On the Overselling of Experimental Philosophy

Responding to the Prospect article that we recently linked to, "Philosophy's great experiment",  Dwight Furrow has some thoughts up under the heading "Overselling Experimental Philosophy" -- and, to be clear, it is the the article that he is accusing of the overselling, not experimental philosophers themselves.  I do think that this is something of a concern; so much of the coverage that experimental philosophy has received really has been rather unfortunately hyped up, in a way that is understandable given the way that part of media's being media is that it hypes things up.  But this has led to some unhappy consequences (I'm sure many readers remember this fooferaw a couple of years back), and maybe goes some way towards explaining some of the strange attitudes that some philosophers have taken up in response to the body of work produced under the x-phi label. 

And I find myself in agreement with a very large part of what Furrow says in his post, and I suspect that most of the x-phi community would agree with much of it as well.  For example:

"X-phi is interesting because it might help philosophers do one part of their job. But it cannot solve philosophical problems."  (I think that latter sentence is supposed to be read as meaning something like "X-phi cannot solve philosophical problems by itself.")

"The real problem with some contemporary philosophy is not the absence of scientific data but the use of odd and fanciful scenarios like the Trolley Problem to unearth how we reason."

"Experimental philosophy may help us determine what people believe and how they respond to various situations. Thus, it can act as a check against unreflectively assuming our intuitions are shared."

Nonetheless, despite these agreements, there are also some parts of his post that I think don't quite hit the mark.  Furrow's post continues after that last quoted sentence as follows: "But brain scans can’t tell us much about why people think as they do, and tracking blood flow or electrical activity is not going to reveal very much about patterns of reasoning."  But brain scans can, in fact, tell us very important things about why people think as they do!  I take it that what's interesting about, e.g., Josh Greene's work in this regard is very much that it is providing some evidence for claims about what psychological factors are or are not in play when people make different sorts of moral judgments.

He continues: "Furthermore, questionnaires and observations of behavior are notoriously unreliable in explaining the motives behind our actions, and are hardly revolutionary."  Well... is any part of philosophy really concerned with "explaining the motives behind our actions"?  Have any experimental philosophers taken on such a question? And I don't see how armchair reflection is going to do a better job of offering such explanations.  This may be a pitch for doing more and better x-phi -- or just plain more and better psychology -- but I don't see how it is much of a critique of the movement as such.  (I'm also not sure what the point is about whether or not using surveys are "revolutionary".  Seems to me it's clearly not at all revolutionary in psychology, which is a large part of why we're willing to use it ourselves; but it clearly does represent a nontrivial methodological shift in contemporary philosophy.  What else is supposed to be at stake here?)

But, anyhow, that's just quibbling.  I think the deeper disagreement is here, and I think it also points to a deeper promise on the part of x-phi for the philosophical project on the whole:

"Most importantly, X-phi could not begin to tell us how we ought to think about reality. It is rooted in what is, not what should be. It can be critical of philosopher’s pretensions but not of the beliefs it purports to describe. It will not be making philosophical discoveries."

Surely Furrow would be right if all he meant is that we will make few direct discoveries by doing x-phi -- for no one thinks that we're going to just learn whether compatibilism is true, or what normative moral theory is correct, or whether knowledge really is stakes-sensitive, just by looking at people's survey responses and brain scans.  But that middle sentence of this last quote seems to aim for something stronger, and something that seems to me to indicate where things have gone off the rails here.  Barring some particularly pure form of rationalism, most all of our thinking about "what should be" is itself thoroughly and ineliminably rooted in our thinking about "what is".  Change your picture of what's going on descriptively in our minds and in our lives, and you will often change your picture of how our minds and lives should be directed.  Just consider, for example, how many issues in the philosophy of mind have been transformed by shifts in the relevant  sciences -- including, very much, our inquiries into the nature of rationality itself.  Or, to pull an example from another literature I'm interested in, consider how enriched recent discussions in aesthetics has been by its engagement with descriptive psychology of the emotions (e.g., in Jenefer Robinson's Deeper than Reason).  For exactly these reasons, I think we can expect that, pace Furrow, x-phi will indeed play a significant role in future philosophical discoveries.

Comments

[Mankel]

Furrow says: "Most importantly, X-phi could not begin to tell us how we ought to think about reality. It is rooted in what is, not what should be."

That kind of statement can only have a point if you detach the "thinking" from X-phil.

What I mean is that what the practitioners of "pure" philosophy don't get is that experimental philosophy is about using data to give the better possible answers to philosophical questions. Furrow writes as if X-phil was just the experimentation or the brain scanning, not understanding that the point is to USE those for the purposes of philosophical enquiry.

[Josh May]

I agree with Jonathan that Furrow makes some good points but still misses the scope of X-phi by holding that it (or presumably other empirical work) can't tell us how we "ought to think about reality" and so on. Mankel is right that people who hold this must be thinking that empirical science only involves generating data and not reasoning about that data. And of course that's not true. Empirical works involves lots of "pure" reasoning; it's just pure reasoning about empirical stuff. As Appiah aptly put it in his awesome NY Times article:

"You can conduct more research to try to clarify matters, but you're left having to interpret the findings; they don’t interpret themselves."

This is largely why I don't get why people think there's a fundamental difference between empirical science and philosophy. If philosophy is just thinking hard about stuff and science is just the gathering of data, then the people we normally call "scientists" are philosophers and their research assistants who gather the data are the only real scientists. :)

[Jussi Suikkanen]

Jonathan,

about this:

"Surely Furrow would be right if all he meant is that we will make few direct discoveries by doing x-phi -- for no one thinks that we're going to just learn whether compatibilism is true, or what normative moral theory is correct, or whether knowledge really is stakes-sensitive, just by looking at people's survey responses and brain scans".

I wonder if this is a bit of an overstatement. Here I quote Thomas Nadelhoffer and Eddie Nahmias's Phil Explorations paper The Past and the Future of Experimental Philosophy:

"Moreover, Greene has gone on to argue that data from his studies on moral cognition [I take it this refers to the dual-processing material] not only support sentimentalist over rationalist theories of moral psychology, but they also purpotedly provide us with reasons to prefer consequentialist theories over their deontological counterparts".

That seems like a fair characterisation of Greene's work (wish I had relevant papers with me to check). If that's right, then there seems to be someone who thinks x-phi can reveal what the correct moral theory is.

[tnadelhoffer]

Jussi,

I think the best example of Greene's work along these lines can be found in "The Secret Joke of Kant's Soul." But I don't think this is limited to his work. In some sense, I think whenever experimental philosophers float performance error theories, they are flirting with using their data to settle first order philosophical issues even if that is not what they are aiming to do.

For instance, if I can show that someone's intuitions about x are produced in biased or erroneous ways or by affective processes not usually associated with rational deliberation, why not say that I have put pressure on not only the reliability of those intuitions, but also their veracity?

Take, for instance, the work on free will by Nichols and Knobe. The explanatory model of their data they seem to prefer is that the non-affect driven incompatibilist intuitions are representative of people's real intuitions about free will while their compatibilist intuitions are the result of an affect driven performance error.

Of course, it is open for them to suggest that people's real intuitions are wrong and the intuitions generated by performance errors are actually correct--but then it is hard to know why we ought to treat the latter as instances of a performance error.

I have to admit that this is an issue that I have been meaning to write about. More specifically, I am interested in the following line of reasoning: The intuition that x is the case is produced by processes in the pre-frontal (or frontal) cortex--the more phylogenetically recent and sophisticated part of our brain normally associated with rational deliberation. The intuition that x is not the case, is produced by a less sophisticated, less deliberative, phylogenetically older part of the brain (e.g., the amygdala). Why not think this is relevant to whether we ought to believe x or not x? For instance, if Kantian deontological intuitions are driven by what some call the "reptilian" brain, why not think we ought to jettison them in favor of the utilitarian intuitions generated by the more sophisticated part?

I take Greene to be adopting this sort of line.
And while I, for one, am strangely comfortable with the move myself, that doesn't mean that I don't worry that lots of questions are getting begged along the way.

Any thoughts?

[jonathan weinberg]

Jussi, I think you've missed the importance of the word "just" in that bit of text that you quoted, in "_just_ by looking..." My claim is very much _not_ that x-phi won't be relevant to debates at those levels, but rather that no one here takes any x-phi result _in and of itself_ to settle such debates. Greene (and Thomas, in the argument he just sketched) are of course bringing still other premises to bear beyond the x-phi results proper, and they are premises that are not themselves just pieces of experimental findings. (They are premises that I think are a tad dubious, btw, but that's another matter!)

[Jussi Suikkanen]

Jonathan,

sorry about overlooking the second 'just' in the sentence. It seems like the auxialiary premises are often not made very explicit so sometimes you do get the feeling that it is the experimental results in and of themselves that settle the debates in the Greene material. But I aknowledge that this isn't fair. It could be made more explicit thought. In the same way I often feel that Libet is trying to solve the free will questions by merely looking at the experimental data but it's true that Mele again is right - some more non-experimental premises are required which can be questioned.

Thomas,

I've been thinking about similar issues. Here's my reason for being sceptical about the Greene's inference that because the deontic judgments are produced by the emotional part of the brain they must therefore be less reliable.

I know you always lose an argument if you mention Nazis but I'm going to intentionally lose the argument here. In the many psychological accounts of Nazis, it's reported that they felt psychologically conflicted. They reason told them to exterminate the Jews but they could never get rid of the nagging feeling of sympathy that was making many of them uncomfortable and even sick.

Now I can imagine an experimental philosopher in this situation doing the brainscans and showing how the kill Jews judgment is created by the reliable reason part of the brain and the feeling of sympathy by the unreliable emotional part. But it seems like a bit of reductio to conclude from this that therefore the Nazist ethical view is correct. So it seems like emotions can in contexts be more reliable morally than reason. I know all kinds of questions are begged here - so sorry about that.

[Bryce Huebner]

As many of you probably know, I have a lot of skepticism about what we might learn about cognition by looking at the results of fMRI scans.

I think that it is important to note that there are some people *even in the brain sciences* that are worried about what we have actually learned from doing cognitive neuroscience. In my neck of the woods, for example, there are clearly brain scientists who are skeptical of what cognitive neurosciences might show us. There is a nice debate on this stuff here:

http://video.google.com/videoplay?docid=-4182662023868835110&hl=en.

(At about the 1 hr mark, Alfonso Caramazza expresses his worries about what can be inferred when fMRI shows that there in increased hemodynamic activity in some region of the brain that don't overlap with areas that, when damaged, yield selective processing deficits).

Beyond these debates, there are recent worries about how hemodynamic activity is actually correlated with local neural activity (see Sirotin and Das, Nature 2009)--it seems that the answer to this question isn't nearly as clear as people have tended to think.

I am worried in general that we don't have a clear idea yet of what we fMRI is really telling us! Thomas, you make the following claim: "if Kantian deontological intuitions are driven by what some call the "reptilian" brain, why not think we ought to jettison them in favor of the utilitarian intuitions generated by the more sophisticated part?"

But there are deep problems here. First, there is a lot of debate in the cognitive neurosciences over how to understand the what the amygdala, and the system in which it is embedded, is doing in our cognitive processing. A lot of people are starting to argue that the amygdala is part of a valuation circuit, and that this circuit can be co-opted by higher cognitive/conceptual structures. If this is the case, I don't see how your argument would go through.

To me, the key question here is not about what structure happen to be active; instead it's a question about the relationship between these affective circuits and higher cognitive circuits. Are we really sure that in the course of development, or even in the course of responding to a particular stimulus, the higher cognitive circuits aren't recruited first, only later activating the valuational circuits in the 'reptilian brain'?

I am inclined to think that we just don't have enough data yet to lay out the precise temporal and causal structure of moral cognition (but, I guess you might have already known that I thought that). The short version of my worry is this: fMRI is a correlational method that is not all that great on questions about the time course of processing. That being the case, the existent data is insufficient to license descriptive claims about the processing involved in moral cognition. So, the attempt to undercut this or that moral theory by appeal to these neurological data strikes me as way beyond the pale of plausibility.

Of course, this isn't to say that other experimental methods don't have important implications for first-order moral theory. My claim is just that we don't have the data to rule out deontological theories (...I can't believe that I just said that...)

[Dwight Furrow]

Jonathan,

Thank you for your thoughtful remarks on my post. As you rightly point out, I am not hostile toward X-phi but rather toward the way it is portrayed in the article in question. But I did want to clarify a couple points. (I apologize in advance for the excessive length of this comment)

X-phi seems to offer the promise that a lot of the intuition-mongering that philosophers often engage in can be replaced with something else that will be more reliable. It seems to me that we don't know yet if that promise will be made good. The scope of X-phi will largely depend on the course of cognitive science. If some of the more reductionist approaches pan out, then perhaps we won't have to rely on intuitions so much. But if the brain is more holistically organized, talking about intuitions may be the only way we can make use of the scientific data. Of course, in that case, our talk about intuitions will be better informed, and X-phi will be a useful tool.

Furthermore, given current technology and understanding, brain scans tell us what general cognitive functions are at work in an agent's response to a situation, but that is not quite the same thing as identifying the complex patterns of reasoning that philosophers try to articulate.

You say: "Well...is any part of philosophy really concerned with "explaining the motives behind our actions?" Have any experimental philsoophers taken on such a question? And I don't see how armchair reflection is going to do a better job of offering such explanations."

It seems to me that much of what we call "moral psychology", with traditions reaching back through Nietzsche, Hume, Aristotle, etc. were in part trying to do just that. Talking to lots of people in many contexts, asking them questions, and observing their behavior is certanly superior to philosophers sitting in a room talking to each other. But you are still talking, observing, and interpreting behavior--but with a larger and more diverse data base, as I suggested in my post. It is the same inquiry and practice, but with more refined and informative tools.

However, you are right to say that there is a larger issue at stake. You say: "Barring some particularly pure form of rationalism, most all of our thinking about "what should be" is itself thoroughly and ineliminably rooted in our thinking about "what is".

This is the issue.

Let me put my point as follows. I have found the work of Joshua Greene enormously useful both in my writing and my teaching. The main points I take away from Greene's work are that our moral decisions are bound up with our emotional responses and are more conext sensitive than the predominant rationalist traditions in ethics had assumed.

I suspect that if we could resurrect Kant and present him with Greene's data, he would say "But of course human beings are emotional and swayed by context! That is why I wrote the Grundlagen--to warn us against it. You have learned something I already knew."

The difference, today, is that many of us in moral theory have made the decision to link "what ought to be" much more tightly with "what is." But we have made that decision, not because of the penetrating insights of X-phi but because we were uninspired by the excessive abstraction and irrelevance of much 20th Century ethics. In other words, it was a philosophical judgment, not a matter of empirical inquiry, that drives this movement. If this constitutes a revolution in philosophy (the jury is still out) it was a philosophical revolution, not one forced on us by psychology experiments.

We (or I) no longer want moral theories that seem to ignore or do battle with human psychology. But notice that once we have made that decision, we still must decide how far our theories can push the boundaries of human psychology. The word "ought" still has a meaning--otherwise we fall into mere conventionalism. Again, these are matters of philosophical (albeit better informed) judgement. (Although I suppose the answer to this in part will depend on how plastic the brain turns out to be.)

I am not quite prepared to say that X-phi is pouring old wine into new skins, because I think you have improved the wine. But it retains a very classic nose.

[Edouard Machery]

Following on Bryce's points, it is also worth noting that there are some problems with the data analyses done by social neuroscientists using fMRI.

See particularly Vul et al Voodoo Correlations in Social Neuroscience. Perspectives on Psychological Science,

Quite sobering.

E

[Joe Paxton]

I'm probably going to regret doing this. But, somehow needs to inject some optimism into this thread...

Following up on Edouard's point, there's growing consensus that Vul et al's conclusions were vastly overstated--so much so that they were forced to change the title of the paper:

http://www.edvul.com/pdf/Vul_etal_2008inpress.pdf

Everyone seems to agree that the correlations in question are in fact biased. But, if you look at Vul's data closely, the bias is rather smaller than the word "voodoo" might suggest: avg r=.69 for the "voodoo" correlations vs. avg r=.57 for the non-voodoo correlations. To think of it in terms of effect size: That's a difference in explaining about 1% of the variance in behavior.

Lieberman et al. make this point and others in their reply, to be published alongside Vul et al. in Perspectives:

http://www.scn.ucla.edu/pdf/LiebermanBerkmanWager(invitedreply).pdf).

For what I think is an even better reply, see the slide presentation by Bob Cox, a mathematician who writes FMRI data analysis software at the NIH (see especially slide 17):

http://afni.nimh.nih.gov/pub/dist/doc/misc/voodoo.pdf

What's nice about Cox's reply is that he doesn't have much of a dog in the fight. So his take is probably less biased than the official replies and rejoinders.

Anyhow, I don't mean to hijack the thread by changing its focus to that one methodological issue. There's plenty to read about that elsewhere.

I'd be interested to hear more from Bryce. The relationship between the FMRI BOLD response and the underlying neural activity isn't yet fully understood, true enough. But we pretty clearly have enough convergent evidence from single cell recordings in monkeys and from lesion patients to be highly confident that the BOLD response can give us quite a bit of theoretically relevant information about human brain function, especially when it's supplemented by TMS, EEG, below-the-neck physiological measures, lesion data, and obviously lots of behavioral data from normals.

(Caramazza's point is not that we can't reliably infer theoretically interesting conclusions from FMRI data. Quite to the contrary: he often does so himself. Rather, his point is that such data is ideally supplemented by other methods for maximal theoretical payoff. He happens to prefer patient data, given that he has access to some patients who have theoretically relevant lesions.)

There's a lot of work left to be done. But it doesn't follow from this that we shouldn't try to draw some theoretically interesting conclusions based on the best data we currently have, keeping in mind (as ever) that scientific conclusions are only ever probabilistic. If we had to wait until we knew the "precise temporal and causal structure of [X]" before we were able to make licensed descriptive claims about the processing involved in X, then even folks in vision science--the most mature sub-field of psychology--would not yet be able to make many licensed descriptive claims.

Instead of being concerned about whether our descriptive claims are licensed, we should be concerned about whether our descriptive claims are true. And whether or not our descriptive claims are true is a debate that will be settled in the journals, not in the blogs. So, for now, I'll leave it at that.

Back to your regularly scheduled skepticism...

[Bryce Huebner]

Joe,

I would second your skepticism about Vul et al's skepticism. While it is pretty clear that were were some offenders in social neuroscience, there were plenty of people who were branded as using bad assumptions in their data analysis who have been more careful than the Vul et al paper sugests.

Also, I think that we are agreed on the skeptical worries about imaging results suggested by Caramazza. The worry was merely that there are some cases where we don't get a clear causal story because legion data and imagining data don't line up nicely. That said, the use of methodologies like TMS surely gives us much better traction on the causal structure of the brain. In the area of moral judgment, for example, the paper by Daria Knoch et al (Science 2006) using rTMS over right DLPFC gives us a lot of insight into the mechanisms that subserve judgments of reciprocal fairness.

My larger problem is this: when we move to more complex phenomena like moral judgment, we don't yet have a clear idea about all of the mechanisms that are operative. We don't, for example, have a complete account of the temporal and causal structure of the neurological mechanisms responsible for our moral judgments. So, I would say: keep doing the work! and: keep adopting the theoretical views that are going to drive the research forward even if we don't have a clear idea about whether they are right yet! That's the only way to do the science.

I didn't make it clear before that I do think that there are some descriptive claims that we can rule out on the basis of the data that we have at this point. Moreover, there is no doubt that as we get more data, we are narrowing the range of plausible descriptive positions. However, with that said, to get to the claim that I took Thomas to be toying with (i.e., deontological views are false because they are driven by evolutionarily old structures), I think would need to have a lot more of the descriptive territory mapped out than we do right now. So, I think that there is a big difference between what we can learn *qua scientists* from the activity reported in these studies, and what we are warranted in inferring from the data *qua philosophers*.

Finally, I agree that there is a lot of cross-task correlation between single cell recordings, legion studies, MEG studies, and increased BOLD in fMRI studies. This does give us strong reason for thinking that hemodynamic methodologies are likely to provide us with good evidence about the regions of cortex that are active for a particular task.

The worry that I was trying to present, which went by far too quickly, is derived from a beautiful study by Sirotin and Das. They used dual-wavelength optical imagery to measure oxegenation and cerebral blood volume in V1 of awake monkeys. They found that although they were able to recover one signal that was indicative of a causal relationship between cerebral blood flow and the presentation of stimuli, they found an additional signal, which was just as strong, that appeared to entrain to the temporal structure of the task *independently of the actual presentation of a stimulus*. This second signal appears to be a predictive signal that entrains to the task and that increases blood flow in anticipation of the presentation of a stimulus.

The worry, then, is that because fMRI cannot decouple these signals, it's not obvious that in every task the BOLD signal is going to be as tightly, and linearly correlated with task-relevant processing as we have typically supposed. Note, however, that this is not a worry about whether fMRI tells us anything interesting about regions of cortex that are implement a particular computational task. I have no doubt that it often does.

My claim was merely that we need to be careful making claims like the one that was suggested by Thomas--and that is why I was trying to offer some reasons to worry about the causal relations that might not be picked out by fMRI--that, perhaps, we can't yet infer the falsity of deontology on the basis of existing data.

[Edouard Machery]

Joe. Thanks for the references. I am still astonished hat people could make such a simple mistake. The autheors are right that this would not be tolerated in a behavioral paper.

I think that Bryce underestimates he frequency of the mistake. Look at the list of offenders and you will find many papers many of which are very well-known.
E

[Bryce Huebner]

Edouard,

I have read the paper and looked at the list of the condemned. I too am astonished that the data was analyzed in this way. However, for each of the experiments in question, to know whether the results are as bad as this suggests, we would need to know 1) how the data come out given better techniques of data analysis; and 2) whether the simple responses that people gave to the survey sent out by Vul et al adequately tracks the analyses that they actually carried out. That said, I admit that the critique by Vul et al is troubling, and I hope that it will lead people to be more careful with their statistics (for that reason, i am glad that they changed the original title!).

[Joe Paxton]

Edouard: I'd highly recommend that you check out those replies, especially Cox's presentation. I think you'll temper your position a bit once you see how much Vul et al. were overstating their case. I tried to give you a flavor for it above, but perhaps I didn't do a very good job of it.

In any case, I'm sure you'll see less non-independent ROI selection in the future, and perhaps that will be for the better. But I fear that that paper has done more harm than good. The bias that Vul et al. have highlighted is quite unlikely to make a difference to the substantive conclusions drawn in the papers that they flagged, for reasons outlined in the replies. But it could make a difference to the confidence that people (and especially funding organizations) have in social neuroscience results, such that we'll all be the worse for their overstatement and the consequent attention the paper has received.

Bryce: Thanks for the thoughtful response. You seem to think that we need to have a _complete_ account of the temporal and causal structure of the neurological mechanisms responsible for our moral judgments in order to assess which theory is best supported by the current evidence. Perhaps this is where we part ways. I don't think a complete account is needed to make such an assessment. Could we have assessed whether evolution by natural selection was the best explanation for the diversity of species prior to the discovery of the double helix? I think we could have. This suggests that theory selection doesn't necessitate a complete account of the causal and temporal structure of the phenomenon under investigation. The same is true in the case of moral judgment. We haven't yet falsified the claim that the affective responses involved in non-utilitarian (deontological) moral judgment are triggered by more reflective cognitive processes, and in that sense we lack a complete account of the causal and temporal structure, etc. But we do have evidence (to pick one particularly relevant finding) that cognitive load selectively interferes with utilitarian, and not non-utilitarian, moral judgment. All else equal, such evidence should lead us to prefer the theory that has it that only utilitarian responses are consistently underwritten by reflective cognitive processes. Don't you think? (And all of that without brains!)

I haven't read the Sirotin and Das paper, yet, so I'm going to have to do so before I comment in detail on that part of your reply. Just to give you my initial impression: I would think that a signal which is indicative of stimulus anticipation would be equally present in all conditions, and so would wash out in a subtraction, assuming you have well-matched, randomly presented conditions. But perhaps I'm not really getting the significance of the finding. I'll check it out and let you know if I have any additional thoughts.

[Bryce Huebner]

Joe,

I think that you're missing the point. I acknowledged that "there are some descriptive claims that we can rule out on the basis of the data that we have at this point." Perhaps claim that "utilitarian" and "deontological" judgments are subserved by identical neural processes is one of them (the scare quotes are there because you know as well as I do that the way in which these terms have been operationalized in moral psych has little to do with the moral theories in question).

So, cog load generates interference with "utilitarian" judgment but not "deontological" judgment. But now what. Can you please fill me in on the argument that is supposed to run from this claim about *differences* in the effects of cog load to the claim that "only utilitarian responses are consistently underwritten by reflective cognitive processes." That move seems to come way too quick on your part.

In the Greene et al cog load paper, the response times for non-utilitarian judgments are hovering at the 5500-6200 ms range (just eyeballing it). Are you really trying to tell me that there is no reflection and cognition going on at 6000ms (high-utilitarian participant/deontological judgment), but there is at 6200ms (high utilitarian/utilitarian judgment? Something seems a little fishy to me here.

[Joe Paxton]

Okay, Bryce, since you asked, here is a more detailed argument. Keep in mind that this isn't an officially sanctioned Greene Lab argument. It's just my best attempt at giving the kind of line that Josh tends to push a decent defense within the context of this thread.

So, again, thinking about the cog load study, if controlled cognitive processes underwrite non-utilitarian (deontological) moral judgment, and if a cognitive load manipulation slows utilitarian moral judgment (relative to no-load), it should also slow non-utilitarian moral judgment (relative to no-load). So, if the cognitive load manipulation slows utilitarian moral judgment, but doesn't slow non-utilitarian moral judgment, this gives us reason to believe that controlled cognitive processes consistently underwrite utilitarian moral judgment, but not non-utilitarian moral judgment, all else equal. And it turns out that cognitive load does selectively slow utilitarian moral judgment. So this experiment gives us reason to believe that controlled cognitive processes consistently underwrite utilitarian moral judgment, but not non-utilitarian moral judgment.

What gives this argument its force is that it stipulates that the load manipulation must work in the case of utilitarian moral judgment before we can believe that the failure of the load manipulation to work in the case of non-utilitarian moral judgment tells us anything of theoretical interest about non-utilitarian moral judgment. This makes the failure of the manipulation in the case of non-utilitarian moral judgment more than just a null result. If the reverse had been true--if the load had slowed non-utilitarian but not utilitarian moral judgment, this would similarly give us reason to believe that controlled cognitive processes underwrite non-utilitarian but not utilitarian moral judgment, all else equal.

Of course, all else would not be equal in that case, because we have independent evidence from neuroimaging that implicates controlled cognitive processing in utilitarian moral judgment, but not in non-utilitarian moral judgment. I suppose you're likely to say that the neuroimaging evidence is "merely correlational," and this phrase is somehow supposed to render that evidence almost completely uninformative with respect to this debate. There's an interesting discussion to be had about that issue, but we don't have to have it today because, again, the cog load study didn't show that non-utilitarian responses were selectively slowed, it showed that utilitarian responses were selectively slowed.

Additionally, you ask why it takes people >5s to make a judgment both in the case of utilitarian and non-utilitarian responses. This is plausibly because all the dilemmas were "high-conflict" (e.g., crying baby). Remember that the hallmark of high-conflict dilemmas is that they take awhile for people to resolve--that's just why we call the dilemmas "high-conflict." So subjects exhibit a baseline level of conflict, no matter whether they end up going utilitarian or non-utilitarian. This baseline level of conflict is reflected by the 5s delay. Further support for this baseline conflict hypothesis is provided by the fact that the high-conflict vs. low-conflict comparison (collapsing across utilitarian and non-utilitarian judgment) from Greene et al. 2004 yielded a robust activation (450 voxels) in ACC, which has previously been implicated in response conflict. (But, again, you could say "merely correlational," "reverse inference," and whatever other buzzwords are supposed to block the use of neuroimaging data in psychological argument. So note that this is _further support_ for the baseline conflict hypothesis. The main point is that the 5s delay is plausibly explained by the fact that high-conflict dilemmas simply take awhile to resolve.)

Just to try to be maximally clear, here's what's most important to note with respect to the baseline conflict hypothesis: what accounts for the load effect is not just that you have conflict (again, there's conflict in both judgment types), but rather that when (and only when) the conflict gets resolved in favor of utilitarian judgment, the judgment is slower under load than without load. This suggests that there's additional controlled processing, _over and above the baseline conflict_, occurring in the case of utilitarian judgment, but not in the case of non-utilitarian judgment. Does all of that make sense?

Regarding the appropriate use of "deontological": you're probably already familiar with the argument that "deontological" is a natural kind term that refers to the output of the intuitive system (and likewise for "utilitarian" and the controlled system). (If you're not familiar with the argument, see Josh's _Kant's Soul_ paper for the first version of it.) That turns out to be quite a lengthly argument, based in part on the evidence cited above, and on an evolutionary claim, which itself gets its own argument. So I obviously haven't yet (and don't plan to) rehash all of that here, which is why I've been using "non-utilitarian" rather than "deontological" above.

In fact, it turns out that it takes about 400 pages to lay out the argument and all the relevant evidence so as to make the strongest possible case (given current evidence) for the dual-process model and its implications for normative moral theory. So I surely am not doing Josh justice here. If you're not yet convinced (and I doubt you are, if previous posts are any indication), I'll just refer you to the book, which will likely be out early next year, and I'll add that additional studies utilizing, for instance, TMS and patient populations are currently underway, and are unlikely to make it into the book. So keep an eye out for those as well.

(Have we learned anything about whether or not experimental philosophy is being oversold? I somehow lost track...)

[Bryce Huebner]

Joe...

Thanks for the detailed reply! I'm not convinced, but not for the reasons that you worry about. I don't think that the 'mere correlation' claim is sufficient to rule out the value of cog-neuro data; and I'm happy to concede a lot of the points that you want to make. I am skeptical about the way in which the term 'controlled' is being used...but that's a long argument.

I don't have the time to give you a lengthy response right now. But I can either give you one later on the blog...or over a beer if you prefer!

(and I think you're right that arguing about the data here probably isn't making much headway on the question of whether x-phi is being oversold)

[Joe Paxton]

I'll opt for the beer.

[Eric Schwitzgebel]

Jonathan, I was under the distinct impression that it was a kerfluffle, not a fooferaw. However, I am open to arguments on the point.

[jonathan weinberg]

Clearly a survey would be in order, Eric.