Is the Armchair Sexist?

Since Plato, intuitions have played considerable evidentiary roles in shaping philosophical discourse.  Yet, given the gender imbalance in professional academic philosophy today, intuitions imported from the armchair mostly belong to a bunch of men.  But do the epistemic intuitions of male philosophy professors represent consensus?  A growing body of empirical evidence suggests they don’t.

In a new paper [Gender and Epistemic Intuition], I present a series of experimental studies showing that men and women intuit about important epistemological thought experiments in surprisingly different ways:

Epistemological Purism.  As many may recall, the Epistemic Side-Effect Effect (Beebe and Buckwalter forthcoming) is an extension of the original side-effect effect for folk knowledge attribution.  We found that participants were significantly less likely to say an agent knows an action will bring about a certain side effect when the outcome is good, and more likely to attribute knowledge when that outcome happens to be bad.  This effect was highly variable to the factor of gender, whereby women were significantly less likely than men to ascribe knowledge to a subject just because the outcome was good.

Gettier Cases.  While there has been some experimental evidence that the Gettier intuition varies among and between cultures (Weinberg et al. 2001), another groundbreaking study has shown that men and women also reason very differently about these kinds of cases.  Starmans & Friedman (2009) report that, pace professional philosophical intuition, women are far more likely than men to deny the so-called Gettier intuition, and attribute knowledge to a putative knower in the particular Gettier vignettes tested. 

In the paper, I develop and test a hypothesis that explains these differences under a single mechanism.  I argue for what I call the normative evaluation hypothesis, or the idea that the gender differences in epistemic intuition arise as a result of the way men and women consult normative factors in formulating their intuitive judgments across domain.  My aim is to show that such demographic variation in intuition seriously undermines an extensive group of epistemological projects that rely on intuitional evidence.  For on what grounds is an intuition characteristic of one group (men) more privileged than another group (women) in that it counts as better evidence to support certain analyses of knowledge?

Additionally, these data may also help explain one factor that contributes to the discrimination against women in philosophy (see Haslanger 2008), as well as the alienation that women sometimes feel in the classroom.  For instance, imagine being a young female philosophy student, presented with the scenario of a famous thought experiment like Gettier, and then asked a question designed to reveal a particular philosophical intuition.  Now imagine that your intuition does not agree with not only the males in the class, but also with the male philosophical majority.  When you try to express that intuition, your professor insists that you are wrong.  It is not hard to understand why any person in this situation would feel alienated.  But why think that the professor’s intuitions about knowledge are better?  In other words, what possible reason could there be to suppose that the content of men’s epistemic intuitions are any more likely to be true than women’s?

Since intuitions from thought experiments are relied on so heavily in professional analytic epistemology, could the lack of tolerance for different intuitions than the ones men tend to have—even if perhaps ever so slight in some cases—be a factor that continually discourages women from entering professional philosophy today?

Comments

[s. welch]

I am so thankful and relieved you are doing this work. I cannot explain how hard it is to have males professor dismiss what you think is so clearly and obviously right intuitively, especially when you're right on track with other women. Looking forward to reading this!

[emily w.]

I echo S. Welch's sentiments. Much thanks for the work you're doing.

[Jennifer Nagel]

I'll be interested in seeing the Starmans and Friedman study when it comes out.

Meanwhile, I'd urge some caution in interpreting the results of any study that seems to show differences in the ways in which men and women ascribe knowledge. In general, mental state ascription is not something that exhibits gender variation. Developmental studies on the emergence of the capacity to distinguish belief from knowledge typically do not show any main effect of gender, admittedly because developmental studies typically work with relatively small samples, and if there is an effect it is so mild that only very large numbers will show it. The best analysis of the issue I've seen is Charman, Ruffman and Clements's "Is there a Gender Difference in False Belief Development?" -- Social Development 2002 -- which concludes that there is at most a weak difference (on average, girls pass the false belief test slightly earlier than boys do).

As far as the adult population is concerned, Charman et al note a few studies that have suggested a slight female advantage, but others that fail to show any such effects. Setting aside data from individuals on the (majority male) autism spectrum,there is no reason to think that men and women are doing anything different in assessing what others think and know.

I trust Walter's report that Starmans and Friedman got performance differences between men and women for the particular probes that they tested. But I think that before I'd start thinking that men and women have fundamentally different conceptions of what it is to know something, I'd want to be extremely careful to rule out performance factors as the basis of the different responses. If narratives are more engaging to one gender than to another, responses may differ simply because some of the subjects are frankly not paying all that much attention to what is going on in the story.

The project of looking for cognitive differences between men and women is something that has been pursued with massive enthusiasm by a wide range of researchers with a variety of interests, noble (as Buckwalter's seem to be) and not-so-noble. It's not hard to get men and women to say different things, to respond to the pragmatics of an experimental situation differently, to be more or less motivated by a given task. But before we conclude that there's a fundamental cognitive difference, let's be very sure that we've isolated the relevant factors and standardized the task appropriately for all participants. It's not surprising that men and women might have different levels of interest in a little story, or that men or women might identify more or less with the protagonist. I'd find it very surprising if there's a systematic difference between men and women in understanding the verb "to know".

[Wesley Buckwalter]

To S. Welch and Emily W,

Thank you for your comments and encouragement! I would be grateful to hear reports from others who also have personally found themselves feeling alienated in the classroom just because they had different epistemic intuitions than the majority.

To Jennifer,

Thank you for your comments! I agree that as experimental philosophers we need to be extremely careful about pragmatics, and that reaching the conclusion of “a fundamental cognitive difference” on the basis of these studies may be a bit premature at this stage. In the paper, I take pains in several places to caution the reader as to the preliminary nature of the results, and to encourage the need for future research as you suggest. I also agree that it is in principle possible that the differences detected here are due only to factors like interest or empathy to particular experimental stimulus, and we’ll only find that out by more experimental epistemology. Specifically, we’d want to replicate the current effect under a wide array of conditions, as well as design experiments to test for evidence that the gender effects presented here don’t arise entirely due to pragmatic factors as you seem to predict.

Given the weight you put on the absence of demonstrated gender differences in mental state attributions, I wonder if you could point me to some discussions of pragmatic differences between men and women to short experimental vignettes. What kinds of factors elicit such differences? Do you have any suggestions of factors that might have elicited them in the studies I discuss?

While such studies are important, we should also be careful to guard against what seems to be a faulty inference when dismissing the current data: that since studies haven't shown demographic variation in a related area of mental processing, there isn’t such variation here. For instance, I agree that passing the false belief task is an important developmental milestone, and that attributing false belief involves a host of different cognitive abilities, one of which including knowledge ascription. But is it still possible that while past experiments in theory of mind have detected little difference in the developmental schedule in one test regarding a certain kind of false belief, men and women might still have genuinely different epistemic intuitions in another test, regarding a famous philosophical thought experiment? Does the former really give us “no reason to think that men and women are doing anything different in assessing what others think and know”?

As you note in closing, “I'd find it very surprising if there's a systematic difference between men and women in understanding the verb ‘to know’”. I agree it would be very surprising indeed if there were a large, central difference in how men and women understand that concept. But, after all, one major response to the Gettier cases for instance has been that they are fairly weird and unusual cases. It wouldn’t be terribly surprising if we’d failed to notice a systematic gender difference in the concept of knowledge, say in the various iterations of false belief task experiments, if that difference was at the fringes of the concept.

[Jennifer Nagel]

Hi Wesley,
I don't have access to the stimuli in the Starmans and Friedman study, so I don't know what factors might be playing a role there.

There's a nice discussion of the role of pragmatic variation in Hugo Mercier's "On the universality of argumentative reasoning" -- which can be found here: http://sites.google.com/site/hugomercier/publications.

I don't think we have any good positive reason to expect a systematic gender difference in the concept of knowledge, even one that would emerge only for ecologically rare cases. Positing gender-based conceptual variation is I think a relatively expensive move to make, in terms of increasing the burden of explaining the acquisition and transmission of the concept; I'd be very keen to investigate less costly options first.

[larryniven]

"If narratives are more engaging to one gender than to another, responses may differ simply because some of the subjects are frankly not paying all that much attention to what is going on in the story."

While this does impact the specific path taken in the paper, I don't think it has quite the saving effect that you're saying it does. Regardless of the stage at which a cognitive difference is introduced, the data is the data: that the women and the men in this paper thought differently is pretty well unquestionable, isn't it? But if that holds up in the general population - which I agree still needs to be demonstrated by way of more experiments along these lines - then women and men do indeed happen to think differently.

There are still places to be cautious, I think, but they're elsewhere. For instance, one needn't say (and I don't read Wesley as saying) that there's an inherent or genetic difference. It's also not clear how much, if any, of this effect crosses racial, economic, and other cultural lines. But in order for one gender (or any other group) consistently "not [to pay] all that much attention to what is going on in the story" when another gender (or whatever) does, you pretty much have to say that there are different thought processes happening.

Perhaps, then, these studies could serve a use outside of the world of philosophy. The nonexistence of similar results in younger populations (i.e., ones with still-developing cognitive faculties) might be a reason to look for mitigating social factors, and so on. In the meantime, at the very least the field of experimental philosophy needs to prove that it's been using valid sample sets, and it seems now like that can't be done without addressing this paper one way or another.

[Bryce Huebner]

Wesley,

I haven't had a chance to read closely as I've been busy with other work. But I'm having a hard time interpreting the import of your data. That is, I'm having a hard time seeing how the significant difference that your report yields a meaningful gender-based differences in epistemic intuitions.

There are a few reasons why I am apprehensive:

First, I am on a paper that's out under review right now where we looked at effects of reported gender in moral judgment tasks. We found some here and there; but they weren't systematic and they didn't always go in the same direction. I worry about making any inference whatsoever on the basis of the small amount of data reported in this paper.

Second, I wonder how substantial the effect really is. The difference in the means is .62 on a 7 pt scale, with whopping standard deviations of 2 points on a 7 point scale. I checked the effect size on the basis of your means and SDs and the Cohen's d was .30. That's a 'medium sized' difference, but what it means is that the reported-male and reported-female curves fail to overlap in 20% of the responses for the examined population. I worry that for the vast majority of responses, people who identified as females gave answers that were absolutely indistinguishable from the answers that were provided by people who identified as male.

So, what about the difference? It could be that there was some smallish group of people who identified as female who claimed that the chairman had no knowledge (-3); or, it could be that some smallish group of people who identified as males claimed that the chairman clearly had knowledge (3); or it could be that more people who identified as female than people who identified as male offered a negative rating.

Can you tell us what the distribution of responses looked like? I think that in this sort of case, the reason why the curves don't overlap where they don't overlap is important to the conclusions that you want to draw. So, i guess that I think that it would be nice to see something more than just an ANOVA in this paper.

Just to be clear, I don't mean to come of as hostile to your claims. I'm just worried that the effect might be spurious given the unsystematic gender effects that we found on moral judgments. And I'm worried that the effect that you have found says more about a small proportion of outliers (male? female?) who gave responses that lay outside the bell of the curve on this case. So, any more data you can provide would be greatly appreciated.

Thanks a bunch, and I look forward to reading the paper more closely over the next few days!

[Jennifer Nagel]

Hi Larry Niven,

Although it is true that systematic differences in performance can be taken as evidence of systematic differences in thought processes, finding differences correlated with gender is not the same thing as finding differences that are arising as the result of gender. If the disparities vanish when we control for, say, level of interest or motivation, then it is level of interest or motivation that is driving the dissimilarity rather than gender per se. The idea then would be that women and men perform epistemic evaluations in exactly the same way, assuming that they have equal interest in the story in question. If the factor that matter is interest in the story, then we don't have any grounds for drawing conclusions about differences in concepts of knowledge.

Various factors can produce different performance in different groups. One easy way to do it would be to fail to disguise the hypothesis of experimental interest: women do underperform men when they are told that a test is expected to show gender differences (the moderating variable there is apparently irritation -- irritated people of any gender don't perform as well on tasks requiring attention as their calm counterparts).

I should say that there is very substantial evidence that with the exception of an extremely limited range of abilities (notably physical aggression and certain motor skills like throwing), gender differences are close to zero. Janet Hyde has a good summary of the literature in her 2005 American Psychologist article; she goes through 46 meta-analyses of gender comparisons, including all the major work on areas like spatial reasoning, moral cognition and verbal ability -- areas in which there is widespread opinion that there are large differences between men and women. She finds that in the vast majority of these meta-analyses, differences were small or close to zero.

I like Jaffee & Hyde's 2000 Psychological Bulletin article refuting the Gilligan hypothesis that men think about morality in terms of justice where women think about it in terms of care. It turns out that this is just false. Whew, that's good news for moral theorists of either gender.

[Mark Phelan]

I think the pragmatics worry is a genuine one. We experimentalists need to be more careful about this. If men and women’s read the vignette’s differently, then they may interpret the relevant questions differently. And if they interpret the questions differently, then they’re answering different questions and there’s no reason to suppose that there are gender differences between epistemic intuitions. But it’s pretty cheap just to raise this worry without having plausible pragmatic differences to point to. Since I don’t have any plausible pragmatic differences to point to, I’ll shut up about that.

I think Bryce’s worries are right on. I would also love to see more info about the distributions.

Finally, a response to this quote from a response: “Positing gender-based conceptual variation is I think a relatively expensive move to make, in terms of increasing the burden of explaining the acquisition and transmission of the concept…”

Surely positing gender-based conceptual variation couldn’t be any more expensive than positing any other conceptual variation. In fact, it’s probably cheapened by the fact that there are other well known gender-differences. So the challenge must be that conceptual variation is expensive in general, in terms of increasing the burden of explaining the acquisition and transmission of the concept. But why shouldn’t we suppose that conceptual variations are widespread, especially when it comes to ecologically rare cases? This is particularly reasonable to suppose when it comes to normative concepts, such as “knows”. Think about a concept like “good”. Surely there are lots of differences between the people regard as good--particularly when it comes to rare or fringe cases. Should we suppose that people don’t share a concept of “good”, or that the transmission of the concept is particularly difficult?

[Brandon Watson]

The expense of positing gender-based conceptual variation, I take it, is partly due to the complications that would necessarily be involved in causal explanations that affect gender in particular. Region-based conceptual variation, for instance, is relatively inexpensive: variation by regions suggests variation in the things to which people in those regions are exposed. But gender-based conceptual variation is not so simple; men and women are mixed in together, and therefore there would have to be a causal factor (or set of factors) that (a) can explain such conceptual variation; (b) affects men and women differently; and (c) does so despite the fact that men and women share an immense amount of their environment. When this is added to the fact that Jennifer mentioned, that attempts to find clear gender divergences (many of which were expected) in a wide variety of other areas has, with only a few exceptions, failed, we also get the problem that the causal factor would have to (d) affect these topics but not (apparently) a wide range of other areas, many of which seem to be relevant to concept formation. That's a pretty steep requirement for a causal explanation, one that is missing if in cases where we might suggest that (e.g.) fairly well-defined groups of people who are exposed to obviously different environments exhibit conceptual variation due to differences in location or class or educational exposure. The factors explaining the variation would apparently have to be very precise in tendency yet very pervasive, and are not at all to be expected from analogy to similar things. Because of that, it would be extraordinarily notable (and important, for the reason suggested in the post) if it were true; but also because of it one would worry much more about the possibility that the result was due to something closer to home -- if we rule out things like statistical fluke and experimental bias, then non-obvious differences in expectations in men and women due to things entirely unrelated to conceptualization and only incidentally related to gender at all (to just pull one possibility out of the air, one would want to rule out that the format of the experiment wasn't picking up on variations of answers due to cultural expectations about how men and women should act, so that the variation would be eliminated if the experimental approach were much more indirect).

In any case, I took something like this to be what Jennifer meant by calling it a relatively expensive move to make; she can correct me if I'm wrong.

[Wesley Buckwalter]

To Bryce,

Thank you for your interest in the experiments! To your first point, I am excited to read your paper that you mention. Is it available? Without knowing the specifics of your work, it’s hard to know how to respond, or how the results there question the ones here. The hypothesis I give in the paper is that certain groups are more likely to consult normative evaluations when forming supposed non-moral judgments. Could you say a little more about how not detecting systematic gender differences in the moral judgment tasks you used bears on this?

To your second point regarding effect size, your point is well taken. I will say though that one thing that has always stuck me about the epistemic side-effect effect is that in the original Knobe probes we used (these ones) the putative knower in the stimulus has lots and lots of evidence. In both cases it seems like it should be a very clear case of knowledge. The fact that there was any difference at all is still completely shocking to me. However, many folks have raised the worry you mention about effect size, and I agree that the epistemological import would be more convincing if it were greater. As we speak, more experiments are in the works to bolster the effect…and many of those efforts have succeeded. One strategy is just to give the subject less evidence, and the data cleans up (in the paper, I reference one such experiment). More research will be needed to see if those differences also break along gender lines.

As for the distribution, female responses from -3 to 0 all got about the same number of participants (average 25 each), while men were all around 10 each. Then, in the positive category, most men were lumped in 3, while women were much more evenly spread out. The curves definitely make me want to look a little closer at what is going on in this experiment!

To Larry,

I think that your suggestions about the possible import of these results outside of the philosophical world are really intriguing. That is, whether my hypothesis that explains the data (in terms of genuine variation in epistemic intuitions)…or the one Jennifer is suggesting (that the differences do not arise at all as a result of genuine differences in epistemic intuition) turns out to be correct, we still might be learning something really interesting for lots of different reasons. Yet, for the purposes of this paper, and the arguments I make about the use of intuitions in epistemology, I'm hoping that the differences detected here are really due to differences in epistemic intuitions. That said, I also agree with you that it seems extremely unlikely that students various attention spans would produce the pattern of results from the three experiments I mention. That is an explanatory hypothesis without any (as far as I know) empirical support.

To Jennifer,

I agree that, “various factors can produce different performance in different groups.” I’ve always found the experiment you mention (about underperformance) completely astonishing! Next, you mention more experiments that fail to find gender differences in spatial reasoning, moral cognition and verbal ability—perhaps pace historical opinion to the contrary. Then you end with another example of how experimentation has refuted common opinion about gender differences in moral theory.

But, I guess I’m unconvinced that: 1) the fact that pragmatic effects exist, 2) that data from bunch of other related areas of inquiry fail to show gender variation, or 3) that data overturns common opinion about gender differences in the past, is completely relevant to the question of explanatory hypotheses of the current data. What I mean is, do you think these things count as strong evidence for the current claim that the gender differences that I report in the three specific experiments here are due only to some kind of pragmatics? I completely agree we should be cautious, but I think we should also support the theory that is supported by the best evidence, regardless of how expensive or surprising. And I think we have the very start of some good evidence in the other direction.

[Anne Jaap Jacobson]

I thought that the Hyde paper on spatial reasoning was saying that the differences that have existed are disappearing. There's lot of evidence that women's mathematical abilities have been affected by all sorts of negative input, limited opportunities for training and so on. (I don't know how many men are told that if they use logic in the conversations, they can forget about having friends. I got that news at around 5, when I discover the power of a logical refutation. My parents were literally horrified.)

I think feministphilosophers.wordpress.com has been tracking some of the unfolding evidence, some of it very recent. A good case can also be made that living in a gender-equitable society makes a difference - the US does not count as one on a number of measures. (If you are interested, you could try searching the blog under "mathematics".)

Wide-spread environmental differences seem much less trivial than the possible causes so far mentioned. They raise a question about whether the intuitions in the classroom have a large environmental cause, if not one precisely on grounded in some more basic bio-gender difference.

I do remember vividly being at a conference where a keynote speaker was claiming that whether one had a good life depended on and only on what happened in the middle stage, when one might be general or powerful CEO, for example. Childhood mattered only in terms of its effects on the middle stage. I thought for a while and it seemed odd to me since I, then the mother of a ten year old child, was pretty sure I wasn't just concerned about the effects of his childhood; I also thought a good childhood was part of a good life. And it seemed indeed that all the academic, exhausted women I knew had broader concerns governing their child-rearing than just the effects on whether the kid would be a success as an adult. So I said so. Of course, I was the only woman in the group. And the response from the men?

"O no," they said, "you are being irrational."

[Anne Jaap Jacobson]

Just to note: the Hyde paper I'm referring to is a more recent PNAS paper.

[larryniven]

Hi again, Jennifer!

"If the disparities vanish when we control for, say, level of interest or motivation, then it is level of interest or motivation that is driving the dissimilarity rather than gender per se."

Okay, sure - but that's still a difference in thought processes, no? Adjusting one's level of interest is a thought process, and the same goes for motivation: each is itself a mental algorithm that takes input (a story, or an experimental setup, or whatever) and generates an output (the portion of mental energy to be devoted to the task) and also part of a larger multi-part algorithm that ultimately results in a data point. I'm certainly not trying to say that gender alone or in itself causes or is responsible for any of the observed differences - that would, I think, be incoherent; gender as such is just not a cognitive factor in either of the ways described above - but I do want to accept the data for what it is and use that as the base for further investigation and hypothesizing, the results of which may or may not ultimately point to a gendered correlation.

"If the factor that matter is interest in the story, then we don't have any grounds for drawing conclusions about differences in concepts of knowledge."

This...is a sticky thing to say. If we're using "concept" in a way that includes unconscious tendencies or linkages, then it in fact does say something about the way that people conceptualize knowledge: it says that one group finds (happens to find) certain kinds of knowledge more interesting or motivating. If on the other hand we're limiting "concept" to the explicit and the consciously recognized, then I think I agree with this.

The only caveat in the latter case is that x-phi needs to consider the unconscious as well as the conscious, so the results are relevant in either case - likewise for the inherent differences (if any exist) as well as differences caused by environmental factors. Just to have an example, let's take your case of irritation affecting performance: x-phi, it seems to me, operates under the assumption that calm intuitional reasoning is more valid (or somehow otherwise better) than irritated intuitional reasoning. But has it justified that assumption? I am, in other words, trying to make sure that, to borrow Brandon's phrase, people understand that there are very, very few "things entirely unrelated to conceptualization."

As for your citations, good! Those and other studies provide the kinds of structures within which we should expect (and, in fact, demand) further research to operate. But if all of the experimental conditions are set up in similar fashion - if, I think we're saying, all of the pragmatics are on par in both your citations and in any given x-phi study - then, as Wesley says, we should be ready to accept the results of the latter with the same level of confidence that we accept the former, yes?

And Wesley, yes - I definitely recognize that you're "hoping that the differences detected here are really due to differences in epistemic intuitions"! I, on the other hand, have no such hope, so you'll understand if my interaction with the data differs from yours. ^_^

[Jennifer Nagel]

Hi Wesley,
I absolutely agree that we should endorse the theory that is best supported by the evidence. When I say that your hypothesis that men and women "reason very differently" about Gettier cases is surprising, I mean that when we look at this hypothesis against the background of a very large body of empirical work on mental state ascription and male/female similarities in cognition, it has a very, very low prior probability. This is to say that we will need a lot of evidence to demonstrate that it is right; in responding to a single study which interprets a performance difference as pointing to an underlying cognitive difference, our first inclination should be to doubt that interpretation of the study's data. If extensive further studies showed a robust effect (and that's a very big "if"), of course one should reconsider.

In drawing attention to work like Hyde's, I also wanted to suggest that many people might have mistaken impressions about the prior probability of the gender disparity hypothesis -- the impression that women and men "reason very differently" on a range of problems is unfortunately quite widespread, although not well supported by evidence.

When I say that the Gettier case gender disparity hypothesis is expensive, I mean that the advocate of such a hypothesis would eventually need to give us an account of how such a gender-based cognitive variation comes to be, and as Brandon Watson has very nicely argued, this is not going to be easy.

Larry, it's easier to explain why men and women might have different interests in a society like ours than to explain how men and women might reason differently about Gettier cases when they don't in general reason differently about mental states.

About whether xphi has justified the assumption that calm reasoning beats irritated reasoning: no, x-phi hasn't done this. But a large body of work in the psychology of reasoning has.

[Wesley Buckwalter]

Hey Jennifer, sorry for the delay in responding.

Throughout your comments you’ve been alluding to “a very large body of empirical work on mental state ascription and male/female similarities in cognition” that among other things you say, has proved that “mental state ascription is not something that exhibits gender variation,” and that in turn we ought to “doubt [my] interpretation of the study's data.” Your claim is that such data should assign the hypothesis that there are sex differences in epistemic intuitions a “very, very low prior probability.” While it may not be entirely productive to belabor this point further, let me just say that all of this seems a little fast to me for two reasons.

1) Is it possible that you are exaggerating the “the very large body of empirical work” you keep mentioning? While there is some specific data on mental state ascription, as well as data showing cognitive and biological similarities between men and women…there is also a large amount of data about how they are not similar. On the biological side: the Institute of Medicine’s Committee on Understanding the Biology of Sex and Gender Differences 2001 report highlights the “need to conduct hypothesis-driven biomedical studies that take into account sex as a basic human variable.” The report “validates the need for research on sex-based differences at the molecular, cellular, and whole organism levels and at different stages of the life span.“ On the psych side as well: many psychological gender differences are nonexistent, some are small but measurable, some are moderate, and some are very large.

I agree that some of these studies (the ones showing small differences) are controversial, and several authors have focused on targeting some of those results, Hyde’s gender similarity hypothesis among them. But Hyde’s meta-analysis based conclusion (that psychological gender differences are in the close-to-zero or small range) is also susceptible to ample criticism (See AP v61 2006 for responses). There hardly exists the consensus that you claim. Instead, as a growing number of scientific fields begin to recognize the importance of research on sex differences (and discovering the societal and biological causes of gender similarities and differences), our question becomes which studies out there best support or question the current data in the three experiments I talk about? You mention your favorite studies; I’ll mention mine. But I’m not so sure that the current evidence out there, like what you mention here for instance, is as decisively against the particular explanatory hypothesis I advocate.

2), How well does the research you mention apply? I feel like maybe the dialectic here is a little confused. It seems like we’ve been focusing on just one part of the paper (on whether Gettier sex differences are real or not) and not on the larger argument. In the paper, the explanatory hypothesis I argue for to explain both ESEE and Gettier results is that certain groups are more likely to consult their moral judgments in applying supposedly non-moral concepts like knowledge in situations where moral factors are salient. Then I test that hypothesis by running an additional experiment about causal intuitions, finding some evidence to support it in another domain. The normative evaluation hypothesis seems very different from some of the data you talk about concerning gender similarity in related areas of mental processing. I was hoping that you might say a little bit more about how you think the past studies you talk about in different areas specifically affect the prior probability of the hypothesis about normative evaluation so strongly.

[Jennifer Nagel]

Dear Wesley,

There certainly are interesting differences between women and men -- I didn't mean to come across as denying that, or denying the value of studying possible gender-based differences. My own recollection of the Institute of Medicine's report was that they were concerned that men are more studied than women especially in drug trials, in part because of concerns about drug exposure in pregnancy. But fortunately research in mental state ascription has not suffered from this problem; male and female children and adults have been studied very heavily, and the consensus is that mental state ascription works the same way for both sexes. I should say that the Charman, Ruffman and Clements analysis showed a slight advantage for 3- and 4-year old girls, but that (1) this advantage vanished by age 5, and (2) the authors were unable to rule out the hypothesis that this advantage was in fact a byproduct of girls' slight advantage in linguistic development during that earlier period, as opposed to anything having to do with mental state ascription per se. After age 5, I don't know of any research which shows significant gender differences in mental state ascription for the non-clinical population. In some fields (perhaps certain areas of pharmacology) we don't know much about gender differences because we haven't studied both genders adequately; we can't say that about mental state ascription.

Your guiding hypothesis is that "women are more likely than men to consult normative evaluations like their moral judgments in a given scenario as informing their epistemological, intuitive judgments." I agree that this hypothesis is somewhat remote from mainstream empirical work on the attribution of mental states. I think it would be helpful to have some account of what it is about gender that is supposed to produce this particular difference in epistemological processing for these particular types of cases; without some plausible story about what the mechanism is here, it will seem more reasonable to continue with the well-grounded assumption that epistemological processing does not vary by gender, and search for another explanation of the variation in performance on the particular probes tested.

[Wesley Buckwalter]

Jennifer,

First, I wanted to thank you again for all of your excellent comments! While I will definitely pursue the hypothesis I presented in the paper, the discussion we’ve had has really made me think twice about how such work would need to fit into the rich and growing data set on sex and mental state ascription!

I thought maybe we might redirect the discussion back to a point related to what Larry was brining up earlier. For the purposes of argument, let’s say that a philosophical intuition is just what famous philosophers have in mind when they use the word “intuition.” In the Stich/Williamson sense for instance, intuitions are, generally speaking, just reflective responses to a thought experiment. Now, by these lights, in the experiments I present, participants do have different intuitions in response to Gettier, perhaps ESEE, and certain causal cases. We’ve been blogging now back and forth about the best explanatory hypothesis to shed light on these different intuitions. They might arise from the factor I mention, but it’s *entirely* possible that any number of other factors could be at work. Even suppose that past evidence up to now decisively suggests the latter. As you note, this would spell lots of trouble for NEH.

But what I was thinking was that the underlying cause of why—for instance, males seemed to have a Gettier intuition while women seemed not to (even if the cause is due to a number of pragmatic things)—does not bear on the claims made here that 1) intuitions so defined are bad sources of evidence in certain philosophical projects, and 2) that such differences have serious implications for the sociology of philosophy.

That is, could members of groups that tend to have different intuitions from those in positions of philosophical authority about major philosophical thought experiments like Gettier get bad grades, no matter their cause? So suppose we bracket off explanatory hypotheses of the data for the moment, and ask how such differences in intuition, pragmatic or otherwise, bear on the use of intuitional arguments in epistemology and on gender underrepresentation on the whole.

UPDATE: Just a quick correction about the grades comment above. I would never want to make any kind of claim about who gets what kind of grades in philosophy courses. Conversely, the point I wanted to make was that it is entirely possible that women could find themselves alienated in *any number of ways* for espousing different intuitions than those of the male majority, no matter what any of the cause(s) of those differences might be.

[Jennifer Nagel]

Hi Wesley,

If the answer to the problem of gender imbalance in philosophy is to give up on using intuitions in arguments, I worry that philosophy will get a whole lot less interesting. What exactly will we put in place of intuition? It would be pedagogically pretty rough to get restricted to scholastic demonstrations from first principles, especially if we are barred from hinting that our principles are evident by the light of nature. Other than pure conditional reasoning for its own sake, or straight history of ideas, I'm not sure what other alternatives would be out there, and if you can think of some good ones, do let me know. For my own part, I find that whatever I do in philosophy at some point I rely on some form of intuitive response -- even when I'm teaching Plato. OK, especially when I'm teaching Plato.

I do think that the very same intuitive response can be triggered more or less successfully in different groups simply by the use of materials that are more or less engaging/irritating to the members of these groups. The very same point about knowledge ascriptions and deviant causal chains could be made with a story about a sporting event, a burglary or an extra-marital affair. For philosophers who already have a strong interest in knowledge ascription, it may not matter much which stimuli are used. (Although I have to say that every time David Lewis uses cricket as a source of intuitive fodder, I have a "what just happened there?" moment, and have to go back later after I see what he says about the intuition to see if I can figure out the content of his example.)

Undergraduates who don't yet see what is at stake may be less inclined to engage in the project of making sense of the facts of a convoluted example if it involves material that they aren't interested in, or material that in some way annoys them (the latter is particularly dangerous because it ties up working memory, which one really needs to keep a deviant causal chain straight in one's head). It takes considerable work just to construe the facts of a case accurately.

X-phi subjects who are filling out a quick survey at a table in a hallway for a trivial reward (or no reward at all) may have even less incentive to think about the case. The problem isn't that they are reading the cases the same way as everyone else and then getting deviant responses: the problem is that they aren't really reading the cases (or maybe only some of them are, because they find the story interesting). This is the old Petty & Cacioppo point that we all are inclined to evaluate arguments either superficially (if we don't care about an issue) or more deeply (if we do). When we don't care much we will be swayed more by factors like font clarity; when we do, it will matter much more whether an argument is cogent (and we will do a much better job of representing and remembering what the argument actually said). At least undergraduates in a class may have a bit more at stake than typical x-phi subjects, who are not performing under conditions of accountability (Angel Pinillos has some nice work that is relevant here). In any event, it's not obvious to me that the data that show variation in performance are actually showing variation in intuition in, say, Williamson's sense of the term.

The Sytsma & Livengood paper recently posted to this blog is also a very useful reminder that people may answer a question differently because they construed the question differently. In the context of a philosophy class, it is possible to disambiguate various readings of a question that might be confounded if the question is read in isolation.

Meanwhile, I've found it helpful to use i-clickers for feedback in large classes -- I can poll the class anonymously to see how many people are getting a given intuition, and then present the same point using a variety of stimuli until I get (almost) everyone on board. But even before using these devices I had known that some cases work better than others (and I'm sure anyone who teaches epistemology has had comparable experiences). Lehrer's Havit/Nogot example works better than Gettier's original 1963 Ford example, because the latter relies on disjunction introduction/addition, which isn't an intuitively acceptable inference pattern for most people, given the pragmatics of asserting disjunctions (Gettier's version is fine for focused students who have the relevant logical training). Plus those names help. But I should say -- while the original Lehrer story was up on screen, 5% of my students selected "Nogot" in answer to my comprehension screen question "who actually owns a Ford?". Keith DeRose suggested to me that maybe "Nogot" isn't obvious enough, and we should change the name to "Doesn't-Own-A-Ford". Sigh. In any event, if some of the variation in responses to examples is coming from our students not feeling motivated to read the cases closely, we can help them by keeping our cases as unconfounded as possible, and giving them a variety of content to respond to. E.g. in a transparently pandering attempt to keep male students on board I use basketball and hockey examples sometimes, despite knowing so little about the leagues that I always have to check some sports website before class to make sure my examples won't be absurd. I also use character names associated with a variety of different ethnicities (being blessed with a very multi-culti student body). Women have never been a problem for me.

As to your suggestion that "groups that tend to have different intuitions from those in positions of philosophical authority .. are probably going to get bad grades" -- I think this is either false or it is evidence that intuitions don't differ by gender: in my experience, there isn't a significant difference between male and female grades. But that would be anecdotal evidence from a female prof, which really wouldn't be a good test of your hypothesis. I wonder if anyone has good data on this point. I know a lot of female philosophy undergrads with good grades end up in law school rather than philosophy grad school. But I don't think variation in intuitions is the driving factor there.

[Wesley Buckwalter]

Hey Jennifer,

So, the suggestion here would be that differences in pretheoretic intuitions between men and women might somehow contribute to gender imbalance in philosophy…not just because in these studies men seem to have different intuitions then women…but largely because of how the disproportionally male discipline might respond to individuals who espouse them. If that is right (just one hypothesis), I wouldn’t suggest that the answer to the problem of gender imbalance is to give up on using intuitions in arguments…but rather to give up on the idea that the contents of the intuitions men tend to have are any more likely to be true or false than the ones women tend to have. Indeed, I would think that research showing demographic variation in philosophical intuition in judgments about novel cases could have something of the opposite effect…whereby more data about differences guides our awareness, tolerance and openness to viewpoints of different cultures or groups.

That said, the fact that certain intuitions are not the unanimous ones philosophers often assume in certain cases—I’ve argued—does belie the efficacy of using those intuitions as good evidence in a *very specific* sort of philosophical argument with *very specific* philosophical goals. But importantly, this is not to say that intuitions are always bad evidence, or that intuitions can never be used in philosophical arguments.

I have to admit, I’m still a little mystified as to the argument for why the differences detected here really aren’t differences in “intuition” in the wide, Williamson sense. First, I was wondering, how plausible is it to think that the problem of participants not reading stimulus could somehow still result in the particular *systematic* differences between men and women in these experiments? Clearly there is some factor that is responsible for different groups answering in this very particular yet different ways across these different IVs…and across these different domains,epistemology and causation.

Second, by intuition here, we are just talking about general judgments in response to thought experiments…whatever their cause (be the potential cause due to levels of interest in the Ford, or different views about what constitutes knowledge, and everything in between). Perhaps this is a reason why Williamson thinks some intuitions (those from experts) are sometimes better than others to use when making arguments in epistemology let’s say, but that is another matter. One might want to argue though that even if level of interest in famous philosophical thought experiments is the relevant factor generating differences in intuition here, that really doesn't seem to affect the hypothesis that groups with underrepresented intuitions (in this wide sense) in response to a particular case could feel alienated in any number of ways.

Third, I would argue that at least in one sense, xphi subjects are actually better than students in your classroom as experimental participants in this regard. Why? Because in xphi tests (these tests were mostly done in the non-philosophy classroom) we don’t try to get anyone “on board” with anything. We just take about 10 minutes to collect snap judgments in response to cases, not tainted by anything a philosophy professor says they are, or are not, supposed to say in epistemology class.