straightforward for moral concepts where conventional understandings of determinism generally regard moral cases as easy.

Hi Joshua,

Thanks for linking to an open copy of the article. The results of the study were very interesting. It seems that this might provide some support for Helen Steward's ideas about the deep folk concept of agency (and perhaps for some version of agency incompatibilism).

I was wondering, though, whether Ondrej Havlicek's comment about people misunderstanding determinism might suggest a plausible alternative explanation. While the studies you report surely show something very interesting about how people tend to categorize actions, I wonder whether they really tell us what people think is possible in deterministic universes. Might they not simply reveal a particular way of misunderstanding determinism?

Murray and Nahmias (2012) (as noted in your article) claim that people often misunderstand determinism, and once people have a better grasp of the position, they report different intuitions. Do you think a misunderstanding of determinism might play some role in how people answer the questions? If my students told me that they thought that if determinism is true then it is impossible to suddenly change their mind about what they were planning to do, I would conclude that they hadn't really understood the nature of determinism. I know your studies involved a comprehension question, but I wonder whether it was sufficiently difficult to test a person's understanding of determinism. This is, of course, nothing more than wondering on my part, but I was hoping you could speak to this worry.

What can we learn by polling the prisoners in Plato's cave on their beliefs about the shadows, and how can that enlighten us about what casts the shadows, etc.? By analogy, what can we learn about free will by asking Joe the plumber? I think the whole empirical turn in free will philosophy is misguided. No substantive conceptual issue will be resolved by it. It cannot determine whether compatibilism is or is not wretched subterfuge. Arguments alone can do that. Polls of the people strike me as philosophy on vacation...


That's a really great question. If I had to guess, I would say that there is a difference between people's judgments about whether it is possible that people *are* punished and whether it is possible that people *should be* punished. The former seems possible even if everyone acts 'passively,' while the latter seems to require that people have a capacity to be 'active' in the relevant sense. So my guess would be that people think the former is possible in a deterministic universe while the latter is not.


I would not describe the distinction as being between arguments and experiments, but rather as being between a priori arguments and experimental arguments. In a typical a priori argument, one might start out with certain propositions that we already believe and then show that these propositions entail or otherwise support a particular conclusion (say, incompatibilism). By contrast, in an experimental argument, one uses empirical methods to explore the underlying processes that lead us to believe certain propositions. This then gives us evidence that helps us to determine whether or not those beliefs are warranted.

The comment above from Angra Mainyu is a great example. Perhaps you find yourself inclined to believe in incompatibilism. But now suppose you learn that this belief arises because you are intuitively drawn to the idea that certain behaviors are impossible in a deterministic universe. You might then conclude that your intuition was driven by a mistaken metaphysics. (Of course, the argument for that conclusion would not be based on further experiments but on a very traditional sort of philosophical reflection.) As a result, the experimental findings combined with a larger philosophical argument could give you reason to infer that your previous beliefs were not warranted.


Thanks for the new link. Your investigation does some great work finding underlying factors of these judgments. The mediation analysis for the active/passive factor is impressive.

I had three small questions about the experimental stimuli and the way people were understanding the deterministic universe. Specifically, I was wondering, did you use any controls to compare judgments in universe A to judgments about these things in a universe more like our own? Did you ask any questions beyond the television set one to gauge people's understandings of how deterministic universes are supposed to work? Do you think that people would answer differently if the stimuli used another example beyond leg movement, perhaps something more cognitively demanding, to illustrate how determinism works to people?

Hi Wesley,

Thanks for your interest in our paper! To answer your questions…

1. Yes, one of the reviewers asked us to look at what behaviors subjects thought were possible in our own universe, and we report a summary of these results in a footnote in the paper. The basic upshot is that, though there is a marginally significant relationship between what people think is possible in our universe and what they think is possible in a deterministic universe, the relationship is small and doesn’t undermine any of our key findings. (If you are curious about the specific means, though, I’d be happy to email them to you.)

2. No, we didn’t ask anything other than the TV question, but to some extent, that was intentional. Because we hypothesized that people might believe certain mental activities couldn’t even occur in a deterministic universe, we didn’t want to assume in the comprehension question that people must be able to reason in certain ways in such a universe. (This was also the motivation for using the “leg” example in the prompt rather than something more psychologically complex, as has been used in past studies—see below.)

3. I do think it’s possible that people would answer differently if we gave a different example, but for the reasons similar to those described above, we worried that introducing a more cognitively demanding example could push around people’s responses in ways that don’t reflect their true beliefs about determinism. For example, if we just stipulated in our prompt that somebody could resist an urge to drink alcohol in such a universe, participants might be more inclined to subsequently say that someone could resist the urge to eat a steak in this universe even if the participants didn’t truly believe this and were simply answering in this way to seem logically consistent with what the experimenters told them.

In any case, these are interesting questions for future work. In this kind of work, it’s difficult to balance the tasks of (a) testing comprehension to ensure participants aren’t just completely confused about what determinism is and (b) getting people’s unadulterated intuitions about what’s possible in a deterministic universe. Perhaps in our studies, we focused too much on (b) and could’ve done a better job of making sure that participants really understood determinism. I’d love to hear people’s thoughts on better ways to address this.


Thanks for your thoughtful responses about the experiments. I was wondering, then, how you might interpret the results in light of the fact that there were only marginal differences (sorry I missed the footnote!) between Universe A and our universe on these items. Specifically, haven't prior findings about Universe A vs an indeterministic Universe B indicated that people judge the two are very different and that B is more like our own? As a result of that research I would have predicted larger differences on your items in Universe A and a world like ours. One hypothesis to explain the marginal differences is that people were basically answering questions how they would in our world too, which given that prior work, is to say, they weren't accepting determinism about the case.
I think these are very interesting findings and I do not mean to say anything to impugn them. But I fully agree it is very difficult to balance people's unadulterated comprehension of determinism in universe scenarios in a way that doesn't unduly impact the very thing you are trying to measure. Or even, what to ask people to show they are applying that understanding. Does that fact ever give you pause about the value of unadulterated judgments in universe scenarios for understanding ordinary reactions to determinism?

Hi Wesley,

Thanks so much for all your comments on this! Your own recent work on this topic has been fantastically interesting, and I'm delighted that you are engaging with these findings.

Just to clarify the result from our study, it's not that there was only a marginal difference between the deterministic universe and our own universe. (On the contrary, there was a very large difference between ratings for those two universes.) The point Adam was making was just that ratings of possibility for our own universe were only marginally predictive of ratings in the deterministic universe. In other words, ratings of possibility in our own universe predict only a small amount of the variance in ratings for the deterministic universe, and all of the rest of the variance is predicted by something specific to the deterministic universe.

Does that help to address your concern?

The comments to this entry are closed.

Books about Agency

3QD Prize 2014: Marcus Arvan