How not to lose friends and alienate people when talking about women in philosophy

paper with a lot of interesting data about the representation of women in philosophy journals has just been published in Philosophical Studies. The authors used data from 25 top philosophy journals and found that, compared to their share of professional philosophers, women are underrepresented among the authors who published in these journals. Unfortunately, the data that are analyzed in that paper are clearly insufficient to draw many conclusions from this fact, because they don’t include any information on the rate of acceptance by gender. Without data about this, even if women are underrepresented in philosophy journals, it’s impossible to know why. Still, the data analyzed in that paper are somewhat at odds with the official narrative, which is that women are discriminated against in philosophy. Indeed, while the analysis shows that, compared to their share of professional philosophers, women are underrepresented among the authors who published in those 25 journals, this is only true for journals that use a double-blind or triple-blind anonymous review process. If you just look at journals whose review process is not anonymous, you see that women are overrepresented.

However, it’s not clear whether the difference between the proportion of women among the authors who published in journals whose review process is not anonymous and the share of women among professional philosophers is statistically significant, but the authors don’t always make it easy to figure out. In another part of the paper, they disaggregate the data and look at that difference for each journal separately.

As you can see, in the case of the Proceedings of the Aristotelian Society and Mind and Language, the only journals the authors looked at whose review process is not anonymous, women are overrepresented but the difference is not statistically significant.

However, this is less clear when you look at these journals in combination, as you can see in figure 7 of the paper above. Even if you try to tell by eyeballing the error bars in that figure, you can’t really figure out if the difference is statistically significant, because it’s not clear what underlying distribution was assumed to compute the standard error. Moreover, the results presented in figure 7 are for 2015 alone, whereas in table 3 the analysis is based on the articles published in both 2014 and 2015. I think it makes more sense to combine the data from both 2014 and 2015, as the authors did when they analyzed journals separately, if only because the test would have more statistical power. So I don’t know why they didn’t do the same thing when they analyzed the data by sorting journals into groups according to the kind of review process they use. It may be that it doesn’t make any difference, but I would still be curious to know. Even if the difference comes out as statistically significant when you do that, however, I think a reasonable argument can be made that we should focus on the results for each journal separately and, as we have seen, the difference is not significant for either of the 2 journals analyzed in the paper that do not practice anonymous review. In any case, it seems that, at the very least, women are not underrepresented among the authors who published in journals that don’t practice anonymous review (although we should be careful not to generalize since we only have data about 2 of them), which contrasts with the situation in journals that do.

This doesn’t really square with the official narrative, so the authors of the paper try to explain away these recalcitrant data, but the result is, shall we say, less than convincing:

Finally, it is striking that journals which practice non-anonymous review tend to publish more women than journals which do not. This result is surprising because it seems inconsistent with some literature on anonymous review (Budden et al. 2008). There are several possible reasons why women do worse in journals that review anonymously. First, the majority of the reviewers at these journals are probably men (Hassoun and Conklin 2015). It is possible that men and women have different opinions on what does and does not constitute important or interesting philosophy (Dotson 2012). Second, biases against women may be affecting the process of anonymous review, despite the intentions of all involved (Lee and Schunn 2010; De Cruz 2014a). Even when journals try to keep author identities hidden, cues like the article’s written style, its theme, and its similarity to conference presentations of which the reviewers may be aware could still reveal the identities of the authors (Rosenblatt and Kirk 1981). Moreover, there are some reports that reviewers seek out the identity of the author through the use of Google (Brogaard 2012). It is also worth noting that some journals have built-in procedures to partially circumvent their otherwise anonymous review processes (Lee and Schunn 2010). In the final step of the review process at Ethics, for example, the editors hold a vote to decide whether an article should be published or not, at which point two of the editors are aware of the identity of the author. As one Editorial Board member explained, associate editors can also invite submissions, thereby circumventing an initial screening by the editor, and guaranteeing that the manuscript will be sent to external reviewers (Author’s personal correspondence: withheld for anonymity). Practices like these, which may occur in other journals, could make it harder for women to publish. In general, such practices may help well-connected authors—who are overwhelmingly men—to subvert the standard double or triple anonymous review process. Third, it may be easier for journals that do not anonymously review to publish more women (Hassoun and Conklin 2015). An editor or reviewer who realizes that their journal is not publishing many women can fix that, if the review process is not anonymous, by giving well-reviewed women authors a chance to publish. It is probably harder for journals that practice anonymous review to remedy the gender disparity.

In other words, if women are underrepresented in journals that practice anonymous review, it might be because people are able to subvert the review process and guess the gender of the author despite anonymity. But the data analyzed in the article suggest that, when reviewers/editors know the gender of the author, women are not underrepresented and may even be overrepresented… So this doesn’t make any sense and, if you ask me, it just illustrates the absurd lengths to which people are willing to go in order to save the official narrative about women in philosophy. (Of course, it’s not the first time that I have observed that phenomenon, as you can see if, for instance, you read the comments of this post on Daily Nous from a while ago.) If you want to save that narrative, a much better option is to say that, since the paper only analyzed data from 2 journals that don’t practice anonymous review, these results could just be a fluke, which is true. To be fair, the authors do say that after the passage I just quoted, but it still boggles the mind that apparently they had no problem trying to explain away this embarrassing result by making a point which is clearly at odds with their own data.

9 thoughts

  1. Thanks for posting this.

    Like you, I’m a little surprised that the authors would draw this conclusion. As far as I can see (I hope someone will correct me if I’m wrong), the data can be explained perfectly by what I take to be two uncontroversial facts:

    1) As Caroline Dicey Jennings’ team concluded from their extensive survey, as I recall, it turned out that women hired to tenure-track positions, and awarded tenure, tended to have significantly fewer publications, and fewer publications in top journals, than their male counterparts, on average.

    2) There is, I have noted, an often-reported preference among some members of the profession to give preference to publications by female philosophers, particularly if doing this would be likely to make the resulting journal, anthology, etc. more sex-balanced.

    Unless these claims are in dispute — and I don’t think they are — then I’m not sure why a further hypothesis is needed to explain the data that are presented here. Given the two factors mentioned above, one would _expect_ that the male philosophers who have attained or are likely to attain tenure-track or tenured positions would be in the habit of producing and submitting publishable articles at a higher rate than their female counterparts (for whatever reason best explains the Dicey-Jennings findings as I recall them), so, under successful blind review, one would expect to find proportionally higher accepted publications by male philosophers. However, given the second factor, one would expect this tendency to be significantly diminished if the editors and reviewers knew the sex of the writer of the paper. And this is exactly what these data show.

    To make the case that this also involves some bias against women philosophers, other data that, as far as I can see, were not provided here would be needed. For instance, it would be interesting if female philosophers’ submissions were more likely to be rejected than submissions from male philosophers, particularly if those submissions were found to be indistinguishable in quality by reviewers operating under more strictly controlled conditions of anonymity (e.g. being unable to use Google to discern the likely authors of the articles). But in the absence of these or any other findings to support the hypothesis that this discrepancy is the result of bias against submissions by women, doesn’t it make more sense to provisionally accept the simplest hypothesis that saves the phenomena?

    1. Yes, this is basically my thinking, but it’s hard to have a reasonable conversation about that. As I note at the beginning of my post, since Wilhelm et al. didn’t have data about the rate of acceptance by gender, the data they analyze in that paper doesn’t have much bearing on whether there is any bias when taken in isolation. From a purely logical point of view, there could be a bias in favor of women even if they are underrepresented, just as they could be discriminated against even if they are overrepresented. But given the other data we already had, which you bring up in your comment, I think your hypothesis should be the default explanation. Of course, evidence could be uncovered that make that explanation less plausible, but in the absence of any evidence that directly bears on that, it’s really weird for the authors to go out of their way to try to undermine that explanation with a hypothesis for which there is no evidence whatsoever and that is at least prima facie at odds with their own data. It’s truly amazing that, even though everyone is constantly talking about how we must increase the proportion of women in philosophy, everyone loses their shit as soon as someone suggests that it’s exactly what people are trying to do, even when it’s the most natural explanation of the data we have…

  2. This is completely off-topic, but since you’re the only French person I’m in contact with, I want to congratulate you on your president’s handshake with Trump.

    Maybe if Macron works out with weights for the next few months, the next time he meets Trump and shakes hands with him, he can send the bastard to the emergency room.

    During the election, without knowing much about French politics, I supported Melenchon, but now I begin to see Macron’s merits.

    1. Thanks, I remember Sesardic and De Clercq’s paper, which I thought was very good. If you’re interested in that issue, I just posted a very detailed post about it, where I discuss the underrepresentation of women more at length.

      1. Yeah, I’ve just read your latest post. Excellent piece as well! I’ve been bothered by the reasoning you criticize for some time now. Wherever there’s underrepresentation of women in a (perceived) high status position, you can easily and accurately predict that what you call the “official narrative” will pop up to explain that difference — even when there’s independent evidence that *undermines* the standard discrimination hypothesis. So I’m often tempted to think that people who embrace this discrimination narrative are already presupposing the truth of a very controversial general theory (namely, that we live in some a society where sexism/discrimination against women is pervasive and happens in a lot, if not most (or all!), of interpersonal interactions between the sexes), or at least assigning an unreasonably high prior to its truth. If this is the case, then it becomes less surprising to see these people accusing you of trying to justify sexism when you point out that there are alternative hypotheses (such as, e.g., the hypothesis that men and women naturally tend to have different professional preferences) that should be seriously and honestly considered. Given that most political activists tend to be irrational and engage in a lot of ideologically motivated reasoning, this reaction coming from them should not be so surprising indeed. But seeing this kind of thing coming from professional philosophers, whose alleged job is to think rigorously (and to be empirically oriented), is really depressing to me, someone who isn’t even as statistically literate as a lot of philosophers who uncritically buy into this standard narrative (and while we’re at, I’d like to ask if you could recommend me a good introductory book on statistical inference).

        1. When I decided to learn statistics, I first read Paul Hoel’s textbook, which is a pretty standard introduction. It’s a graduate-level textbook and is a little bit dated, so I’m sure there are better options to start. If you don’t mind reading something in French, I can recommend Michel Lejeune’s textbook, which is excellent and more up to date. If you can’t read French, I’m sure there are other up to date textbooks in English, but none of them comes to mind. I recently bought this book, which looks really cool even though I haven’t read it yet, but it’s probably not a good idea to start with this. One thing I strongly recommend are the little green books on statistical methods published by SAGE. Each of them is about 100 pages and can be read in a couple of hours. They are great to get a grasp of various statistical methods quickly. For instance, there is a very good one about regression analysis, a method which is ubiquitous in social science. I have read several of them and I have been consistently impressed by how clear they are. If you don’t care about theory much but want to be able to understand social scientific articles, you can probably just read a few of them as needed, though I think it’s always useful to have a theoretical foundation.

          1. Thanks a lot! These suggestions are already very useful and probably will be good enough for a start. I have also found out about this little book called “Causal Inference in Statistics” by Judea Pearl et al. that looks like a good introductory approach to causation in statistics.

Leave a Reply