Trying to Test Smilansky's Free Will Illusionism

As some of you already know, Eddy Nahmias, Jason Turner, Steve Morris, and I ran a number of studies a while ago with the goal of probing folk intuitions concerning free will, determinism, and moral responsibility. The details of these studies--and our analysis of the results--can be found here. Our most surprising finding was that participants were less likely to have incompatibilist intuitions in response to cases involving determinism than philosophers (especially incompatibilists) have traditionally assumed. Leaving aside for now the results fof some interesting follow-up studies that have been run by Joshua Knobe and Shaun Nichols (the details of which can be found here)--I wanted to say a few words about a project that I am working on with Adam Feltz (a grad student here at FSU and a contributor to this blog). When we looked more carefully at Saul Smilansky's free will illusionism, we realized that his particular (and peculiar) position within the free will debate is markedly more dependant on empirical claims about folk intuitions than other positions. Indeed, we argue that if the relevant data does not come out his way, then there is no way of motivating illusionism at all. The details of our argument need not concern us now.

For present purposes, I only want to discuss the results of some studies we ran in an effort to test two empirical assumptions about folk intuitions that play a critical role in Smilansky's attempt to motivate his view. The first assumption is that the majority of people have illusory beliefs concerning the existence of libertarian free will. The second assumption is that if people were disillusioned about this--i.e., if they came to realize that libertarian free will is both incoherent and non-existent--this would have negative effects at both the societal and personal levels. On the one hand, people would be markedly less likely to behave morally. On the other hand, people would no longer find meaning and value in their lives. To test these assumptions, we ran some new studies--studies that mentioned determinism explicitly rather than simply implicitly building it in to the vignettes as we had done in our earlier studies. Participants were 105 undergraduates--none of whom had studied the free will debate before. They were broken into three groups--two groups of 40 and a group of 25 (FW1, FW, and FW3--respectively).

Participants in FW1 received the following questions:

1. Do you think that human beings have free will?

2. Do you think that our actions can be free if all of them are entirely determined by our genes, our neuro-physiology, and our upbringing?

3. Do you think that free will is necessary if we are to be ultimately morally responsible for our actions?

4.Now, for the sake of argument, assume that in the future scientists discover that all of our beliefs, desires, choices, decisions, and actions are entirely determined by our genes, our neuro-physiology, and our upbringing. Now assume that this scientific discovery leads you to conclude that humans are neither free nor ultimately morally responsible for their actions. Would you be more inclined to behave immorally in light of this knowledge?

Responses were as follows:
1. 90% said yes
2. 30% said yes
3. 83% said yes
4. 35% said yes

Participants in FW2 received the same set of questions, except that Q4 was phrased in the third person rather than first person, as follows:

4. Now, for the sake of argument, assume that in the future scientists discover that all of our beliefs, desires, choices, decisions, and actions are entirely determined by our genes, our neuro-physiology, and our upbringing. Now assume that this scientific discovery leads people to conclude that humans are neither free nor ultimately morally responsible for their actions. Do you think they would be more inclined to behave immorally in light of this knowledge?

Responses were as follows:

1. 90% said yes
2. 48% said yes
3. 80% said yes
4. 70% said yes

Before discussing the results of FW3, I wanted to say a few brief words about the results of FW1 and FW2.

First, obviously participants were less likely to give compatibilist friendly responses to the definitional question 2 than they were to the cases we ran earlier. Presumably, this is partly explained by the fact that Adam and I used the word determinism whereas in the earlier Nahmias studies we built the determinism in without actually using the word "determinism." Nevertheless, I think that philosophers will be surprised to know that when we add up the responses to question 2 for FW 1, FW2, and FW3, we find that 44 out of 105 said that we would be free even if all of our actions were entirely determined by our genes, neurophysiology, and upbring--nearly 42%. And while this means that a bare majority of people may be incompatibilist, a robust minority are not--even when they are explicitly told that all of their actions are entirely determined, a number of people are still willing to say that we are free. Although I am nevertheless curious to hear why people think we got different results in these studies that we got in the earlier studies. Perhaps someone--like me!--should run another definitional study that does not specifically mention determinism. If the results are similar to those of FW1-FW3, this would suggest that it is not the word "determinism" that is having the effect, rather it is that people respond differently to definitional questions than they do to analyses of particular cases--an issue I have already mentioned in some recent posts. Unfortunately, until more studies are run, we won't know for sure. In the meantime, I look forward to hearing your thoughts about the new studies.

Second, when we compare the responses to question 4 in FW1 and FW2, we find that people are twice as likely to think that other people would be more likely to behave immorally if they found out that determinism is true than they are to think that they themselves would be more likely to behave immorally in light of this knowledge. Since one of Smilansky's claims is that people would be more likely to behave immorally, it is unclear which of these results are salient to testing his asumption. Minimally, I think the first person-third person assymmetry here is interesting, and I also think that it is something that should receive more attention (I have run some trolley problem studies aimed at testing for a similar assymetry in people's responses to trolley problem cases--the results of which I will post as soon as I can make some sense of it). Minimally, if we add the first and third person results together, we find that little more than half of the participants thought that finding out about the non-existence of free will would have a negative impact on people's propensity to behave morally.

In FW3, we simply wanted to see whether people believed that if they discovered that we do not have free will, they would find their lives to be less meaningful or valuable. The first three questions were the same as before--but the fourth question was as follows:

Now, for the sake of argument, assume that in the future scientists discover that all of our beliefs, desires, choices, decisions, and actions are entirely determined by our genes, our neurophysiology, and our upbringing. Now assume that this scientific discovery leads you to conclude that humans are neither free nor ultimately morally responsible for their actions. Would you therefore conclude that your life—as well as human life more generally—is less meaningful?

The results were as follows (keep in mind, this group only had 25 students):
1. 92% said yes
2. 52% said yes
3. 80% said yes
4. 28% said yes

The most important result for Adam's and my purposes is the responses to question 4--where only 28% of the participants thought that disillusionment about free will would lead them to conclude that life was less meaningful. This was, by our lights at least, a surprising result. Indeed, it was the one question which we thought would come out the way Smilansky would have predicted.

OK, since Adam and I are still actively working on the paper--I don't want to spend much more time trying to flesh out the details here. Let it suffice to say that we think Smilansky's free will illusionism is on fairly shaky empirical grounds. And, if our analysis of his view is correct, this poses a serious problem for his view more so than any other view in the free will debate. Once we have a working rough draft of the paper, I will post a copy here. In the meantime, I am curious to see what others think about this new data--especially in light of the data that has already been collected by myself and others.

Comments

[sennoma]

How many study participants would typically be considered adequate for such questions? (I ask because you are careful to note the small size of one group.)

Further to that idea, surely such things as "folk intuitions" (I am going on dictionary definitions, so if that's a term of art I may well be misunderstanding it) vary from group to group. Or do they? Would my ideas about free will be different if I'd grown up in a commune, or a socialist dictatorship, or...? It might be interesting to cross-reference demographic information on your participants with their answers.

[Neil]

Hi Thomas,

I have multiple problems with the set-up. First, I'm not sure that we can probe beliefs about strange counterfactuals (or, indeed, tacit beliefs) just by asking. Compare: would you shock someone all the way to XXXX in the Milgram experiment (this question, in a 3rd person form, was actually asked by Milgram of a control group: they predicted that some low % would go all the way - about 5%, I think). Now contrast that with people's actual performance in the experiment. Or, more directly about beliefs: would you conclude that someone close to you had been replaced by a replica if you ceased to feel the normal emotional response upon seeing them (which is the Ellis & Young theory of the origin of Capgras).

Second, I think there are additional difficulties with asking 'would you behave immorally'? First, the contrast between 1st and 3rd person responses: compare the response to these questions: are people heavily influenced in negative ways by violence on TV (a fairly large proportion respond 'yes'); are you heavily influenced (etc) (very few people answer 'yes'). Second, there is a problem with defining 'immoral': some subjects may think that if determinism is true, it's not possible to act immorally, since morality requires libertarian free choice. So the claim that I wouldn't act immorally if determinism is true is not equivalent to 'I wouldn't act in ways that would be considered immoral or are immoral if determinism is false'.

As for the last question, about meaningfulness in life, I suspect people simply don't know what they're being asked about.

[tnadelhoffer]

Neil,
I can always count on you to be long on criticism concerning the set-up and short on helpful suggestions for ways to improve them. Luckily, the problems you point out are usually helpful (and besides, having a resident malcontent keeps us honest!). In this particular case, we have already thought about the problems you point out (and your analysis of the asymmetry between first and third person responses to question 4 in FW1 and FW 2 is surely along the right lines--which is precisely why we ran the studies in both first and third person). We have also given some thought to your worries about whether asking people if they would do x under certain conditions is a good way of determining whether they would actually do x under the specified conditions. We tried to come up with some ways of running some behavioral studies--but all of them would require the type of funding and resources that we lack as philosophers--which is why I hope to work more with psychologists in the future. Jonathan Schooler has run some pretty cool studies recently on this very issue--but since the studies are still in the very preliminary stages, he is understandably tight lipped about the details (he talked about the studies at the recent SPP meeting in Raleigh, NC). Hopefully, we will hear something more about the results of his follow-up studies sometime soon.

In one respect, I feel like a number of your criticisms of the sorts of studies I run apply to a number of the studies run by social psychologists more generally--as such, I feel less compelled to defend myself in this respect since if these kinds of studies are good enough for well-known psychologists (e.g., Bertram Malle or Jonathan Haidt), they should be good enough for us philosophers who sometimes put on our psychologist's cap. Your worry about counterfactual scenarios--however--is certainly an important one that we should all think about more carefully.

So, even though I don't feel it necessary to defend my general methodological approach to the kinds of studies I run, I am nevertheless going to dig in my heels when it comes to the particular studies I discussed in the post. After all, Smilansky's view *requires* that the folk (a) have the illusory beliefs he says they have, and (b) that if they were shown the errors of these beliefs, bad things would follow. If either of these two assumptions are false (and absent any empirical support, they are little more than assumptions), then there is no way to motivate his otherwise curious view.

It is obviously fine (and not very surprising) that you are dissatisfied with our preliminary attempts to test his view--and we are very clear that they are little more than that--but since Smilansky's view is intimately tied to some explicitly empirical assumptions about folk intuitions, we are simply pointing out that he must sink or swim with the data about these intuitions. And we have merely suggested that the gathering data make it likely that he will sink.

If it turns out that more sophisticated studies vindicate his view down the road, that would be fine by us. Our goal is not to falsify his view--indeed, I have to admit that I have always been quite fond of it--rather, we were simply trying to figure out how we might going about testing it. Since his is a paradigm case of a philosophical position that is entirely beholden to folk intuitions, we simply wanted to highlight that fact as yet another way of motivating experimental philosophy. Surely you can agree that so long as he is going to rely on assumptions about folk intuitions to motivate his view, we are justified in asking for evidence that supports these assumptions. And in the event that there is no evidence--above and beyond his own intuitions of course--we are presumably justified in trying to figure out (a) what kind of evidence would do trick, and (b) how one might go about coming up with the relevant evidence. You seem to doubt whether we have come up with an adequate way of settling (b)--which is fair enough. Perhaps you have something helpful to add about (a)--which is something we talk about in the paper, but I wouldn't want to spoil the surprise.

[Neil]

Thomas,

Don't take my criticisms to be expressive of the view that I can do better! Experimental work is *hard*; that's the main reason why I've not yet done any (I am actually planning some soon, involving administration of beta-blockers to subjects; that should get me some exp. phil. cred). I concede the aptness of saying that I am long on criticism and short on constructive suggestions; I'm not Robinson Crusoe there. But I do think these are real problems. A nice illustration: how good do you think people are reporting whether they have engaged in overt action? Turns out, not very good: a study which asked people never to smoke inside got a number of participants who, knowing they had been filmed, confidently asserted that they had not smoked inside - until they were confronted with the video.

About the only helpful suggestion I have is this: you can do more to 'operationalise' words like 'meaning' and 'morality'. Morality is easier, if my claim that people have implicit libertarian assumptions about it is correct: describe it in overt behavioral terms; eg. would you (or others) be more likely to kill, steal, rape, lie....?

Meaningfulness is harder to operationalise. You might ask about relationships and love. You might explicitly distinguish it from happiness and welfare (eg. 'would you conclude that though lives can be happy, they cannot be meaningful?')

But I do think that these are just hard questions, and we shouldn't expect to find experimental methods that settle them definitively.I don't know the answers to *any* of your questions: that is, I can't confidently answer yes or no to any of them, speaking just for myself.

[tnadelhoffer]

After re-reading my comments a few minutes ago (and before reading your response to them--which I was unaware you had already posted), I ammended them slightly to make it clear that I genuinely do always welcome your criticisms. Your suggestion about operationalizing "morality" is something Adam already suggested--but at that point, I had already run the studies with the old wording. And as it is hard to come up with the participants for these studies--I ran over 12 studies with over 500 students last week!--I won't be able to try it again for a few weeks. We thought in the future we would simply use particular types of immoral behavior such as theft or lying--figuring that if we used something like murder, people who be less likely to either (a) imagine themselves engaging in that type of behavior, or (b) to be honest about the liklihood of their engaging in such behavior.

Figuring how to operationalize meaningfulness would be much more difficult--indeed, much less thought went into that study as it was not something we had originally planned on testing. I nevertheless thought we might get some interesting results, so we ran it anyway. All of these studies are just first run throughs we needed in order to be able to write something off to send to some conferences. We will run some bigger (and hopefully improved) studies in the spring that will hopefully be part of the final paper. For now, we just wanted to be able to (a) point out how uniquely Smilansky's view is dependent on the empirical data about folk intuitions, (b)look at some of the data that have been gathered to date, (c) present the results of some new studies aimed at probing the relevant intuitions, and (d) suggest what kind of future studies would be needed to settle the matter for sure. Hopefully, we will end up accomplishing all four tasks--independently of whether some of the methods we used in the current studies end up being unsatisfactory.

In our defense, we were trying to run studies that were as charitable to Smilansky as possible. For instance, just asking people point blank whether they think that we can be free even if all of our actions are entirely determined seemed a much more direct approach than the one we adopted in my earlier studies with Eddy and crew. And as I said in the original post, I expected the results of the study to go Smilansky's way. In any event, once we have run the new studies I will post the results for you to criticize :)

I also look forward to reading about your upcoming foray into the experimental realm. In many respects, the kind of hard science that you will dabbling into is more interesting to me than the sort of surveys I have been doing. I nevertheless believe that the surveying that we have been doing as of late will help lay the groundwork for more interesting empirical work in the future involving brain imaging, fMRI, etc. Indeed, Eddy is fortunate enough to have some new colleagues in the Brain and Behavior program at GSU who are interested in precisely this kind of work--which is in many respects a logical extension of the stuff we have already been doing. But for now, I have to be content with simply polling people--something I know you aren't all that fond of, but that is the topic of another post altogether.

[Eddy Nahmias]

Just a couple points about these interesting studies. First, I am very surprised by the relatively high number of "compatibilist friendly" responses to question 2, first because we had earlier found that many people define "determinism" to *mean* "the opposite of (or lack of) free will" and second because your description has the reductionistic ring that I always worry about in descriptions of determinism and that I think (mistakenly) drives many incompatibilist intuitions. Your question mentions genes, neurophysiology, and upbringing which seems to suggest that these factors alone are the immediate or essential causes of all actions, leaving out any causal role for, say, one's desires, intentions, deliberations, conscious choices, etc. (all of which can have a causal role in our actions even if determinism is true). Anyway, it'd be nice to know more about why people are answering the way they do on that question.

Second, the final question is tricky to interpret for several reasons, but the one I'm most worried about is that almost half the people had just answered that determinism would *not* take away free will (or, presumably, moral responsibility), but then they are told to believe that they think it *would* and to consider how they think that belief would affect their own or others' behavior. That's some tough counterfactual imagining to do on top of the difficulty of interpreting whether what people say they would do has any bearing on what they would actually do.

Having raised these issues, I should say that these surveys, like all you do, still provide us with interesting data to consider, try to interpret, and try to fit in with our philosophical theorizing.

[Dan]

Thomas,

1. I think that you should analyze the results in more detail, in part to respond to Eddy's concern about the tough counterfactual imagining. You can break down people into groups based on their responses to the first 3 questions and then see how the responses for the last question vary between those groups.

2. Neil's account of the first person/third person difference on the last question is probably a big part of the story - as in the Milgram study, people like to think that they will do the right thing regardless of the situation. However, there's another, less interesting explanation that the data don't rule out, based on the difference between individuals and groups. Suppose that exactly one third of people would be more inclined to behave immorally, and that everyone has bias-free knowledge about what will happen. Then 1/3 of people will answer "yes" to the first person question, but everyone can answer "yes" to the third person question, since (one third of) future people will be more inclined to act immorally.

3. The methodology here is not very strong (though it's a start, and operationalizing immorality and meaningfulness would be an improvement). People are being asked to predict their behavior and experiences (or others' behavior and experiences) in a hypothetical world with a fairly abstract, conceptual difference from our world. There are plenty of reasons to think that people are not good at this sort of prediction. People are bad enough at predicting how they'll feel a few years after being denied tenure (Gilbert et al. 2002), or what divinity students will do if they pass a man in need when they're in a hurry. Haidt's research on moral intuitionism uses hypothetical scenarios in a methodologically stronger way, requiring people's reactions to concrete scenarios (which is what much of moral judgment is anyways - giving your opinions on situations that you hear about). You might consider a correlational study - are ordinary people who believe in determinism & incompatibilism more immoral, and do they consider life less meaningful? Although finding a large enough sample of such people might turn out to be a much more ambitious project, and it the correlational study would have its own flaws.

4. I can't wait to see what you've done with the trolley problem. I have wondered about how people might respond to different variations on it - like asking what their judgments would be as a witness, rather than what their decisions would be as a participant; or what they'd do if the fat man might fall to his death and stop the trolley if they didn't do anything. I expect that, if you give several versions of the problem to the same people, their responses will be highly dependent on which version you give them first. If you want more minds trying to make sense of your results, I'd enjoy taking a look at them. We can talk by email if you have any interest.

Citations:
Gilbert, D., E. Pinel, T. Wilson, S. Blumberg, & T. Wheatley (2002). Durability Bias in Affective Forecasting. In Gilovich, T., D. Griffin, & D. Kahneman (Eds.), Heuristics and biases: The psychology of intuitive judgment. New York: Cambridge University Press, 292-312.

[tnadelhoffer]

Dan,
Thanks for the comments. As I am presently exhausted (and soon to be heading off to bed), I will have to be somewhat brief--promising to revisit your comments tomorrow or Tuesday (you make a number of interesting points and suggestions that I might currently be too sleep deprived to fully appreciate). For now, I just wanted to say that our main goal was to see how people responded to the definitional question (Q2)--since that is the key assumption for Smilansky's view. The questions that asked people to make predictions were admittedly problematic--for all of the reasons that everyone has so helpfully pointed out. In the future--assuming we even decide to pursue this particular line of reasoning any further--we would certainly be more context specific concerning the immoral behavior in question. Moreover, we would try and uncomplicate some of the counter-factual reasoning required on the part of the particupants--as Eddy is certainly correct to worry that it is far too hard as stated (hell, it might even be incoherentto some participants for the reasons that both Eddy and Neil mention). Looking at some of the participants' comments support these concerns.

Leaving aside for the moment your worries about the method we used for these questions, I wanted to say something about your suggestion concerning running some correlational studies since this is something I just wrote about tonight while I was working on the paper! In at least one important respect, Smilansky seems to be his own counter-example since he along with most free will skeptics (myself included!) presumably live meaningful lives, behave morally, have loving and caring relationships, etc. Hell, most compatibilists seem to get by fine without libertarian free will as well.

Of course, it doesn't follow from the fact that trained academic philosphers can give up their belief in libertarian free will and ultimate moral responsibility without any great perceived loss, that non-philosophers will be able to do so as readily or easily. But before I would be willing to adopt such an elitist and paternalistic stance towards the folk, I would need to see some relevant data that supports such a stance. If you can think of an interesting way of generating that kind of data, I would certainly like to hear more about it.

OK, I gotta run for now. I will try and return to your ideas and suggestions--as well as some of the things posted earlier by other commenters--sometime soon. I will also post the results from the trolley problem studies I ran recently--which I have not given much thought since I have been trying to make it through the Smilansky paper. But since we should have a working rough draft in a few days, I should have some time to give the other studies I ran last week some more attention. Thanks again for your very helpful comments.

[Dan]

Thomas,

I'm not very familiar with Smilansky, but it looks to me like the people who are most relevant to Smilansky's argument are those who answer yes, no, yes to the first 3 questions. These are the people who believe in free will but are at risk of being disillusioned - if they came to believe that genes, neuro-physiology, & upbringing determine actions, then they might become less moral and their lives less meaningful. Question 4 should make sense to these participants, since it only involves that one change from what they already believe. So there would be an advantage to looking at the answers to question 4 from just these subjects. That pattern of responses seems to be common enough for you to already have a sample size that is large enough to be meaningful.

The correlational study I was thinking of has a pretty simple design. First you need some way of measuring how immoral people are and/or how meaningful their lives are. You could probably find measures that already exist that you can use, possibly in modified or abbreviated form. Then all you need to do is recruit a bunch of subjects, ask them their beliefs about free will, and give them the morality and/or meaningfulness measure. If there is a correlation between free will beliefs and the other measure(s) in the direction that Smilansky predicts, then that's a good sign for him; if not, that's bad for him. Of course, you don't want to put too much weight on this evidence, since there are plenty of ways that such correlations could arise without disbelief in free will causing immorality & meaninglessness. As a second step in increasing methodological rigor, you might look into the research on the morality & meaning measures that you use to see what factors are known to correlate with them, and then ask subjects about those factors and control for that when you look at the relation between free will beliefs and morality/meaning.

I would caution against generalizing from philosophers to the folk. Philosophers are a highly non-representative sample. Haidt found that they think differently than most people - reasoning is not just the tail on the intuitive dog for philosophers. Further, philosophers know a lot more about the free will debate - the various positions, their implications, and so on - so believing that there is no libertarian free will means something different to them than it does to ordinary people. Philosophers have other concepts that they can use once libertarian free will is taken away. Additionally, a lot of people think that morality and meaning depend on God/religion/spirituality, while most philosophers take a different approach, and I imagine that this will have implications for the relation between free will and morality & meaning.

[Matt James]

There seems to be a significant underlying issue in the comments by Neil and Thomas about the nature and scope of experimental philosophy. There seems to be a rush to legitimize experimental philosophy by using laboratory psychology methods, but once we start administering beta-blockers to people, it looks to me like we would just be doing psychology. (Not that there's anything WRONG with that.) Is the goal of X-Phi to find out people's intuitions and how they think and reason about philosophical issues, or to find out which cognitive processes are responsible for the production of those behaviors? The former to me sounds like X-Phi, and the latter sounds to me like psychology. As much of psychology has shown, our reasoning about a situation can be inconsistent with our behavior in such a situation, and it's not clear to me that X-Phi need be equally interested in both.

Obviously we shouldn't ignore the empirical data on cognitive processes, and should take them into account when making any descriptive claims, but it seems that X-Phi should be more interested in people's reasons and justifications for their judgments, and should draw from the experimental data only to show that these reasons may not be what people always act on. Of course, views like Smilansky's with strong descriptive claims are still vulnerable to empirical refutation, and theories like virtue ethics may be vulnerable to charges of descriptive vacuity, as John Doris points out, but what exactly is the work of X-Phi beyond invalidating descriptive claims and getting base rates for new ones? It may be that the surveys are the ideal method for X-Phi, and while we would need to work closely with psychologists, and (gasp) even do some "hard" empirical studies ourselves, we should keep in mind which are which. It may save philosophical debate and help the survival of the movement if we more properly define the scope of X-Phi, although by no means do I claim to have done it here.

[Dan]

There's one other issue that I thought of. Smilansky says, roughly, that if people don't have the illusion of free will then bad things will happen. But, as a first approximation, there are two ways that this could be true:

1) Things will be bad for people who have the illusion of free will and then become disillusioned.

2) Things will be bad for people who grow up without ever having the illusion of free will.

I don't know which claim is more plausible. Maybe the first problem isn't so severe because those people are already accustomed to morality and meaning, or maybe the second problem isn't so severe because those people would get used to living without free will and would not suffer the shocks disillusionment. The study reported here mostly speaks to the first claim. I don't know how you'd test the second claim - perhaps by studying the children of people who don't believe in free will? If immorality and an absence of meaning are going to be long-term problems, and not just transitional problems, then you really need to test something like the second claim.

Alternatively, you might take a position that combines aspects of 1) and 2), perhaps following Paul Bloom ( http://www.edge.org/3rd_culture/bloom04/bloom04_index.html ) in claiming that every person naturally develops a belief in libertarian free will (because of genes and early childhood experience). That way, every person would face at least some sort of disillusionment. Testing this claim would probably require research in child development - you'd want to keep up on what developmental psychologists are doing.

I don't know if Smilansky has taken a stance on what process is at work, but in order to do empirical tests you really have to be clear on which process you are testing. That could be one advantage of the experimental approach to philosophy - it forces philosophers to be more specific about mechanisms.

[Tamler Sommers]

I'm late to this, and new to this blog but I want to make one rather obvious point. Even if the results came out entirely in Smilansky's favor, that STILL wouldn't vindicate illusionism. It would only show that our first instinct is to be wary of denying free will and moral responsibility. (This goes for Schooler's studies too, I saw him give that paper...) But it would not rule out the common folk becoming convinced over time that life without free will and moral responsibility is not so bad. Think of a theist who takes a survey on how he would feel if there was no God. He'd probably say: I'd be immoral, everyone would be immoral, life would have no meaning etc... But one day, after reading some books by Carl Sagan and Richard Dawkins perhaps, he might come to realize that a Godless world is not so bad.

And that's a role for philosophers, especially those who like me who deny free will and moral responsibility: to explain to people who are initially pessimistic about this position why we can still lead happy, meaningful, love-filled lives...

Of course, this would be hard to test in a study, but I've often wondered why follow-up questions and interviews can't be a part of these surveys (in the way that Haidt uses follow-ups in his moral dumbfounding studies). And if you look at the Monterrosso paper posted on The Garden of Forking Paths, you'll see that some of the most illuminating data comes from the follow-up interviews...it's where the implicit dualism in the subjects' views on moral responsibility most vividly reveals itself.

But now I'm rambling, too much cold medicine. Very interesting study!

[Neil]

Matt,

It is no part of my (planned) experimental work to set an agenda for X-phil, or to invalidate others. I happen to be doing work on beta-blockers, from which the experimental work arises naturally. It's still broadly philosophical, inasmuch as I hope to discover something about the nature of reasoning, but that's not the primary motivation.

So far as I can see, survey-based X-phil is appropriate just insofar as philosophers appeal to intuitions - no more, but no less. And appeal to intuitions is so common that there's plenty of room for such an approach.

[tnadelhoffer]

I just got home from tutoring LSAT for three hours--so until I have had the chance to consume some (more) coffee and wake up a bit my comments may be less than fully coherent (indeed, they not even be coherent *after* the coffee--but that is neither here nor there!)

First, I wanted to thank everyone for participating in this thread--it makes the time and energy trying to generate interest in the movement worth it. Indeed, the more dialogue like this that we have, the more we are forced to think about what exactly the movement is attempting to accomplish in the first place (as well as what it is not trying to accomplish). It was especially nice to hear from Matt--who I haven't spoken with in a while--and from Tamler--someone whose views concerning free will and moral responsibility are similar to my own--but who is new to the blog. Perhaps he (i.e., Tamler) might be interested in posting something at the Garden about this thread here so that others might join in--I thought about sending them an email, but I never got around to it.

Second, I think Matt is right that only certain kinds of experiments/studies might qualify as experimental philosophy (I still don't like the abbreviation X-Phi for some reason--although it is certainly useful for brevity's sake). Although I would ask what we are supposed to call the work that psychologists do--especially when it is in many ways indistinguishable from what we do. Consider, for instance, Bertram Malle--one of the pioneers of examing folk concepts concerning intentions and intentional actions in both an empirically informed and a philosophically informed manner. Clearly--or at least clearly to me--he is doing psychology and not experimental philosophy. So, why is it that when I do it, I am doing X-Phi and not psychology--especially when a number of the kinds of vignettes, methodologies, etc. are the same? I suppose that some people will use my posing of this question as evidence that we aren't really doing philosphy after all when we put on our psychologist's hat. But I am curious to see how others will answer the question.

Third, I think Tamler is right that it is unclear precisely what Smilansky's Disutility of Disillusionment Assumption (DDA) amounts to. Is it a prediction about what would happen if people who believe in libertarian free will ceased to believe in it, or rather a descriptive claim about people who have never believed in free will? Since he limits himself mainly to "the West"--we assumed he had the former prediction in mind (coupled with the assumption that most people in "the West" believe in libertarian free will--an assumption that we are calling the Ubiquity of Libertarianism Assumption). I think Dan's suggestions concerning how to test DDA are certainly helpful. Hopefully, in the spring I will have the opportunity to run some studies that are along the lines that he suggested in earlier comments. Indeed, based on all of the criticisms that have been put forward in this thread of our first attempts to get at the relevant intuitions, we have decided not to include that part of the data in the paper at all--except as a footnoted example of how *not* to try and test DDA.

Fourth, I wanted to say something briefly about Tamler's suggestion concerning exit interviews--which I agree are a very good idea. Unfortunately, without any funding or assistance, it is hard to set aside the time to do them. Last week, for instance, I ran nearly a number of studies with over 800 participants. Adam helped me with the 105 free will surveys, but the others were done solo. I simply don't have the time to conduct any interviews (although we did try some exit interviews when Eddy, Jason, Steve, and I worked on the phenomenology of free will paper--but there were four of us and only a handful of participants). This is another instance where I think as philosophers become increasingly engaged in running studies, etc., we should start eliciting help from our psychologist friends concerning how to get funding, etc.

Finally, I wanted to remind everyone that one of the first X-phi philosophers to publish something about folk intuitions concerning free will and moral responsibility was Shaun Nichols--whose interesting paper concerning children and libertarianism entitled "Folk Psychology of Free Will"--which can be found on his homepage at: http://www.philosophy.utah.edu/faculty/nichols
Everyone should check it out.

OK, that's it for now. Thanks again for participating. Hopefully, I will be able to post a rough draft of our paper soon. It has been greatly improved in light of you helpful comments (I'll make sure to give credit where credit is due!)--which is precisely why I posted some of the data.

[Matt James]

Neil,
By no means was I demanding that you set an agenda for X-Phi (although if you could, that would be nice). My point was intended to be that while your criticisms of the survey method, as in the contrasts you made between people's surveys and actual performance in the Milgram experiments, may not be the kind of thing that would be the primary work of X-Phi, and thus not relevant criticisms of its experimental methods. This is the familiar point that a philosophical theory is not necessarily invalidated by people's failure to conform to it, unless strong descriptive claims are part of that theory, such as appear to be in Smilansky's.

For another example, take the familiar Nisbett and Wilson pantyhose studies, in which people consistently chose the pair on the right despite the fact that the products were identical, and made up explanations for their preferences. In these cases, the biased cognitive processes were responsible for the production of behavior. Should an X-Phi approach focus on the intuitions and reasons people give for their choices, even as spurious as they can be, or should X-Phi count the cognitive processes that choose objects on the right as people's true "intuitions"? If the cognitive processes are the right "intuitions", then it seems to me that we have stretched intuition well beyond its normal use, and intuition pumping just becomes cognitive psychology.

So it seems that in order to avoid the collapse of intuitions into cognitive psychology, we should place some restrictions on the notion of intuition, such as separating it from actual behavioral performance, and thus it is no criticism of survey methods that they don't capture people's performance. In your reply to my comments, I think we agree that X-Phi should focus on philosophical theories that appeal to intuitions. I just wanted to make the point that X-Phi shouldn't necessarily take people's actual performance as its primary concern. There are issues, as I think was your intent to point out, with what counts as a belief and whether we hold a belief if it conflicts with performance. Yet for the moment I am pushing the line that in such cases, X-Phi should be more concerned with verbal reports, reasoning, and responses to abstract questions, as issues of performance involve many more processes that muddy the influence of the intuitions that X-Phi is concerned with. I'm not sure that we disagree, I'm just trying to hack out what are relevant criticisms of X-Phi methods and what are not, and arguing for the claim that X-Phi shouldn't worry about testing strange counterfactuals by asking; in fact, that's exactly what we should be doing. Anything more begins to look like psychology.

[jonathan weinberg]

Matt, why shouldn't experimental philosophy come to be pretty much indistinguishable from cognitive psychology? Our overall theoretical aims will be different from psychologists' (since we're looking most fundamentally to answer philosophical questions, and they aren't), but I don't see why our methods shouldn't aspire to be practically identical to theirs. I don't see this as a threat of "collapse", but more like a promise of liberation!

[jonathan weinberg]

Does anyone know how to make italics show up in the commentary? My and html tags seem to just disappear, without actually changing the text at all.

[Matt James]

Well Jonathan, that's an excellent question, and I'd be the first to admit that there are days when I wish philosophy could be "liberated" in such a manner. I think my answer is that the difference in theoretical aims between X-Phi and psychology inevitably leads to different experimental methods. (Whether this is a general rule or not requires more philosophy of science than I know how to do or to post here.) The goal of X-Phi, at least at this point, seems to be to articulate what people's intuitions about philosophical issues really are, and perhaps later to elucidate the reasons (or lack thereof) behind them for further evaluation. The devices of cognitive psychology generally do not aim to capture our surface intutions, instead taking them for granted, but the causal mechanisms and processes responsible for production of behavior. In fact, most of the great cog sci work of the past few decades consists in finding that people's behavior violates their own intuitions. The answers that cog sci gives succeed in establishing that reasons and rationality are not the governing forces in our behavior, but to stop the story there would leave no room for reasons and rationality, especially if you think with Cosmides and Tooby, Gigerenzer, and others that all these heuristics and biases are a good idea. It would liberate philosophy by making it epiphenomenal.

Now it may yet turn out that philosophy actually is ephiphenomenal, and questions of evaluating reasons and rationality should be replaced with sharpening our cognitive processes or designing better mechanisms, but there is nowhere near the evidence needed to support that claim now, and as a philosopher by training I think we still have reason (.) to think that there is philosophical work to be done in the evaluation of our intuitions and reasons, while cog sci seems bent on refuting them. Perhaps I am just arguing from past cog sci work, and work that could be done using cog sci methods like reaction times could be useful for X-Phi. However, while studies of reaction times could tell us that people have difficulty with some moral decisions more than others, we already know that, and while fMRI studies of trolley problems may tell us whether there are specific areas of the brain dedicted to moral reasoning, which I am sure will cause some to revive some metaethical arguments about moral realism (in what I expect will be a rather unsound fashion), the project of philosophy seems largely unaffected. A more interesting question is whether X-Phi can be distinguished from social psychology, which often relies on surveys in conjunction with more cog sci methods, but I don't think X-Phi has a stable enough research program yet to be able to say one way or the other. Experimental philosophy, it seems to me, is being done to fill some descriptive gaps in the philosophical literature, and not to liberate philosphy by reducing it to the operation of mechanisms, but there are probably as many agendas as there are experimental philosophers. I'm writing this as much to convince myself as to defend a position to you, as I've been in one of my "save philosophy" swings for the past few months, but I think it helps to try and set these issues out there.

[jonathan weinberg]

This part isn't quite right, and I think it's an important mistake here:
"The devices of cognitive psychology generally do not aim to capture our surface intutions, instead taking them for granted, but the causal mechanisms and processes responsible for production of behavior"

In order to get a theory of the underlying mechanisms, you've first got to get a pretty good theory of what our judgments & behaviors are in the first place. The theory of the behavior provides the data for the theory of the mechanisms to explain. And our own rough commonsensical/intuitive/introspective judgments about how different people would behave in different circumstances turn out not to be terribly good. (Just to be clear: I am using "behavior" there to include not just bodily actions per se, but also things like what you are calling "surface intuitions".) So one part of X-Phi is just using standard empirical methods to get a better picture of those behaviors themselves. What X-Phi is reacting against, in part, is analytic philosophers' tendency to take a very thin & biased slice of those behaviors (like attending only to attributions of the term "free will" to weird hypothetical cases), and try to build an account of the philosophical phenomenon (like agency itself) directly off of that rather bad sample of data. But nothing in that kind of critique depends on saying anything like the underlying mechanism is what philosophers really should be studying, or anything like that.

Having said that, information about underlying mechanisms can certainly become relevant to at least some philosophical disputations. At a minimum, the better account we have of where intuitive judgments come from, the better able we will be to separate credible intuitive judgments from ones that we ought disregard. Debates about the a priori depend in part on what sorts of structures we might find existing in the mind in some appropriately perception-independent way, and whether we do or don't have such structures is a question for cognitive science. And if some metaethical or moral-psychological views have empirical commitments (e.g., about what does or doesn't motivate us under various conditions), then those commitments can and must be tested empirically. Thus we can hope that the relevant science will give us new material upon which to base our philosophical theories, and to help in our selection among different competing philosophical theories.

So I think that X-Phi can, in that sense, be liberating for philosophy without having "to liberate philosphy by reducing it to the operation of mechanisms" (which I agree sounds a bit like 'liberating' the fish from its pond and into the oven). There's nothing here about the reduction of philosophy to anything else. It is, rather, about spelling out & testing empirical commitments that are already there in the philosophy itself.

[Kip Werking]

I just finished Smilansky's remarkable book. It was a pleasure to read a book that the Catholic Church certainly would have banned in the 16th century. I hated half of the book. And I loved the other half.

Smilansky's an eloquent spokesperson for free will denial, as articulated by Galen Strawson. But I think his philosophy could be more rigorous. For exampe, although Smilansky defends Strawson's no-causa-sui -> no fw position, he refers to the promise of libertarian fw and "lacking libertarian fw." But libertarian fw (fw enhanced or enabled by indeterminism) is very different from being causa sui (and Strawson explicitly notes that indeterminism doesn't help one be causa sui). But this is a quibble.

Some points about this research:

1. The answers to question 4 in the third study are surprising and counter intuitive (which is not to say that I disagree with them, the value of incoherent things like being-causa-sui is controversial). Considering the small sample size, one would like to see this result confirmed.
2. Why leave aside Joshua Knobe and Shaun Nichols fascinating suggestion that emotion inclines one towards compatibilism?
3. I strongly agree with the commenter who mentioned the Milgram studies. Self-reports are notoriously problematic, and I suspect these problems would be complicated by asking people to make pessimistic conclusions. People are just hesitant to make pessimistic or negative conclusions in general. The first person/third person assymetry probably illustrates the problem with self-reports. That said, I doubt one could take people who believe in fw, convince them that fw doesn't exist, and then test whether they become more immoral (or could you, muhahaha!).

[tnadelhoffer]

Kip,
Thanks for the comments. For our purposes, we don't bother talking about the strength (or lack thereof) of Smilansky's argument against libertarian free will. On the one hand, we don't see how he adds anything to the stuff Strawson already did (which I, at least find convincing). On the other hand, we are more interested in the empirical assumptions about folk intuitions that he needs in order to get his view off the ground.

As far as the Nichols and Knobe studies are concerned, we will talk about their data--especially as it relates to studies I ran earlier with Eddy, Jason, and Steve--in the longer version of the paper. For now, we are only trying to get a conference-length version put together. But it should suffice for now to point out that their explanation of the data--as convincing as I happen to find it--cannot explain why so many participants in our new studies answered the definitional question in the way that they did. After all, they are merely asked the straightforward question:

Do you think that our actions can be free if all of them are entirely determined by our genes, our neuro-physiology, and our upbringing?

As far as we can tell, it is hard to explain the results of this question away under the banner of emotional impact, etc. Indeed, we used a definitional question partially in order to avoid emotional biases and the like. And as I said earlier, I genuinely expected far fewer participants to answer the question in the affirmative. Indeed, I don't see any other way of explaining the responses of those who thought we could be free determinism notwithstanding, except to assume that 42% of the participants either (a) already had compatibilist intuitions, or (b) ended up with compatibilist intuitions once they were asked to reason counter-factually about the existence of determinism.

Minimally, Adam and I think these preliminary results provide perhaps the strongest evidence to date that far fewer people have incompatibilist intutions than philosophers--especially Smilansky--have assumed.

[Kip Werking]

I would note that the specific question you highlight just refers to "free action." Many philosophers who deny the existence of free will can entertain the existence of free action (Hobbes comes to mind). The question makes no reference to the will, freedom of the will, or free will (or any of the capitalized or hyphenated or conjugated variants of these). It does not distinguish between religious/metaphysical/philosophical and political freedom, nor does it suggest any distinction between Ultimate/Incompatibilist/Origination freedom and "superficial"/Compatibilist/non-controversial freedom. Of course, the survey refers to "free will" in the other questions, and so the survey takers might not have exploited this ambiguity. But perhaps some did. This is a problem I've raised before, and one that might be corrected by giving survey takers a menu of options. I try to do this on my own survey:

http://people.wm.edu/~ktwerk/survey.htm

[Paul Torek]

I agree with Matt's point that "contrasts ... between people's surveys and actual performance in the Milgram experiments, may not be the kind of thing that would be the primary work of X-Phi", at least in this case. Consider the 2/3 who claim that they would not act immorally in the absence of Libertarian free will. Does it really matter whether they can walk the walk? (OK, duh, it matters, but does it matter to the assessment of Smilansky's view?) I think what matters is whether they talk the talk: whether they perceive their belief in Libertarian free will as a necessary underpinning of their moral behavior.

[tnadelhoffer]

Paul,
In one respect, you are certainly correct--althogh there is the assymetry between the participants' intuitions in the first and third person cases that must be taken into consideration. The only reason I think that it is more important to determine whether disillusionment would actually produce negative personal and social consequences than simply determining whether peolpe believe it would is that I presume that if the data showed (as ours does) that people don't share Smilansky's concerns, Smilansky would likely just respond by suggesting that all he was ever really interested in was whether disillusionment would actually be deleterious, not whether people believe it would be. So it would be nice to simply focus on that issue--although you and Matt may be right that settling that issue may take us beyond the confines of experimental philosophy. I, for one, see no obvious reason to assuming that to be the case. But as more people wrap their minds around some of these metaphilosophical and methodological issues, more light will be shed on the matter.

[Scott Hagaman]

Do you run any counterfactual control questions to test how good these students are at counterfactual reasoning?