Blog Coordinator

« Causal Determinism and Going with the Flow | Main | Thanks! »



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


That's very plausible, and it explains all the data that I'm aware of on concrete/abstract effects. Which is less than the data you're aware of. But I want to put my competing hypothesis out there, which can also explain all the data I'm aware of.

On this lighter-folk-commitment hypothesis, the folks' automatic implicit understanding of agency has it that the agent can resist mental tendencies. (Note that in everyday observation, only mental tendencies, not physical traits, seem relevant to explaining decisions.) In this, "the agent" is unspecified other than, by implication, it has to go beyond simply mental tendencies. It could be dualistic or physical, indeterministic or deterministic. Which is not to say that the folks will find all these possibilities equally believable. Many of them will have explicit theories saying that agents are dualistic and contra-causal. Even those who don't, may find a physicalistic view over-complex and far-fetched. But as far as the automatic implicit understanding of agency goes, "the agent" is mostly a blank check, to be filled in later.

This is just a variant of Eddy's "theory-lite" approach, as far as I can tell. But if not, I'll take the blame for it. It's the same old thing I've been getting at for three threads now - but hopefully, new aspects are showing up.


That's surely one of the contenders, and it might be correct, or partially correct (i.e., maybe it explains the effect with respect to some people, but not all).
I'd like to also add a competing hypothesis: in another experiment, you found that most people believe that a number of ordinary actions wouldn't even be possible in a deterministic universe.
In the abstract scenario, people are told that the universe is deterministic, but they aren't told which actions are possible. So, it seems very probable to me that when assessing whether there is free will, moral responsibility, etc., in a deterministic universe, they're basing their assessment partly on what they reckon is possible and isn't possible in such a universe, and because some relevant actions aren't in their assessment possible, they conclude there is no free will, no moral responsibility (or no full moral responsibility), etc.
On the other hand, in the concrete scenario, there is a specific description of what's going on in the deterministic universe, which (at least in the case of many, perhaps most people) may well override what they would otherwise have considered possible in a deterministic universe.

I'm going to take Paul's courageous comment as opportunity to chime in with a likely amicus brief. There are some very different possible motivations for embracing incompatibilism and compatibilism: theology versus science, the abstract versus the concrete, and so on. But one strong motivation for compatibilism that must have some role to play in the empirical work is pragmatism, particularly moral pragmatism. What I mean by this is that in general compatibilists have been motivated by rejection of what they see as failures of incompatibilists to close the deal on how control can be at once non-causal yet susceptible to reasons--a very abstract concern. So they look for things like reasons-responsiveness, lack of external coercion, identification with second-order desires and so on, all of which aim *to work* with assigning responsibility in a way that aligns with something like familiar moral judgments. To me that suggests an overall pragmatic motivation. But, if such pragmatism rules in compatibilism, then maybe, just maybe, it is that same motivation that dominates the folk as well when presented with both abstract and concrete examples. When abstract examples of determinism are presented but then concrete examples of choice are included, maybe people just generally default to what they generally know as what has worked for them and society. If it's agent-resistance of an apparent contra-causal kind that's needed, then fine. So maybe it's a moral pragmatism related to "normal" moral standards that's in play here, and not some presumption about what the self is.

Hi Josh,

you wrote: “In essence, the difference would not be a matter of how people decide whether the kind of agency they are imagining is compatible with free will or moral responsibility. It would be a matter of what kind of agency they are actually imagining in the first place.”

I am wondering if the last sentence is meant to encompass a conception of agency according to which the possible actions from which people may choose are themselves put in place by social force relations, that is, relations of power put in place the limits of possible conduct and, furthermore, operate effectively because subjects assume that their acts have been “freely chosen:” that they are autonomous and free.

The conception of agency that I have described in the previous paragraph summarizes some of Foucault’s ideas about the relation between the subject and power. It is (still) widely believed that, taken together, Foucault’s claims did not leave room for agency. This (mis)understanding of his work assumes that freewill (and autonomy, and so forth) and force relations are mutually exclusive, that is, there is an unadulterated freewill that stands apart from power. Yet, Foucault wrote against such a conception of the relation between power and agency. For Foucault, subjective agency always operates within the constraints of social force relations.

If you don’t endorse Foucault’s ideas about the relations between power and agency, is there some other conception of the relation between power and agency that you do endorse?



When people think abstractly about agency, the training they’ve received from science comes heavily into play. When people think concretely about agency, their own experience comes heavily into play. Science argues that all of the control associated with an agent is predeterministic in nature, and science is (currently) unwilling to acknowledge that our thoughts have an emergent property of control, which adds into the mix and affects the path forward. When people think concretely about agency (instead of abstractly using the ideas of science), people directly *experience* the control exerted by their thoughts. Since the scientific viewpoint (i.e., predeterminism) contradicts their experience, that’s where the tension comes from and causes people to believe predeterminism is likely false.

It seems to me that mankind is waiting for neuroscience to discover that our thoughts exert new emergent control.

Hi Josh, I want to push back a bit here by pointing out that you do not really seem to be offering an explanation for why people are drawn towards compatibilism, as the start of the post suggested. You are saying (most, all?) people have an implicit commitment to contracausal free will (call it AC for agent causation) and that when they respond that an agent can have FW or MR in a deterministic universe, they are simply ignoring or misunderstanding the determinism because they are being told the agent makes decisions and that triggers their AC intuitions. So, this is an error theory for why some people offer responses that may be misinterpreted as suggesting they have compatibilist intuitions. Not that there's anything wrong with that! I do the same thing in reverse to explain away incompatibilist-looking responses. We do need to figure out better ways to test between these two opposing views and to test the extent to which folk psychology has a single 'content' here (compat/theory-lite or libertarian/AC), such that we have to explain away contrary responses, versus there being different psychologies that explain conflicting responses (Feltz's view?) and if so, which is more prevalent, where, and why.

I won't repeat my alternative error theory, but just highlight that it starts with the exact same move of saying that concrete scenarios that highlight agents making decisions will trigger people's folk psychological processes more so than abstract scenarios. I just think those processes do not invoke any metaphysical content as substantive as AC (as Paul explains nicely). It's also not clear to me what such content would do for us if we were trying to understand others' actions or our own.

Hi Eddy,

Point taken. You are right to say that the hypothesis I propose here might not be rightly described as an explanation of people's compatibilist intuitions. Instead, it might be described as saying that, on some deeper level, the intuitions that appeared to be compatibilist aren't actually compatibilist at all. I am not sure which of those descriptions would be the best.

In any case, I hate to be imposing extra work on you, but do you think you might be able to write your alternative explanation for the abstract/concrete effect here? Then everyone reading this would be able to compare the two explanations, and we could see which sorts of evidence might speak in favor of each option.

p.s. Responses to the other comments coming soon! My apologies for the delay.

Hi Shelley,

I agree that there is a lot of potential in thinking about how the project I am pursuing here relates to the work of Foucault. My project is about understanding people's ordinary conception of the self, so to bring this project into dialogue with Foucault, it would be essential to focus on Foucault's work on contemporary conceptions of the self.

I am thinking, for example, of his discussion of the 'Californian cult of the self,' in which 'one is supposed to discover one's true self, to separate it from what might obscure or alienate it, to decipher its truth thanks to psychological or psychoanalytic science.'

A question then arises as to whether Foucault would characterize contemporary conceptions of the self in a different way from the one I have proposed here and, if so, whether future studies could benefit from taking advantage of his work. Of course, you are the scholar, so I will defer to you on this question.


This approach sounds intriguing, but could you say just a little bit more about how exactly it explains the difference between the abstract and the concrete?

(This is not supposed to be some kind of implicit objection, just an honest question.)


Nice! This hypothesis actually makes a really interesting new prediction. It predicts that the abstract/concrete effect should go away when we consider agents who do something wrong by going with the flow. (For example, there should be no effect when we consider the action of an addict who just continues using.)

In my view, this is definitely an intriguing avenue to explore.

Hi Alan,

This is a very nice suggestion, which makes a helpful counterpoint to my own. Basically, the difference seems to be a matter of whether one explains the abstract/concrete effect by positing an extra level of complexity in people's conception of the self or by positing extra complexity in people's moral judgments.

The hypothesis I develop suggests that something surprising is happening in people's way of making sense of the self and then, with that idea in hand, explains the effect without having to posit anything terribly surprising about people's capacity for moral judgment. It seems like your proposal does the opposite. You make a promising and very plausible suggestion about people's moral judgments. Then if this suggestion turns out to be right, the effect can be explained without suggesting that anything surprising is happening in people's conception of the self.

I guess the reason I was drawn to the former hypothesis is just that existing work finds surprising effects of the abstract/concrete difference even on questions that don't seem to have anything to do with morality or free will in themselves. For example, there are striking effects of the abstract/concrete difference in the studies by Nahmias and Murray:


Regarding the prediction, I'm not sure it's that the effect would go away in those cases, or rather, just that the effect would be significantly reduced but perhaps not disappear in those cases.
That's because even in "going with the flow" example, there are people - even if a minority - who think they're impossible in a deterministic universe.
So, when given the concrete scenario, those people will be given info that would override what otherwise would have been their assessment in a deterministic scenario (namely, they will assume such behaviors possible), and some of them may well deem the behaviors in question immoral, unless perhaps they consider that moral guilt is prevented by some other feature they believe the deterministic scenario has; but it's unclear to me that all of them will think so, especially once they have incorporated the info that the deterministic universe is considerably different from what they would have anticipated.
So, in brief, I'm inclined to think the alternative hypothesis I suggested predicts at least a significant reduction of the abstract/concrete effect, but I'm not convinced it predicts it would go away.


Good question. I suspect that in Abstract cases, people who have an explicit Agent Causation (AC) theory simply deploy it. But they can't easily do that in Concrete cases because the description of the case militates against the explicit theory. The subject is forced to consider an agent who seems quite like us, and who lives in a deterministic universe, and the subject finds the notion surprisingly easy to accept. Your survey is somewhat turning them into compatibilists-in-principle, although they may still believe that in *our* universe agents act contra-causally. Compare a person who goes into the movie Blade Runner thinking that artificial beings could never be conscious, but comes out thinking differently.

Addendum to my last comment: here are a few ways the Concrete cases could militate against subjects' explicit theories. (1) People in this deterministic universe still look ahead to consequences of actions, instead of being pushed from behind. (2) People ponder their perceived desires and wonder whether to go with the flow or against it (Angra's point, roughly?). (3) People's decision making is not bypassed but plays a vital causal role (Eddy's theory).

To the extent the Concrete cases state, show, or suggest any of these features, that could explain the concrete/abstract differences.


Yes, that would be an approximation to the hypothesis I suggested. I would just add the following two points:

1. The hypothesis I suggested is not limited to "going against the flow" cases. Even though the category "going against the flow" was the only one statistically significant in the study, it seems that a good number of people judged behaviors that involved "going with the flow" impossible in a deterministic universe; I don't know whether that would have some impact in a study, so I'm leaving the door open for an impact in cases in which (say) people are not "pushed from behind", or people's decision making is not bypassed, etc., depending on the description of the deterministic universe. The "going against the flow" cases might be just some of those in which there is some conflict between the concrete description and what some (most) people believe or would have believed if presented with the abstract case.

2. The hypothesis is silent on whether the theories are explicit. In other words, I'm suggesting there might be conflict between the beliefs about the deterministic universe that some (maybe most) people have or would have if presented with the abstract deterministic scenario (whether those beliefs are based on explicit theories or not) and what's possible in the deterministic universe according to the description of the concrete case.

It seems to me that if the hypothesis I sketched turned out to be right, the one you posited about explicit theories might be right as well, since it seems to be a more specific one.


Thanks again for a series of great posts. I apologize for being late to this particular discussion, but my hard drive crashed and it's taken me longer than expected to get my computer fixed. That said, I wanted to offer another explanation of the concrete/abstract findings that is based on the recent work by David Rose, Wesley Buckwalter, and Shaun Nichols. While their target in the paper were intuitions about neuropredictions, the explanation of the findings they develop might help explain the concrete/abstract asymmetry as well. In short, they suggest that people's indeterministic metaphysical framework "intrudes" upon their intuitions about deterministic cases. On this view, there is a kind of imaginative resistance to deterministic cases such that people tend not to internalize or properly understand that the cases are deterministic, properly speaking. So, while people may give compatibilist-looking responses in some cases, it's just because they are not really viewing the cases as deterministic. As such, their responses don't provide evidence for folk compatibilism but rather evidence for "intrusion effects." Might something similar happen in the abstract vs. concrete cases? In abstract cases, because people don't feel a "moral tug," they are less inclined to experience intrusion effects and have imaginative resistance to the determinism of the cases. Nothing hangs on their judgments after all. But when the cases are concrete, suddenly something hangs in the balance--namely, the desire to blame the agents in the vignettes for their bad behavior. Here we might expect intrusion effects and imaginative resistance to play a more prominent role. Folk indeterminism is triggered by the specific agents and particular details and the desire to judge people in real situations. So, people give what appear to be compatibilist answers in concrete scenarios not because they judge that people can be free and responsible in deterministic situations but rather because they judge that the people are free and responsible, they simply don't internalize the deterministic nature of the scenario. On this hypothesis, it is the desire to blame and punish which leads to intrusion effects and imaginative resistance, which in turn misleadingly make it appears as if they are giving compatibilist responses. In low or no stakes abstract cases, on the other hand, it is easier for people to abandon their indeterministic roots. Just another possibility to consider. I have collected a ream of data on this issue with David, Wesley, and Shaun--but we can't find the time to analyze them and write up the results. Hopefully, we will get to it soon. In the meantime, people can find an earlier draft of their interesting paper here:

Hi Thomas,

This is all super interesting to hear. I didn't even know that you were working together on this.

Anyway, the hypothesis you describe is obviously deeply similar to the one I propose here. However, it seems like your hypothesis differs in that it posits a role for punishment and blame. I would love to hear more about the results from your studies!

Hi Thomas,

Regarding the "intrusion" hypothesis, that's very interesting. I do have a couple of worries, though:

1. The interpretation of "could" in this context might not be a matter of contention. Is it not the case that she could have chosen otherwise?

2. Even going by that interpretation of "could", I think there might be some difficulties with the original scenarios in the first place (and generally with brain-prediction scenarios).
For example, one question is: "After the pattern of brain activity occurred, could Jill have voted for a different candidate for Governor instead of Green?"
I suppose here answering "yes" would be interpreted as the result of intrusion. But if I had been asked that question, I wouldn't be inclined to answer "no". If I have to give a yes/no answer, I would be struggling, and definitely not because of libertarian/indeterministic intrusion.
For example, let's say that a couple of seconds after the brain pattern based on which the scientists predict she'll vote for Green happens and before she's aware of a choice to vote for Green, Jill gets a message on her phone from her sister.
She takes a look and the message says "Did you see the Green scandal? It's on TV right now". As it turns out, Green has just been arrested for some heinous sex offense, and there is a leaked video that's going viral and seems to clearly catch him in the act. Some media outlets are reporting he's already confessed.
Could Jill (even given that the pattern happened already) refrain from voting for Green? If she couldn't no matter what info she's given, I'd be inclined to say something is very wrong with her brain, and wouldn't know whether she's decided freely (more info about the choosing process would be required, but probably not).
Or maybe she gets the message after she's aware of a choice to vote for Green. There are many variants.
Perhaps, most people are including an "all other things equal" condition when interpreting the scenarios, which excludes any further relevant information, threats, etc. But I'm not sure they would do that in these sort of context, when "could" questions are asked. I don't think they're likely explicitly coming up with scenarios like the one I just described, but I think more information is needed to tell what is driving their judgments.
On the other hand, in the "manipulation" case, people may be getting an impression that Jill has to vote for Green regardless of further info (or they might be getting that impression more often than in the prediction/no manipulation case).


Thanks for your comments. I wasn't trying to defend the intrusion effects hypothesis as it was tested by Rose et al. with the neuroprediction cases. So, I will leave it to them to defend their approach. For present purposes, I was simply suggesting that the hypothesis could be put to work when it comes to the concrete/abstract asymmetries.

In the follow up work I did with David, Wesley, and Shaun, we didn't use new cases (nor did we focus on neuroprediction cases specifically). Rather, we went back and tested old cases that have elicited compatibilist-looking responses in the past. Except this time, we asked tons of questions per vignette rather than just one or two to make sure that the participants were understanding the determinism built into the scenarios rather than reading their indeterministic beliefs and intuitions into the otherwise deterministic scenarios. This post reminded me that I need to get in touch with my collaborators. We collected data from over 1,500 people. But analyzing the results was pretty far down each of our respective to do lists! Now seems like a good time to pick up where we left off!

Hi Thomas,

This is really extraordinarily exciting. To see whether this thing explains the abstract/concrete effect, you could just randomly assign participants to either abstract or concrete and then check to see whether intrusion mediates the impact of condition on moral responsibility judgment. You haven't tried that, have you?


Thanks for the suggestion. No, we didn't try that. At the time, we weren't trying to get at the abstract/concrete issue. Our goal was to see whether intrusion effects and imaginative resistance might explain away some of the earlier work by me, Eddy, and others that seemed to show compatibilist intuitions. So, we used some old "classical cases" (e.g., the supercomputer case, the rollback case, and others), manipulated some additional variables (e.g., agents vs. robots), and asked a bunch of more detailed questions to get at whether indeterminism was creeping into their understanding of the vignettes. I never liked how many people failed the comprehension checks when we ran the original studies ten years ago (yikes, it's been that long). This recent work was an attempt to suss out why that may have been the case. Emailing David, Wesley, and Shaun is on my list of things to do tomorrow. We collected the data last summer, I was up for tenure last fall, and then I had neck surgery in December. So, I simply dropped the ball. Now I will try to encourage them to help me pick it back up. I will keep you posted. In the meantime, let me know if you'd like to work on something similar specifically focused on the abstract/concrete stuff--which, as you've said, would dovetail nicely with what you're already doing. Just a thought!

Which intuition (incompatibilist/compatibilist) would win out if you ask people both types of scenario? So if you ask an abstract scenario first would that then influence their later answer to a concrete scenario? Would people come back and change their answer after considering both types of scenario, or just be inconsistent about it?

Hi Thomas,

That sounds like an incredibly exciting project. If you are interested in pursuing the abstract/concrete angle, I would definitely be more than happy to help out. (I probably wouldn't be able to be a full-blown coauthor, but I'd be happy to do some work on this all the same.)

p.s. That sounds like a pretty intense time you're going through. Hope everything is okay!

Hi Greg,

Good question. We ran a very small pilot study along those lines, which we describe on p. 680 of the paper at:

In our study, experimental participants who got both the abstract and concrete versions were pretty much evenly split between compatibilism and incompatibilism.

It is possible that other researchers have gone after this question in a more serious way since the time we wrote that paper, and if so, I hope they can write in with further information.

The comments to this entry are closed.

Books about Agency

3QD Prize 2014: Marcus Arvan