Sociology of Gender Bias in Science

Sociology of Gender Bias in Science
Sociology of Gender Bias in Science. Photos: Texas A&M University-Commerce, CC 2.0

A new study by Dr Corinne Moss-Racusin and colleagues has analysed the public’s comments in response to a prominent study on gender bias in Science, Technology, Engineering and Mathematics (STEM). The researchers find that men are more likely to post negative comments in response to scientific findings about sexism in STEM careers. To provide a flipside illustration, I share some examples of what it is like to be a woman moderator of a large, international science community on Google+. This case study will illustrate the recurring arguments used to invalidate the science on inequality in STEM. These arguments are focused on biological (mis)understandings of gender; stereotypes of what motivates men and women; and a desire to police the boundaries of science. Denying that sexism exists is a common tactic to invalidating the science on gender bias in science, and attacking the social sciences is concurrently used to discredit findings on inequality, as well as support the idea that inequality does not exist in STEM.

Gender Bias in STEM

In a study published in Psychology of Women Quarterly, Moss-Racusin and colleagues analysed 831 comments in the New York Times website, Discover Magazine blog, and the IFLS website, all of which reported on a highly cited and reputable study on the gender bias in faculty hiring committees. The original study was published in PNAS in 2012, and was also authored by Dr Moss-Racusin leading an interdisciplinary team of researchers. The PNAS study was chosen because it was the one science story on gender that gained the most online comments in 2012. The three websites were chosen for analysis because they have a high public following and because all three included a link to the original study, which the public could read freely should they have further questions on the findings. The PNAS study involved an experiment where biology, chemistry and physics professors at research-intensive universities evaluated job applications for a lab manager position. They were presented with identical CVs with either a man or a woman’s name. The participants rated the male applicants more favourably along every measure and offered them a higher starting salary, even though the women’s CVs had the exact same information.

Science Inequality in Comments
Photo: Texas A&M University-Commerce, CC 2.0, via Flickr

In the present study, the researchers find that 433 comments were negative. That is, the commenters refuse to believe the findings on gender bias. Men are more likely to do so. There are eight types of negative responses, that draw on a range of justifications from biology to social conventions. Men especially used subjective and sexist observations about women’s supposed innate inability to succeed in science. Ideas that prevailed included “women get pregnant and leave their jobs!” Women’s interests are also cast as fundamentally different to men’s:

I think that a proportion of the gender divide can be accounted for by a division in interests.

Men also evoked ideas of personal choice, such as arguing men are “hungrier” for success and work harder than women, or that men work harder in STEM because women push them to make more money. (Men would not be sexually competitive, they argue, if they didn’t succeed.) I note that in all these discussions, “men” and “women” are discussed as two discrete groups, implicitly drawing on biological narratives of gender. Transgender and alternative genders don’t feature; racial and other socio-cultural differences are not considered; and heterosexism prevails. That is, the discussion is centred on the presumption that everyone is heterosexual, and that all women and men want the same things. In the study, men are more likely to deny that inequality exists, or conversely they blamed women for inequality and said that gender bias targets men (“I’ve experienced it in the opposite way so far.”). These men evoke a general discontent with affirmative action laws, or they raised other unrelated social issues as examples of “bias against men,” such as divorce laws and custody of children, and saying women get “the most stuff.”

Finally, some commenters dismissed the findings on the basis of its social science methodology, but without actually critiquing the scientific data or methods in any specific way. For example, “This report is JUNK science. There is no data here,”and “This is PC BS propaganda.” I will return to this further below.
In summary:
  • The majority of sexist comments against women are made by men (95%);
  • Between 79% to 88% of comments that justified bias using biology or social arguments, and those who blame women, are men; and
  • Men predominantly mentioned that sexism targets men and they overwhelmingly critiqued social science (68% to 85% of comments).

Gender Differences

Photo: Laurent Neyssensas CC 2.0, via Flickr
Photo: Laurent Neyssensas CC 2.0, via Flickr

Men are more likely to refute the science findings on inequality by stating that they work in STEM (75% of men’s comments). In comparison, women shared personal and detailed stories about the gender bias they’d experienced at work, but only 25% justified their opinions by saying they work in STEM. Women shared stories like:

Speaking as a female computer geek, who seems to be unemployed twice as often as my male counterparts – YES. Gender bias definitely still exists. My instructor told me he generally believes women are bad at math but they’re great if they don’t catch you staring at their butt! Whatta jerk!’’

The gender difference here is that men use blanket statements about biology and innate differences, as well as using personal opinion (society is biased against men) to refute scientific evidence about gender bias. Conversely women use personal anecdotes to illustrate the scientific findings. The first strategy – to deny the science on inequality – is used largely by men to invalidate science on sexism in support of the status quo. The other strategy, used mostly by women, supports the science using personal experiences of bias to challenge the status quo. Neither approach is scientific as personal anecdotes are not science, but the first approach rejects science evidence, saying things are fine the way they are, while the other approach embraces the science, to say things are unfair and should change.

Social Science of Inequality

Photo: Ben Seese, CC 2.0, via Flickr
Photo: Ben Seese, CC 2.0, via Flickr

One strategy that men used to invalidate the PNAS study was to establish themselves as a science expert, by saying they work in STEM. Men also critiqued social science but did so on moral grounds and using emotive language. Social science is often categorically excluded from the umbrella term of STEM. Social scientists rarely use this phrase to describe their practice. While sociology was set up by our early founders to mirror the practices of the natural sciences (for example Durkheim), more recent traditions are expressly critical of the natural sciences for contributing to the marginalisation of women, minorities and vulnerable populations (Foucault is a key critic). Nevertheless, social science is very much a scientific practice – we offer valid methodologies for the critical study of society. We collect data and use established theories to draw conclusions about the social influences on behaviour. Unlike the men who refute the science on gender bias in science, we do not use emotional arguments to dismiss scientific studies. We draw on our training and credible peer reviewed science studies. I run several science communities, and no single issue (other than climate change) draws more heated debate than posts about social science studies on science inequality. I present examples of posts that I’ve authored, all of which draw on social science research, as well as a couple of other examples from other social scientists and non-social scientist women who write about inequality. The common denominator is that whenever inequality in science is raised as an issue, this is immediately met by cries of bias, almost exclusively by men. When the authors are women scientists, we encounter even more push back.

Case Study: Science on Google+

Science on Google+ is an ever-growing community with over 503,300 members, most of whom are general members of the public with an interest in science. We also have host a Google+ page with over 521,200 followers, many of whom are scientists. Even with a slightly higher following, our page rarely descends into personal attacks, possibly because our followers are predominantly science practitioners who use Google+. Our community generates many excellent discussions as we have practising scientists who share their posts, but posts about gender almost always become unruly.

Royal Society

Consider a post where I discussed the Royal Society’s own data that showed gender inequality in their science fellowship program. When I shared this post on Science on Google+ it did not take long for a man to angrily cry that I was asking for special treatment for women simply by writing about women’s experiences of inequality.

New York Times

Via STEM Women
Via STEM Women

In another post which I authored for STEM Women, social science is also called into question. I had presented a scientific critique of a study published in the New York Times. The study argues sexism is dead in academia. I showed that the study’s methodology was flawed. In the Google+ discussion, a man argues that my analysis (of a social science study) is biased because of my sources (also social science studies). Another male mathematician posted the original NYT article and used it to attack psychology as a science, but he also offers his personal experience saying that the authors (both psychologists) are correct in their assessment that there is no sexism in academia. In these examples, we see how social science is malleable to the public and non-social scientists alike, who either attack the study based on its argument (that gender bias exists) or discipline. Social science findings are welcome when they match someone’s world-view that inequality does not exist.

Nobel Prize

Image: STEM Women
Image: STEM Women

In a post I wrote for STEM Women, I discuss data published by the Nobel Prize committee, showing that less than 3% of laureates have been women. Again, men (and some women) cry foul. One man does this by calling into question my scientific credentials, even though he is not a scientist. He wants to argue that women may simply not be good enough to win Nobel Prizes but he has no data to back this up.

True scientists would not discard the politically-incorrect possibility of intellectual differences out of hand. (My emphasis)

I have presented scientific data. He has presented a “possibility” that the science is wrong and argued it with great emotion (see the thread). He persists in arguing that I and my fellow women science moderators (who are biologists) are simply being “politically correct.” He then goes on to copy paste a series of statements from Wikipedia. As it turns out, he is inadvertently referencing a series of social science studies. He tries to dismiss the social science studies he doesn’t like, by using other social science studies he thinks support his argument that women are inferior to men. The problem is he hasn’t actually read these studies and he is not trained to read them critically. I am. So are my fellow women scientists. As we show, not only has he cherry picked his examples, but the studies actually validate my original argument about inequality in science. Other men show up and give the predictable examples of “My wife has a PhD…” Personal examples are used to try to discredit the science.

Confirmation Bias

Another post by diversity specialist and lecturer, M. Laura Moazedi, on the science of confirmation bias (how stereotypes are used to justify outcomes by men and women) leads a man to argue she is being biased.

What about confirmation bias of “scientists” searching at all costs the gender inequality in stereotypes, while ignoring biology?

Note the quote marks, which are used to discredit social science. The fact that the study is a piece of social science research rather than biological science is also used as a rationale to discredit the findings (even though it is a study of social behaviour) . There are many more examples I can give from our community. (I will do a follow up post on how I manage these types of arguments.) The point I want to illustrate here is that when women speak up about science inequality, the science is dismissed. The responses are gendered in other ways, however, as our male colleagues generally face less push back. Nevertheless, resistance still rears its head when a woman scientist speaks up.

Men & Women Moderators

Lewis's Law: “Comments on any article about feminism justify feminism.” – Helen Lewis, journalist
Lewis’s Law: “Comments on any article about feminism justify feminism.” – Helen Lewis, journalist

Take this post by a male moderator, who speaks up about the level of sexism in our community. Most of the comments are positive until a woman moderator, speaks up to confirm her experience. A man promptly begins to argue against another group we co-moderate, STEM Women, saying that it reflects how sexism affects men more than women.

the fact that +STEM Women on G+ exists shows women have sexism problems against men, you are asking for the opposite.

He then argues women are biologically inferior and unable to join the military (in his eyes). The thread rages like this for a couple of days. Compare these two posts about Dr Maryam Mirzakhani’s Fields Medal win. The Fields Medal is colloquially referred to as the “Nobel Prize for Mathematics.” Dr Mirzakhani is the first woman in the award’s history to be recognised. One community member simply posts the news of the win. The second comment is by a man commenting on Dr Mirzakhani’s looks. Another male community member calls out the commenter, asking if he’d make such a comment about a man. I step in as moderator and remind the original commenter about our guidelines that expressly ban sexist comments. A different man jumps in saying:

Zuleyka Zevallos your comments offend me. There was nothing derogatory in Vincents comments, neither was it sexist. You have taken it upon yourself to portray yourself as the almighty of the science discussion community group by suggesting that his free speech was not allowed. This may have been handled better by yourself had you possibly been complimented about your looks, i dont hold out for that compliment or the stepping down from your high horse. #neigghhhh (My emphasis)

I am chastised for enforcing our community rules, in my role as moderator, and I receive a sexist comment to boot. This discussion goes on for two days with my fellow women moderators jumping in. The thread only quietens down when a male moderato repeats our rules against sexism. In a second post about the Fields Medal win, this time written by me on behalf of our moderator team, it takes only 15 minutes for a sexist comment to appear. In almost all of these cases, it is men who deny the science or who make sexist comments. These are, by and large, White men who are heavily invested in protecting the boundaries of science to remain the exclusive domain of (White) men. Writing about inequality in science is critiqued for being “biased against men,” and social science is dismissed for not being biology, and when this tactic fails, other social science is evoked to (erroneously) discredit the original post.

“Focus on the Science”

On her personal Google+ feed, astronomer Dr Katie Mack noted that her large public following loudly revolts when she publishes on issues about equality in science. She notes that her followers shout to have her “focus on the science.” She argues this makes little sense since science is practised by human beings; therefore scientific practices impact on scientists.

So, no, I won’t just “focus on the science” at the expense of actual human scientists. I will keep talking about the ways we can make scientific culture better and more welcoming to anyone who has a contribution to make.

The idea that women should not talk about inequality in science dominates public discussions of science. This happens in science communities such as ours, which expressly state that sexism is grounds for being banned. Women who speak up about inequality are accused of bias or they are otherwise targeted for personal attacks. This includes the women moderators, who are practising scientists with PhDs and a strong knowledge of the science on gender bias. It also happens to women scientists writing about inequality on their personal social media profiles. A rowdy sub-group, largely men, want to read about science and talk about how much they love science without hearing about what it means to practice science. They want to follow women experts but they demand that these women not discuss issues of inequality as a scientific concern. They want to be members of science communities without having to see posts on inequality, even when the community expressly supports such posts. In all these cases, simply ignoring posts about inequality is not enough. Men feel a need to vehemntly disagree with science on inequality even when they have no data and when they do not understand the science. Why? As I’ve previously shown, the sociology of science shows that people are more likely to speak up against science issues in which they have an ideological vested interest. The science about inequality in science is polarising because it is tied to personal identity and deeply-held values. Some men want to imagine science as being uncontaminated by women. If inequality does exist, surely women are inviting it, by virtue of their biology, by their choices, by their mere existence. Women should just shut up and do science. Their science should be seen, but their experiences not heard. Above all, though, these women cannot be allowed to write about the science of inequality in science because this is an encroachment on White men. It goes against nature that women don’t simply accept inequality. It’s unscientific to want to address inequality using science. It’s biased for women to talk about gender bias. It’s censorship to remind people not to objectify women scientists and to stick to community rules when talking about science. Or so the logic of sexism goes.

Moving Forward

Photo: Argonne National Laboratories, CC 2.0, via Flickr
Photo: Argonne National Laboratories, CC 2.0, via Flickr

The research by Moss-Racusin’s team presents a framework for thinking about why men react negatively to the science of gender bias in STEM. Being able to educate the public and STEM professionals to recognise personal gender bias is the first important step in making STEM a more equitable space. Moss-Racusin and her colleagues argue:

“Simply put, women are likely to perceive potential personal gain from research that may ameliorate men’s privilege, whereas men may believe that such research can only harm them. More broadly, because people are often more receptive to information that confirms their existing worldviews… it may be critical to understand participants’ pre-existing attitudes toward gender bias and diversity when creating effective interventions.”

In better understanding the types of arguments used to sustain gender inequality, educators, managers and policy makers can begin to target attitudes that undermine gender inclusion in STEM. I have shared some examples of the issues I encounter as a science moderator with the aim of further illustrating the flipside of Moss-Racusin and colleagues’ findings – what it means to be a woman dealing with these comments. I have more to say on this issue and will return to it again. One take away message for now is that, despite the problems, it is worth speaking up on these issues. Moss-Racusin and colleagues’ study shows that comments on science sites are largely positive. Despite the negative experiences, continuing to discuss the social science on gender bias in science is important. First, because without social science data, the public would bicker about inequality using solely subjective examples based on dangerous stereotypes that undervalue women’s contributions. Second, while there is plenty of evidence that gender inequality in STEM is not based in biology, we still need to keep elevating this science precisely because the message is still not getting through. A very vocal sub-group of people, most of whom are men, want to see women stay quiet on this issue. They see bias in social science but they see no bias in themselves. Why they feel a need to police the boundaries of science is central to moving forward. By crying out that scientists should just “focus on the science,” they are actually calling for the maintenance of White men’s dominance in science. Lack of diversity in STEM impedes innovation. So, in fact, while these men disguise their bias as concern for science, they are in fact saying that they don’t want science to answer new questions. In a nutshell, by protecting men’s perceived dominance in science, they are, in fact, advocating for scientific stagnancy. How much can one really love science if they are heavily invested in gate keeping scientific inclusion, thereby ensuring that new ideas will fail to flourish?

Learn More

To read more about how social science can be used to debunk gender myths in science, see my post, “Science Inequality in the News: Avoiding Dangerous Gender Narratives in STEM,” in Minority Postdoc.

Science Inequality in the News- Avoiding Dangerous Gender Narratives in STEM


Top image: photos 1 and 2 via Flickr.

Connect With Me

Follow me @OtherSociology or click below!

Other Sociologist Twitter

Other Sociologist Facebook

Other Sociologist Google+

Other Sociologist Instagram

Other Sociologist Tumblr

Other Sociologist Pinterest

44 thoughts on “Sociology of Gender Bias in Science

  1. Excellent observation, and unfortunately this is correct Adam Liss. People who disbelieve the science on sexism look for evidence that confirms their beliefs to the exclusion of evidence that contradicts their position (“confirmation bias” for others reading this thread), but science evidence on inequality paradoxically makes their belief stronger!


  2. Thanks for sharing this, Zuleyka Zevallos !

    As depressing as it is… it’s unfortunately perhaps not surprising that men are the ones denying gender bias. After all, they’re the ones who are privileged enough to not notice it unless it is pointed out to them.


  3. But if I’m reading your analysis correctly, men and women deny sexism for different reasons, and the evidence to the contrary makes each of them believe those different reasons even more strongly.

    This is the perfect (and by “perfect,” I mean “totally not perfect”) end to a day of interacting with anti-vaxxers and deniers of evolution. If you can’t trust scientists to follow the scientific method, well, who’s left?

    I’m going to relax by convincing myself this study is wrong, and then I’m going to bed.  🙂


  4. Thanks for your comment Jonah Miller; agree that the findings are not news to many of us who do public outreach. On the one hand, this is validating, as it empirically demonstrates something that women in STEM and our allies experience first-hand. On the other hand, it is sad, as you note, that we are still having to prove over and over that inequality is a problem, given that the body of evidence already establishes this firmly. Nevertheless, I’m optimistic that, by identifying the types of responses men use to deny inequality, we might better tailor public education programs.


  5. That’s a good point, Zuleyka Zevallos . Data specifically on how people respond to discussions like this one is valuable.

    It reminds me of the recent surge in research in physics education. Of course many physics educators know what they are doing when they taught their students.. but the data showed a large number of educators merely think they know what they’re doing.

    The result has been a clear prescription for how to teach introductory, college level physics. It would be nice to see something similar for us outreach folks.


  6. I think you’ve got it, Adam Liss! Women overwhelmingly believe sexism exists because they’ve experienced it; they agree with science that demonstrates gender bias; and they use (often detailed) personal experiences of bias to support the science. Men are more likely to disbelieve sexism is a problem because it doesn’t impact on them personally or they think women have it easy; they disagree with science that shows otherwise; they use general statements about biology and social differences to argue against the science. 

    Both strategies are similar in that men and women are using subjective examples to illustrate that the science is either correct or incorrect. The key difference is that personal experience with sexism matches the science, and personal opinion that sexism doesn’t exist does not match the science.

    Good analogy to anti-vaxxers. I mention the example of climate science on my blog post as also being similar to denial of science on sexism. Topics that are politically loaded will lead people to disbelieve science for different reasons. This is because any science that demands a shift in personal identity (I should support women) or any measure of personal change (I should vaccinate my child) will be attacked on a personal level. People are defending their right to not change in the face of evidence that change is necessary.

    The research also shows that people who disbelieve science are more likely to attack the science publicly using confirmation bias. As you no doubt have experienced, these people seek out articles or posts that negate their personal beliefs. People who don’t feel strongly either way are less likely to speak up. (We need more of these people speak up.) Science discussions in the public are polarised and sometimes set up a false dichotomy that there are two sides that are equally valid.


  7. Hmm. This is very interesting, and I’m going to have to read the data carefully. It doesn’t surprise me that 95% of the people denying it were men, but combined with the statistic that the majority of comments were positive, it raises the question of the reverse fraction: what fraction of men were denying it? That could tell us a lot: is it a small number of noisy people, or is it a broader phenomenon?


  8. Very interesting Jonah Miller that physics education research has been embracing a critical look at what teachers think they know versus what they actually know. To be honest, much of academia is a baptism of fire. We don’t get taught how to teach, how to do peer review, and so on (we largely learn by doing). Similarly, we’re not taught how to do outreach, and most importantly how to manage negative public comments that attack science. It’s really tough! I’m glad that members of the Science on Google+ moderation team are able to privately discuss issues that arise.

    I think your point is about measuring the impact of outreach. I agree that this needs more scientific study. Are some ways of communicating science ideas better than others? When we speak up on loaded topics, what are the best strategies for keeping threads on track? 

    I didn’t cover this in my post, but Corinne Moss-Racusin’s study finds that only a tiny minority of commenters said that reading about the science changed their minds on the topic. Only eight of 831 comments said they previously did not believe in gender bias in STEM, and now understood differently. Eight! (This doesn’t necessarily mean that other minds weren’t changed; it’s that only a minute number of people felt motivated to comment to this effect.)

    What we know is that people only comment on science when they feel strongly one way or the other. What we need to measure is all the undecided people or the people who read but don’t comment. Other research suggests that the majority of people using social media are “lurkers.” Does sharing this science change their minds? We don’t really know, but we need to address this gap.


  9. Zuleyka Zevallos In a case of my own bias…  physics education research probably hasn’t been embraced everywhere. I may have a skewed opinion because it is very prevalent in both universities I have attended as an undergraduate and a graduate student respectively. (But the interest at both universities is as recent as the overall trend, which is quite new.)

    The fact that only one percent of all commenters said they changed their mind is indeed staggering! I wonder if, we followed the same commenter as (s)he read several posts discussing the same topic if (s)he would be more or less likely to change his or her mind.


  10. After reading through the article, it looks like the way they coded their dataset doesn’t allow counting the number of unique commenters, which prevents calculating that reverse fraction. That seems to me to be a failure of experimental design more than intrinsic, though, so I’m sad that they didn’t do it. I would really love to understand this aspect of sexism in STEM more.

    My gut hunch is that there’s an interesting distribution of sexism, ranging from the loud and overt to the quiet to the purely subconscious, and understanding that could have lots of practical applications. Time to do more science to it. 🙂

    (However, this comment does highlight one slightly odd methodological choice that they made, which was to code “criticism” of social science as a whole, of the particular study, and of the study’s methodology all as a single, negative-valence type interaction. They give a few examples in the text: a commenter who wishes they had compared academic to non-academic settings, “why are your y-axes not labeled?,” and “all these studies and articles only provide employment to those who wish to keep this issue of men versus women alive.” One of these things is not like the other, and the decision to code them all the same and as negative valence makes me cock an eyebrow. My own wish for more data would be coded as a negative opinion in this methodology, and in the bulk analysis, it would be grouped together with disagreement with the results. That’s just whack. Only 13% of the negative-coded comments hit this category, though, and 6% of the gender-codable comments, so this shouldn’t affect a lot of the key results too badly.)


  11. Hi Yonatan Zunger 754 comments agreed with the science on gender bias in STEM. Amongst those who agreed with results, there were four sub-categories:

    1) 560 comments agree that sexism exists: 71% are women; 29% are men 

    2) 15 comments say the results are obvious: 80% are women and  20% are men (2) 

    3) 74 comments express sense of pessimism that sexism exists and likely can’t be changed: 68% are women; and 32% are men 

    4) 105 comments made varying types of constructive comments, such as saying that sexism needs to be actively targeted. There was a relatively even gender split amongst those who noted that gender bias.

    As far as men being the noisy sub-group of disbelievers, the key take away is that among those disbelieve in gender bias in STEM, the overwhelming majority of those who comment publicly are men. The reverse is true of women who comment. Are there men who support this type of science but don’t comment? As I noted above, this may be the case. We obviously need these men to speak up. But this study specifically measures people who comment. Future studies might look at why people choose to comment on gender bias in STEM, but as I’ve discussed in other research, at the moment, only people with polarised views about science tend to comment. We need to better understand why others don’t participate in public discussions, and work on ways to engage them.


  12. Yonatan Zunger  As for why the researchers code comments that mention social science as a negative, this is because the fact that the study is a piece of social science research is used to discredit the findings. These comments are superfluous that don’t actually speak to the validity of the methods of the original study published in PNAS. These are not people saying: “here is evidence of why their methods do not match their conclusions.” These are people saying things like: “This report is JUNK science. There is no data here,” and “This is PC BS propaganda.” 

    As I illustrated in my blog post, this is one of the recurring (but certainly not the only) derailing tactic we face on Science on Google+ whenever we post about gender bias in STEM. Social science studies human behaviour. Our methods are scientifically valid as any other field. Some studies are methodologically flawed and deserve criticism; this happens in all areas of science. That is why we have peer review and why this process does not end at publication. We write about and critique published studies. The PNAS study on gender bias in STEM is not being critiqued scientifically. It is being dismissed by men who think sexism doesn’t exist, and by a minority of people who think that social science can’t ever demonstrate gender bias. Unless of course, as I’ve shown, they cherry pick random social science examples they think show women are biologically and intellectually inferior to men. 


  13. Zuleyka Zevallos This thread is an excellent example of what was illustrated by your data. Everyone who has so far questioned aspects of your experiment (albeit politely) have been male.

    I’m not going to question it at all as I have experienced gender bias for over ten years in the computer sciences. However I don’t for a moment believe that men don’t see the rude comments and sexual harassment that takes place. They are protecting the status quo. 


  14. Zuleyka Zevallos  Thanks for catching those numbers; I had missed some of them!

    Please excuse what I’m about to do: I wanted to understand the numbers better, and so I just spent a few hours reverse-engineering their calculation to get better details. You can find the results of this calculation at the very bottom, marked with a “####.” The net outcome is that these results may be even more worrisome than the authors indicated.

    The first number you listed, the 71/29 split on on comments agreeing that “sexism exists,” is the one that ultimately worries me the most. Diving into it a bit more, it looks like:

    831 total comments, of which 560 (67%) stated that sexism exists. (Good sign)

    433 total comments with some negative-valence statement*, which means that at least 162 of those (37%) also stated that sexism exists. (Good sign)

    423 comments which could be coded by gender. (52%) They note that codeability didn’t correlate strongly with any of the response categories (good).

    Of those 423, 295 (69%) agreed that “sexism exists,” in agreement with the overall fraction.

    However, of those 295, 210 (71%) coded as female, and only 85 (29%) coded as male. (Bad news)

    Now I wonder about the non-codeable subset. My experience from sites like G+ has been that non-codeability is not independent of gender: women have significantly more reason to cloak their gender than men do. 

    Given that the overall “sexism exists” fraction was so similar in the codeable and noncodeable sets, and that codeability didn’t correlate with any of the responses, and that the gender ratio for “sexism exists” on the codeable set was highly robust (χ²=52.97, p<0.001) then we might consider the hypothesis that the 71/29 split is also the gender split of the noncodeable set. 

    If so, then we would expect that of the 265 noncodeable “sexism exists” statements, 188 were women (for a total of 398 of the 560), and 77 were men (for a total of 162 of the 560). 

    Another data point is that the study states that if we drop the “sexism exists” category, there are 194 remaining positively-valenced statements. Oddly enough, 560+194=754, the exact total number of positively-valenced comments they listed, which would imply that every single one of the “sexism exists” statements had that as its only positive-valence statement. That seems somewhat odd to me (was there something in the coding which only marked this category if others weren’t present?) but I’ll take it at face value. 

    So using all of this data, we can hypothesize that the dataset looks like:

    831 total responses, of which 433 had a negative code, 560 had a “sexism exists” code, and 194 (nonoverlapping with the 560) had another positive code.

    Between 162 and 433 total responses therefore included a negative code as well as a “sexism exists” code. 

    560 people, 398 female, 162 male, said that sexism exists, had no other positive-coded responses, and between 162 and 433 of them had a negative-coded response.

    194 people had at least one other positive-coded response, and did not say “sexism exists.”

    77 people had at least one negative-coded response, and did not say “sexism exists.”

    (Here I’ve assumed that each comment comes from a distinct person, which I think is reasonable at least at first order)

    Coding the 194 and 77 for gender seems hard. The 194 has no subgroups with a strong enough statistical strength to do the same kind of generalization. While the 77 does, (especially biological justifications for sexism), we don’t have that wonderful “why yes, this set is disjoint” data that made it possible to separate out the 560 from the rest of the group.

    However, as all of the negative-sentiment categories skewed strongly male, I think that an estimate of 80/20 for the negative-sentiment group as a whole (the 433) may be reasonable. I’m basing that on the numbers for biological justification, since those have good χ² and a higher N than the other groups. Non-biological justification gives at least roughly similar numbers. Overtly sexist statements give a much higher ratio (95/5) but since those are likely to be tied to the most extreme members of the group, they seem less likely to scale.

    So if that partially blind guess is right, then of the 433, 346 are male and 87 are female. If 79% of those women (68 people) and 21% of those men (72 people) were in the “sexism exists” category — which is assuming that “sexism exists” and “negative sentiment” are IID, which is probably wrong but is at least not insane — then we would be seeing:

    330 (F, SE, n)

    68 (F, SE, N)

    72 (f, SE, n)

    90 (f, SE, N)

    19 (F, se, N)

    274 (f, se, N)

    (Where the bits are female, sexism exists, and made a negative statement, lower-case meaning feature absent and capital meaning feature present; the 19 and 274 are the 21% of the 87 women who said something negative and didn’t say sexism exists, etc)

    This tells us that the population of (N, se) is 293. Since we know that the total population of N is 433, this means that (N, SE) contains 140 people. However, we know that this is impossible from the above: by the pigeonhole principles, the size of (N, SE) has to be at least 162. This means that the hypothesis about the partitioning of the 433 people must be wrong: in particular, the probability of saying “sexism exists” while giving a negative-valence answer is not distributed across genders in the same way that saying “sexism exists” as a whole is. 

    Basically, we guessed that P(SE | f, N) = 0.21 = P(SE | f), and P(SE | F, N) = 0.79 = P(SE, F), and this must be wrong. Numerically, we find that the size of the group (N, SE) must be equal to 346 P(SE|f, N) + 87 P(SE|F, N). 

    Now, even if P(SE|F, N) were equal to one — that is, every woman who had a negative-valence statement also said that sexism exists — then we would find that we would need P(SE|f, N) to be at least 0.22. This assumption seems kind of extreme, but let’s be conservative and go with it:

    87 (F, SE, N)   (all women who offer negative sentiment say SE)

    0 (F, se, N)

    311 (F, SE, n)   (since 398 women said SE)

    346α (f, SE, N)   (writing α for P(SE|f, N))

    346(1-α) (f, se, N)

    162-346α  (f, SE, n)  (since 162 men said SE)

    And from our earlier constraint, |(SE, N)| = 87 + 346α  ≥ 162. Thus we conclude that 0.22 ≤ α ≤ 0.47. 

    If men who exhibit negative sentiment are roughly as likely to say that sexism exists as men who do not — the minimum α — then we would find

    87 (F, SE, N)

    0 (F, se, N)

    311 (F, SE, n)

    76 (f, SE, N)

    270 (f, se, N)

    86 (f, SE, n)

    0 (f or F, se, n)   (since there are 831 people total)

    whereas if men were far more likely to say that sexism exists if they also exhibited negative sentiment, we would find

    87 (F, SE, N)

    0 (F, se, N)

    311 (F, SE, n)

    162 (f, SE, N)

    184 (f, se, N)

    0 (f, SE, n)

    87 (f or F, se, n)

    Now, we know that we have a separate population of 194 people who said something positive, but didn’t say that sexism exists; these belong to (f or F, se, n or N). Now in the first case, we would discover that all 194 of them must therefore be in N — i.e., all 194 people who had said something else positive had also said something negative! In the latter case, at most 87 of them said nothing negative; therefore at least 107 people (55% of those who said something else positive, gender unknown) said something positive other than “sexism exists,” and also said something negative. Compare this against a total of 184 people who didn’t say that sexism exists, and we find:

    87 (F, SE, p, N)

    0 (F, SE, P, N)  (Nobody said SE and P)

    0 (F, se, p, N)

    0 (F, se, P, N)

    311 (F, SE, p, n)

    0 (F, SE, P, n)

    162 (f, SE, p, N)

    0 (f, SE, P, N)

    184 (f, se, p or P, N)

    0 (f, SE, p, n)

    0 (f, SE, P, n)

    87 (f or F, se, p or P, n)

    If we assume that everyone coded for either SE, P, or N, then this last row is actually

    87 (f or F, se, P, n)

    0 (f or F, se, p, n)

    We therefore find that at least 107 of the P’s must be in (f, se, P, N) — men who said both a P and an N statement. The rest must be in (f or F, P, n).


    And we can finally do a gender breakdown of the whole set. If we let x be the number of women out of the (se, P, n) group, then there are 87-x men in (se, P, n), and thus 107+x men in (se, P, N), and 77-x men in (se, p, N). Thus 0 ≤ x ≤ 77. Since I find it very hard to believe that the 19 men who made overtly sexist statements were also in SE or P, (“Sexism exists, but women simply aren’t capable of complex thought?”) I’m going to guess that the size of (f, se, p, N) is at least 43, and so x ≤ 34.


    87 (F, SE, p, N)   (Sexism exists, and a negative statement)

    311 (F, SE, p, n)  (Sexism exists, and no other statement)

    x (F, se, P, n)  (Some other positive statement, nothing else)

    Total population: 398 + x, all of whom either said that sexism exists or made some other positive statement. About 20% also made a negative statement.


    162 (f, SE, p, N)  (Sexism exists, and a negative statement)

    107+x (f, se, P, N)  (Some other positive statement, and a negative statement)

    87-x (f, se, P, n)  (Some other positive statement, no negative statement)

    77-x (f, se, p, N)  (Just a negative statement)

    Total population: 433 – x. 

    Or by their behaviors:

    Negative statement only:

    0 (F, se, p, N)

    77-x (f, se, p, N)

    (Total between 43 and 77 people, 0% female)

    Negative statement qualified by any positive statement:

    87 (F, SE, p, N)

    0 (F, se, P, N)

    162 (f, SE, p, N)

    87 – x (f, se, P, N)

    (Total between 302 and 336 people, ranging from 29 to 26% female) 

    Any positive statement without a negative statement:

    311 (F, SE, p, n)

    x (F, se, P, n)

    87 – x (f, se, P, n)

    0 (f, SE, p, n)

    (Total 398 people, (311+x/398) — between 81 and 90% — female)

    That’s an even more jarring statistic than their headline: while 398 of the 831 people were wholly positive, almost all of those were women, and 80% of women were positive while only 20% combined positivity and negativity. On the other hand, only 14-20% of men were wholly positive; 80-85% of men combined both, with 11-19% of them being wholly negative.

    This suggests a few extremely distinct populations. The “positive people” were nearly half the audience, but almost entirely female; only 10-20% of them are male. The “negative people” were also nearly half the audience, but they had the opposite gender distribution, and that group contained a hard core, about 10-19% of it, entirely male, which was entirely negative.

    That’s bad news.


  15. To summarize the above giant comment: By doing some creative data massaging to the paper, I can make an estimate of what this population of 831 comments (I’ll call them “people,” assuming each one has its own author) looks like.

    398 of these people are the “positive people:” they said only positive things, ranging from simply agreeing that sexism exists to something deeper. This group is 80-90% female.

    The other 433 said at least one negative thing. This group is 80% male, and within it there’s a smaller hard core of between 43 and 77 men who only said negative things, and nothing positive.

    If you instead divide people up by sex, it looks like there were around equal numbers of men and women who commented. About 80% of the women were strictly positive, and 20% said some of both. Of the men, about 20% were strictly positive, about 65% were mixed, and the rest were strictly negative.


  16. Zuleyka Zevallos Oh, I’m all in favor of coding people who criticize social science as a whole as negative. It’s an incredibly obnoxious derailing tactic. I was just worried about comments like “you didn’t label your y-axis,” which really don’t seem like they belong in the negative bucket.

    But after doing a few hours of computation I’m quite convinced that this couldn’t have affected their results in any significant way. 🙂


  17. Carole Wozny I was thinking about that too, and it worried me as I wrote about this. I wonder if part of this is attitudinal: when I encounter a paper that really interests me, I’m trained to tear into it and analyze the hell out of it, and I know very well that this can come off as hostile. (And that training came out of working in STEM for the past several decades, so this is by no means uncorrelated!)

    So I should probably make my meta-statement really clear: This is a really interesting study, and I think they did a great job of it. It was a well-designed experiment with a clear and very interesting output. My nits are minor and I’ve confirmed for myself that they don’t materially affect any of the results. I just spent several hours digging deeper into it because I think this paper is highly worthwhile.

    In fact, the one thing I can tell from this analysis is that the problem is every bit as bad as the worst thing you might imagine from reading the abstract, and possibly more so. It’s not merely that men are prone to disbelieving results about sexism in STEM more than women: it’s that the population responding to this has nearly sex-segregated, with the men hugely on the negative side. Only 20% of the purely positive people were men, and only 20% of the men were purely positive. And given the large ratio of men to women in STEM as a whole, the fact that so few men are even willing to recognize that sexism is a problem, or even to accept a study that says that it is, is really terrifying news.


  18. My wife just came up with the perfect summary of why I find this result so disturbing: “So you’re saying that seeing the numbers makes it harder to be in denial about denial among men?”

    Yup. That would basically be it.

    Sigh. There is a lot of stuff to be fixed here.


  19. Hi Carole Wozny I very much appreciate your comments. I am more than happy to answer questions from both Jonah and Yonatan since I know they are active advocates of women in STEM. I can’t say this for a lot of other men who say they believe in equality and then proceed to do as you point out: pick apart the evidence, or make excuses – this isn’t sexism, it’s something else, they’re just ignorant, etc. 

    I do think it’s interesting that  you’re the only woman who’s commented so far. I wonder who else might be reading and not commenting and why. 

    I’m sorry to read you’ve been through sexism in  your career in computer science. I commiserate with your comment about noticing that men see the abuse but not everyone helps. If only more men spoke up when they see it, equality and inclusion would progress more quickly. I once had a male manager say to me on my last day of the job: “I saw  you suffering and I felt bad, but I thought: she’s not my problem.” This phrase will linger with me always because it is the epitome of what is wrong with STEM.

    Whenever I write about inequality, which is almost exclusively what I write about, I have to set aside the most time when I write about inequality in STEM. I wrote the first draft of this blog post days ago but publishing it meant waiting until I could plan enough head space and time to deal with trolls. That is the reality of being a woman scientist writing about inequality: I have to factor in men’s anger, disbelief and how much energy I have to deal with negative comments, let alone how resilient I’m feeling about having to manage my own emotions about being trolled.

    Thanks for sharing your story.


  20. Jonah Miller Interesting that this is a developing trend in the two universities you’ve attended. Reflexive teaching and pedagogy are a step in the right direction. I was thinking about the way curriculum/syllabi are put together. They cover the major theories that have always been covered, with some new studies also weaved in. But most STEM disciplines do not make the time to discuss diversity issues. Women, people of colour and minorities aren’t integrated as part of everyday learning. Their science is not part of teaching; they are rarely brought into classrooms; and the issues they experience are not given any air time. I hope this also changes sooner rather than later.

    As for the one percent of people changing their minds – there’s been other research showing that people bring with them biases when they read comments sections, for the reasons I discussed above. (People who spend a lot of energy trying to discredit science are motivated by personal values.) Some of this research was used to justify the closure of Popular Science’s comments section in 2013. But as I wrote at the time, there’s no point doing public outreach if we’re not going to put in the time to moderate comments (

    That’s why we work hard on Science on Google+ to have our moderators jumping into threads and speaking up as experts. There’s also a difference between a news website reporting on science findings. They tend to sensationalise. That’s why I like G+: scientists can answer questions directly from the public. Is this type of outreach more effective at changing minds than writing an article in a big name media outlet? That remains to be tested!


  21. Yonatan Zunger Thanks very much for your detailed calculations and thoughtful comments. One issue that struck me was about gender concealment (half the dataset which could not be coded by gender). I’m not sure that the patterns of G+ hold in other mediums. I know from speaking with other women on G+ that some choose to conceal their gender as a direct tactic to avoid sexist trolling (and to evade the G+ algorithm that shows different content to different gender groups). I’m highly active on Twitter and the number of men who set up anonymous accounts to troll women writing about women’s issues is overwhelming. Women scientists on Twitter are constantly writing about this issue and how to best deal with it. Probably best not to assume that the non-coded are mostly women or that women have the most to gain from concealing gender on this issue. Gender concealment on science websites may vary according to the topic.

    Other research suggests that even with anonymity, people will still give away clues as to their gender/race/sexuality by the things they say and, most relevant here, what topics they’re willing to speak up on. When topics cover traditionally “masculine” issues, otherwise anonymous male commenters will dominate conversations and the reverse is true for  anonymous women who dominate traditionally “feminine” topics (e.g. ). This research suggests that anonymity does not reduce gender inequality.

    This may suggest that the anonymous comments may be male if they were motivated to defend STEM as an exclusive space for men. We can’t really know, but it’s interesting to consider why people hide their gender to comment on science.

    Your wife’s comment was perfect, by the way. 🙂


  22. As a father to two daughters (11 & 6) I am very concerned about encouraging my girls to pursue STEM fields due to the bias against women. I want them to do something they love and makes them happy in their careers, but it is very difficult to recommend an industry that is in my view against their gender when it has the potential to make their lives miserable. The realities of lower pay, lower ‘ranks’ and unequal opportunities wants me to encourage something that would be in an industry where it is not slanted in order to maximise the potential for happiness. The other side and why I do continue to encourage computer programming or gamer interests or science or maths or design, is for the concept that maybe my girls will be part of the group of talented women who destroy this inequality.

    Thanks for the share Zuleyka Zevallos, not so much for the depressing confirmation by metrics.


  23. Travis Koger​ but there will be significant personal gratification for themselves if they are aware and believe that their struggle to succeed in STEM is for a greater good. Bottom line is that humanity needs more women and diversity in these powerful fields or we will all pay the price, in numerous ways, for this inequality if we continue down this path.


  24. Zuleyka Zevallos Absolutely. The G+ case simply made me curious; I didn’t use it for any numerics. My gender scaling was done for the 560 who said sexism exists from the gender-codeable subset (on the theory that that subset showed such a strong statistical strength, plus the fact that the “SE” fraction was the same in the codeable set as in the non-codeable set). For the people outside that 560, I basically put in two free parameters (α and x) for gender ratios, and could use pigeonhole principle-type arguments to bound them and show that the remaining statistics were largely independent of their actual values within those bounds.

    And in fact, if my estimates are right, the sample was actually fairly close to 50/50, and there was no significant gender skew in the people whose gender wasn’t clear. Which isn’t crazy: you might see cloaking on Facebook (although FB works hard to prevent people from doing so), but on sites like Discover and the NYT, people are often using ad-hoc names and so on for reasons far simpler than long-term cloaking strategies.


  25. Thanks for sharing your story Travis Koger. I understand your concerns and it’s positive that you’re thinking about the issues that your daughters may face in the future. Despite all the roadblocks women face in STEM, it is a wonderful path to follow. No matter what happens, no matter where I go, no matter what people say or do, no one can take away my knowledge and skills gained from my degree, nor wither my love for science. It is the best investment I ever made on myself. I wouldn’t trade my STEM career for anything. I think you’re doing a great thing by reading up on women’s experiences, processing them, and contributing to public discussions about how to raise awareness and create change. I wish your daughters all the best; it sounds like they’re very lucky to be able to share their love of STEM with their dad! 

    Echoing Katy Kasmai’s words: little will change if we simply steer women away, and we rob the world of more brilliant women with the potential to change how we think, how we use technology, and how we solve problems.


  26. D Mane I just received a notification about this comment. Research shows that so-called “paternity fraud” is a marketing gimmick invented by commercial DNA testing labs, sold to fathers disputing custody battles. Reliable rates of misattributed paternity place the proportion between 0.7% to less than 3%. Even in these small cases, it is not about deception, but rather due to complex overlapping relationships. (

    Research on inequality in the family court system is also a myth perpetrated by misogynist men. The majority of custody cases are settled outside the court system, and those that go to court favour shared parenting, with the onus on children’s wellbeing. (E.g.

    Regardless this thread is about gender bias in science, which has nothing to do with your desire to get other men to join you in your abject hatred of women. Best of luck to you in working out your projected loathing.


  27. This is a great post. Thank you. At the same time, it disturbs me quite a bit to realize that my fellow male scientists (or at least men who represent themselves as scientists) cannot apply objective, rational thought to issues such as gender inequality in STEM disciplines, and other than arguing with them and trying to pound sense into their heads — metaphorically, if not physically 🙂 — I’m not sure what the solution is.


    1. Thanks for your comment rhwoodman! The way to address this is to bring the conversation into the centre of STEM. At the moment, some men support gender equality actively, by speaking out against sexism on social media, or by joining groups that discuss these issues. Other men only support gender equality tacitly; they believe in gender equality but don’t really do anything about it. A minority of men call out sexism at work, but not enough do. Yet gender issues are not discussed as part of scientific training, and so we get chronic denial about the pervasiveness of gender inequality and how it affects women’s careers. Lecturers, Principal Investigators, Heads of Departments, and Chancellors at Universities need to make gender equality central to how we do science. By law, universities and research centres have gender equity policies in place, but we know from other research that these don’t really work as they are not actively enforced. We need gender discussions at every level, from undergraduate to postgraduate, and we need gender equity to be part of management’s everyday work. There is little accountability for managers to ensure that all staff understand gender issues. The Athena SWAN Charter in the UK is one model to watch. It sets out clear goals for universities and it ranks universities based on gender measures, making accountability transparent.


  28. I am a man and have worked in a few research and developement labs. Women were always very few. I just assumed engineering was way more interesting to men than women. We guys have been building things for a million years so could it be that statistically we are better at it than women? You know that women genetically women are more selfish, less compassionate and generous than men. Of course msnbc took the oxford study to mean men are weaker than women. America (liberals) have made being a free victim very popular.


  29. Charles Ward Your sexism is ridiculous. I’ve written something using science, your comments are based on your subjective ideas and lack of education on women. Since I’ve been answering comments like yours here and on my other posts I won’t engage you further except to say: read the science; get educated.


  30. Your study is a farce. You break everything down to who’s negative and who’s positive as if one side is right and the other is wrong. Your article is negative towards men and of course you are going to get more negative responses from men. Women (dems) have taken the role of the victims so of course they are going to be all about it. Were you really looking for unbiased responses when your blog essentially says women are great and men suck. We’ve been listening to that for decades now. You dems have been spewing your calumny for so long that you think its knowledge. You should look into why you are so passive aggressive. I’m sure most of your stories go something like this. Just try to be better at what ever it is you do and as they say the cream will rise to the top. I’m sure my personal life story was much harder than yours and i made it. I’ve played my part in having a very profound effect (for better or worse) on everyones’ life in this country and many around the planet.


  31. +Charles Ward​​ Ad hominem attacks are the last resort of a weak argument. I draw on peer reviewed research and nothing you say here actually addresses my post. Your misplaced anger towards women has no place in a discussion about science. Hope you get educated on gender equity issues. 


  32. johno doh While it’s impressive you are both incapable of reading a scientific post and that you misunderstand the meaning of religion and yet you still feel threatened enough to comment on a post about women in science, alas, you will have to live out your extreme fear of gender equality elsewhere. Before you leave, try to be less scared and angry; educated women are not the enemy, and clearly not your concern. Chao


Comments are closed.