Intuitions, Thought Experiments, and Expertise

Over the course of the last year, I have written extensively about intuitions, thought experiments, and expertise.

In "Experimental Philosophy of Language," a chapter co-written with Steve Stich for a Companion on the philosophy of language, I review the reasons for why a growing number of syntactians distrust the appeal to experts' intuitions, and favor the experimental study of linguistic intuitions. We argue that these reasons carry over to the linguistic intuitions philosophers of language have appealed to.

In "Expertise and Intuitions about Reference," forthcoming in Theoria (as part of an exchange with Michael Devitt), I provide some tentative evidence for the claim that expertise does not improve the reliability of experts' intuitions about reference.

In "Thought Experiments and Philosophical Knowledge," forthcoming in Metaphilosophy (together with some papers by Papineau, Williamson, etc.), I argue that a plausible view about intuitions (defended, e.g., by Williamson) leads to some form of skepticism about many, perhaps most, intuitions elicited by thought experiments (the argument differs from the one put forward by Jonathan Weinberg) I also say a few words about expertise. 

There are several positions that emerge from this body of work.

I reject the idea that philosophical expertise improves the reliability of intuitions in general. Whether or not philosophical expertise improves their reliability, and in which area of philosophy it does, is an empirical question, and available evidence casts doubts on an improvement of intuitions in ethics and philosophy of language. 

Furthermore, I hold that if intuitions are to play a role in some philosophical argument, they should be surveyed experimentally.

Finally, I argue that we have reasons to doubt that the intuitions elicited by many thought experiments are reliable, and this leads me to question whether that most of these intuitions can really justify the premises of philosophical arguments. 

Comments on any of these papers would be welcome.

Comments

[Jonathan Livengood]

Edouard,

I have a general question: if it turned out that intuitions in some area were stable and uniform, do you think that would give them any evidential value? Why or why not? And how much value?

[Peter B. Reiner]

I have been thinking a fair bit recently about the role of empirical data and normative considerations in bioethical discourse. Many commentators have suggested that there is a role for both - the folk view of contentious issues is worth considering, but experts who have reflected long and hard on the topic have a deeper understanding of the matter. Your position would suggest that we might consider further devaluing the normative.

On one level I think that is correct, but I do worry about the overall validity of non-reflective thinking on the part of the folk. Has any work in experimental philosophy compared responses when individuals are asked to answer quickly (as they usually do) versus asking (or even requiring) them to take some time and reflect on the issue? Would the results change?

[Damian Szmuc]

Edouard,

(again me, from Argentina)

I would like to ask a specific question.

If empirical evidence was likely to be found, and suppouse it is, and for the sake of the expertise defense it coincides with philosophers' intuitions for each time and for each topic we make experiments: wouldn't the claim that says "philosophers have correct intuitions because they are experts (and all what it entails)" rely itself in another intuitive theory (for instance, a theory about induction)?

I feel like even an extended empirical investigation of the rightness of philosophical experts' intuitions wouldn't reject a more general skeptic point of view, due to the fact that they [the experiments checking rightness of intuitions] would be only a weak support to philosophers' intuitions -just a defense of them considered as particular "outputs". That elicits, for me, the need to a more deep argument in favor of the role and importance (for the task of justification) of our intuitions generally considered as a human capacity or faculty.

But maybe I am wrong, and I would be pleased to know your opinion.

[Shen-yi Liao]

Hi Edouard,

The philosophy of language papers is very interesting. It's definitely worthwhile to think whether and how philosophical intuitions are like linguistic intuitions.

However, it is not obvious to me why syntactic intuitions are suitable comparisons for philosophical intuitions. Grammaticality judgments don't seem to be anything like responses to thought experiments. For one, if Chomsky's right, then grammaticality has the kind of innateness that philosophical intuitions typically lack. For another, linguists need not claim themselves to be the relevant experts, as philosophers do. I'm sure a Taiwanese linguist would agree that an ordinary native English speaker's grammaticality judgment is more reliable than hers. So it is puzzling to me why the methodology appropriate for syntax research would offer any guide to the methodology appropriate for philosophical research. The article does not spell out the link in detail.

If anything, semantic intuitions seem to be better comparisons, especially since philosophy of language more frequently overlap with semantics (see, e.g., the recent literature on epistemic modals) than syntax. But there, as you note in the article, less experimental research is done, compared to in syntax.

More importantly, a recent paper provides experimental support for the traditional semanticist methodology. The authors say, "We conclude that there is no empirical, logical, or statistical reason to think that the informal experiments routinely performed by linguists are unreliable. In fact, we show evidence that these experiments might be, in some circumstances, much more powerful than formal experiments with naïve participants." (http://ling.auf.net/lingBuzz/001075) So I am wondering how your claims would comport with this result.

[Edouard Machery]

Jonathan

Stability and uniformity would give us a reason to believe in their reliability, and they thus would have evidential value. Uniformity, for instance, suggests that demographic variables do not influence these intuitions. Thus, we have eliminated one reason to doubt their reliability, and I take this to be a reason to believe in their reliability.

Of course, there can be other reasons to doubt their reliability, which would have defeated if they are sufficiently salient.

Does this make sense?

Peter,

I doubt that the distinction folk/expert intuitions map neatly onto the distinction non-normative/normative. There are folk intuitions about the normative, and expert intuitions about the non-normative.

Concerning your second question, I do not know whether experimental philosophers have been doing exactly the kind of study you are interested, but there are studies that try to elicit more reflective intuitions. See for example Angel Pinillos's paper in Mind & Language. Justin Sytsma and I also have been doing this in a forthcoming paper in Consciousness and Cognition.

In any case, it is not entirely clear that slow intuitions are more likely to be true than quick intuitions. There is some work in the domain of preference that shows that intuitive choices are sometimes better than slow, reflective choices.

Damian,

It is very important not to challenge intuitions in general. This leads to skepticism, and this is not a position I endorse. My skepticism is, as I say in the paper, well-circumscribed: It bears primarily on intuitions elicited by many, perhaps most, thought-experiments.

Thus, it would not necessarily be a problem if a theory were supported by means of some intuitions, provided these are not of the problematic kind.

Furthermore, it isn't clear that a theory of induction would have to be defended by means of intuitions. See, e.g., Reichenbach's work on the question.

Edouard

[Clayton Littlejohn]

Hi Edouard,

I had some questions about your view of evidence. You say, "Can these judgments [about thought experiments] provide evidence for the premises of philosophical arguments? If they can’t, then thought experiments cannot play any evidential role in philosophy, and their contribution to the growth of philosophical knowledge is severely limited."

At various points, you write as if judgments provide evidence because the judgments are the evidence (e.g., "The judgment that it is permissible for the driver to turn the trolley onto the side track in the switch case is evidence that in the circumstances described by this case it is permissible for the driver to turn the trolley onto the side track.") Can't judgments provide evidence without being evidence? Can't friends provide happiness without being happiness?

Here's a worry about what might be an important assumption in the paper. Consider three claims about evidence:
(i) Evidence is propositional (why? Because evidence constitutes a reason that bears on whether to believe and nothing non-propositional could do that);
(ii) Evidence is factive (no false propositions constitute evidence);
(iii) Your evidence will include any proposition you know non-inferentially.

Now, consider three claims about what your evidence includes when you make a correct judgment in response to some thought experiment:
(1) It's the proposition that you've judged such and such that's evidence;
(2) It's the proposition that you judged to be true that's evidence;
(3) Both.

While your arguments (e.g., your remarks about the chair and judgments about chairs) might support (1), I don't see why they should lead us to reject (2) or (3). Sure, our evidence for believing that there are chairs here or there might include facts about what we believe, but why can't my evidence for the belief that there's at least one chair in the room include such facts as the fact that there's a chair right there (or that there's a visible object that has the look of a chair)? Seems like you assume that certain liberal forms of foundationalism are false (e.g., views on which we can have non-inferential knowledge of an external world based on experience).

I have two worries about the view that accepts (1) but rejects (2) and (3).

First, you get the skeptical conclusion trivially if the three claims about evidence mentioned above are correct. Now, you might have reasons for thinking that our evidence cannot include anything but facts about our own mental life, but I don't know what it is.

Second, you force those who want to resist the skeptical argument into an odd sort of view. Take the moral case. Here's what you say:
"The judgment that it is permissible for the driver to turn the trolley onto the side track in the switch case is evidence that in the circumstances described by this case it is permissible for the driver to turn the trolley onto the side track. The judgment that it is not permissible to push the large man in the footbridge case is evidence that in the circumstances described by this case it is not permissible to push the large man. That it is permissible to turn the trolley onto the side track but not permissible to push the large man onto the track reveals that there is a moral difference between the two situations"

One view that you might have is an intuitionist view on which some moral propositions are known non-inferentially. Which propositions is subject to debate, but suppose you go for a view on which they include verdicts (e.g., it's permissible to push; it's not permissible to push). An intuitionist might say that what you judge is evidence (e.g., that it's impermissible to push in the circumstances described is evidence against certain consequentialist views). On your gloss of what our evidence is limited to (if I get your view right), our evidence won't include the moral fact that it's impermissible to push, but the non-moral, psychological fact that that's what we think. But, if our evidence is so limited, why think that it confirms or disconfirms some view about general moral principles? What background assumptions would we need in place to think that our evidence raises the probability of some hypothesis when the evidence is psychological and the hypothesis is moral?

Later in the paper you offer some reasons for thinking that it's in some sense inappropriate to appeal to the sort of evidence that I've suggested the non-skeptic should say we have when we consider thought experiments (e.g., the observation that our intuitive judgments vary across cultures, that there's disagreement amongst the experts, etc...), which I think is important. Still, one way to raise the worry is just to say that given what you say our evidence is in the beginning (remarks that aren't supported by any empirical work), your view of evidence supports a skeptical conclusion on its own. Why did we need the empirical work since you could get your skeptical conclusion for free?

[Edouard Machery]

Sam,

In our chapter, Steve Stich and I are mostly focused on the intuitions that are relevant for the philosophy of language, and I am not entirely sure that the lessons we draw apply to all intuitions used in philosophy, although it seems likely that they (or many of them) apply to most.

There are obviously differences between syntactic intuitions and philosophical intuitions. However, note first that the extent of these intuitions depend on your views about the nature of philosophy. If you have an "internalist" view of philosophy (roughly, philosophy is conceptual analysis), then there are many important similarities between syntactic intuitions and philosophical intuitions. Particularly, in both cases, what one wants to understand by looking at intuitions is something in the head (UG or concepts). The similarities are less striking if you have an "externalist" view of philosophy (e.g., Devitt, Williamson, and others).

Second, I am not sure to see why the differences (whatever they may be) would prevent the concerns about experts' syntactic intuitions to carry over to philosophers' philosophical intuitions. Consider theoretical bias, or the possibility to ignore the diversity of intuitions due to the demographic homogeneity of experts (syntacticians or linguists), or the incapacity to understand how different cues influence intuitions without large sample sizes and factorial designs. Why would it matter whether intuitions are innate? Whether they are meant to provide evidence about something in the head or in the world? etc.

Thanks for the reference, but I was unable to get the paper you cite from (the link connects to dozens of papers!). Could you send it to me?

In any case, the use of experiments in linguistics is not entirely uncontroversial, as we mention in the chapter, I think.

However, from the quotation, I get the impression that the paper targets a strawman. Nobody holds that linguists' intuitions are often unreliable. Often, actually, there are no differences between experts and lay people. But they can be unreliable, and it is impossible to know when on a priori grounds.

As for their claim about the superiority of linguists' intuitions, since I have not read the paper, it's hard to assess. It is true however that the study of lay intuitions is not without problems.

Clayton: I'll try to respond to your post tomorrow.

Edouard

[Kristof Molnar]

Hi Edouard!

With a little google-fu, I found the paper you are looking for:
http://ling.auf.net/lingbuzz/@IhfZufFFZiIztOuA

I hope this helps.

[Edouard Machery]

Kristof: thanks

[Edouard Machery]

Clayton,

Thanks for this thoughtful comment, and sorry for the delay in responding.

I am not sure I will be able to do justice to all the points you raise, but I want to say a few things about the three views of evidence you mention in the middle of your post.

(Before doing this, I must confess that as a philosopher of science I find this way of thinking about evidence alien, which of course does not mean that it is misguided: I just do not have definite positions about this kind of issues concerning evidence. So what follows is a bit tentative.)

I am inclined to say that pieces of evidence are facts. A piece of evidence is yours (it's your evidence) when you know this fact holds. For counterfactual situations, pieces of evidence are the facts that would hold in them. I think this claim is compatible with (i) to (iii) above (although I am not sure that facts are necessarily propositional, but in any case facts give you reasons.).

Those facts include, but do not reduce to, psychological facts, including one's own and others' judgments. You and I observe an object that could be a chair (it has a weid shape...). You judge that it is a chair: your judgment (the psychological state itself) is evidence. The properties of the object are also evidence.

So, I do not accept (1) below as a general view of evidence. In fact, I endorse (3).

But I don't see how it helps because in the case of thought experiments the only facts that are part of your evidence are psychological - the judgments themselves.

Now, you seem to think that this is a very odd position, because it is very unclear why one would think that judgments are evidence. I am puzzled here: If people are reliable judgers, then their (honest) judgments are evidence. If they answer randomly, then their judgments are not. (There is probably more to be said here, but this line of response seems perfectly available.)

So there is a real empirical question: are judgments reliable? Evidence of bias, of divergence across demographic groups, failure to find improvement due to expertise seem relevant to answer this question.

I have to stop here for now, but I may have missed some important points in your post. Let me know!

Edouard

[Clayton Littlejohn]

Edouard,

Thanks for calling my comment "thoughtful", I thought it was a bit disorganized.

So, it looks like we agree on some points. You seem to agree that evidence consists of facts and that these facts needn't be facts about a subject's psychology. But, you still say that a subject's judgment is evidence. That's puzzling since I think judgments are in a different ontological category than facts. Is there an act/object ambiguity issue going on here? By judgment, did you mean something like what you judged to be so rather than the judging of something to be so?

Here was the sort of worry I had in mind when I posted. Your paper started out saying things like the evidence we have to evaluate a theory after reflecting on thought experiments are our judgments. (I took these judgments to be psychological episodes or thought you might have meant that it's facts about psychological episodes that gets into our evidence but never the facts that these judgments are about.) Given this conception of evidence and the equation of evidence with knowledge (one you might not like, but one that I think is pretty close to right), you seemed to be assuming a skeptical thesis from the get go. So, consider a simple and patently question-begging skeptical argument:
(SA1) In reflecting on a case (e.g., a trolley case), your evidence is not some moral fact (e.g., it's wrong to push, there's a reason not to push), but a psychological fact.
(SA2) p is part of your evidence iff you know p.
(SAC) Thus, reflecting on cases is not a way of coming to know some moral facts (e.g., that it's wrong to push, that there's a reason not to push).

Once (SAC) is accepted, it's hard to see how we could have good reason to then reject certain ethical theories such as simple forms of consequentialism given the evidence we had. After all, consequentialism is consistent with the evidence so understood. Moreover, it's not clear that the evidence so understood raised the probability of non-consequentialism. I don't know what background assumptions I'd have to have in place to argue that facts about my psychology raise the probability of some form of non-consequentialism.

I don't think accepting E=K (or some alternative view in the neighborhood of it) helps to deal with the skeptical worries that you raise in your paper. After all, it's plausible that if you confront me with lots of data that suggests that intuitive moral judgments about cases are unreliable, it's not then plausible to ascribe to me moral knowledge unless there's some reason to think I'm reliable in ways that others are not. And, given E=K, since my moral judgments don't constitute knowledge, my evidence will be quite limited. Presumably, it would be limited to non-moral facts. My worry was simply that the view of evidence you sketched at the beginning of the paper seemed to be one that got you a skeptical conclusion even without any additional support from the empirical observations that figure in the later parts of your paper.

So, I think we're in agreement that once certain empirical evidence is made available to the relevant parties, it will be harder to justify the claim that my evidence for rejecting consequentialism on the basis of thought experiment consists of the fact that it's wrong to push the man into the path of the trolley. That's because I think the empirical evidence can defeat the justification you might have otherwise had if you were non-culpably ignorant of that empirical evidence. I think where we might party company is in describing cases where the subject isn't yet aware of the evidence that suggests that her moral intuitions aren't reliable. Here, we might distinguish two cases:

Case One: The subject's moral intuitions are reliable and her moral beliefs are true and non-inferentially justified. In such a case, I'd want to say that her evidence could include facts such as these: there's a reason not to push the man in front of the trolley; it would be wrong to push the man, etc... Once she acquires the evidence that people like her are unreliable (evidence that's misleading, I suppose), her justification is defeated and her knowledge is lost.

Case Two: The subject's moral intuitions are unreliable, a fact she's non-culpably ignorant of. Perhaps her beliefs are justified in some sense, but they don't constitute knowledge. As such, we might not say that her evidence includes the fact that there's a reason not to push, it would be wrong to push, etc... Once she acquires the evidence that people like her are unreliable (evidence that's not misleading, I suppose), her justification is defeated and she never had knowledge to begin with.

My preferred description of Case One seemed to be at odds with your claims about judgments as our evidence if that meant our evidence was limited to psychological episodes or facts about such episodes.