> It is important for us to know that females begin in high school to perform less well than males on mathematical problem-solving tasks. Problem solving is critical for success in many mathematics-related fields, such as engineering and physics. In this sense, mathematics skills may continue to be a critical filter.
It speculates that this may be because
> the content of problem-solving items on those tests may have heavy representation of masculine-stereotyped content, which has been shown to produce better performance by males in some studies, although results on the issue are mixed
So the Figure 1 the slide uses from the paper has smeared the critical skill, from a software-development perspective, across a large set of other math-related skills. The results of the paper with respect to that skill are inconclusive, not surprising since it's a 27-year-old meta-analysis. Surely there are better results to cite by now?
Yeah, just reading the abstract tells a different story than the graph she uses.
Google or a college level computer science course is specifically pulling from adults who are in the very top percentiles in logical/quantitative skills.
The abstract of the paper notes that while the math difference is small among the total population: "differences favoring men emerged in high school and college" and "grew larger with increasingly selective samples, and were largest for highly selected samples and samples of highly precocious persons."
So the slide deck obscures the fact that in the relevant sample, men have better skills, by including the general population (women, average folks) where there is no difference.
Well, it's refuting a lazy argument. Why would computer science be so different from other STEM fields that have radically higher proportions of women? [0] In particular, why is math so much better? [1] If biological differences are so important, why has it varied so much over time, including a big decline since the 80s? [0] I mean, we know that gender roles play a big role in what jobs people do and that sexism is a problem in tech, so it's honestly kind of weird to think "actually, it's mostly caused by some hitherto unobserved large biological difference."
Nonetheless, there is research indicating gender differences in math vary greatly between countries and appear to be correlated to other measures of gender equity [2]. Even in countries with relatively large discrepancies, there's not a big difference. I couldn't find any meta analysis of specifically programming skill, but an analysis of one CS department didn't find gender differences due to ability [3]. An older study showed women and men scored similarly on an introductory class, but women were less likely to get an A or A+ [4]. A small study of elementary students showed that girls and boys did equally well on easy problems, but boys did better on hard problems [5].
It's a much smaller field, with far fewer lucrative jobs. For any given smaller fields, there could be a number of reasons why it is more or less equal. My hypothesis is that men are biologically more inclined toward geeking out and doing coding or such for fun. If men who are good at math are also geeking and out and coding, while women just do their high school math homework, women would have a comparative advantage at plain math as opposed to computer science.
This would suggest that the gender gap could be reduced by forcing women and men to do computer programming as part of a standard high school curriculum. But the idea of forcing women to do something other than what they do in their free time, just to close a gender gap, seems to me pretty appalling totalitarian.
Nonetheless, there is research indicating gender differences in math vary greatly between countries and appear to be correlated to other measures of gender equity
Reading the studies, some things are positively correlated, some are negative. Some things have no correlation. There are lots of controls and data manipulation going on, hard to tell if the dredged for the results or not. And if you look at other types of studies you see the opposite result -- http://slatestarcodex.com/2017/08/07/contra-grant-on-exagger...
I couldn't find any meta analysis of specifically programming skill, but an analysis of one CS department didn't find gender differences due to ability
If the scores are equal, this would suggest no bias in selection. If women are unfairly excluded from computer science, or face a higher bar due to discrimination, then since they were more selected their scores should be higher.
A small study of elementary students showed that girls and boys did equally well on easy problems, but boys did better on hard problems
Keep in mind girls mature faster and differently, so differences in elementary school aren't all that predictive of differences later.
Thanks! However after looking through the slide deck, it seems that the author tries to argue that because mathematic ability isn't so important in CS, biological differences can't explain the disparity between the men and women in CS. That's a quite a leap, I think.
If CS doesn't involve a lot of math, what does it involve that might be impacted by biological or psychological causes?
(P.S. I went to a STEM high school and I know plenty of women who are absolutely stellar in mathematics and science, so I know that's not the issue here.)
No, that's not the argument, perhaps read all the slides, in particular slides 20/21. It's true that cs doesn't involve a lot of math, but that's really just a side point and a little in joke for mathematicians. The argument is the disparity is so small that it could not account for the massive disparity we see in tech.
Re your question on biological or psychological causes, why limit your consideration to those areas? The evidence points to social or economic causes, given the large shifts in women over just a few decades:
When I looked into this a lot, looking into all sorts of data, reading tons of stuff, my conclusions were:
1. Women are generally more conscientious and studious. I've seen data they do more homework, get better grades, even in math, and are more likely to make the honor roll.
2. The more difficult the math test, and the higher the score bracket, the more men outscored women. Average SAT score for men was only a little higher. But among those scoring 800, men outnumbered women two-to-one (this has changed a bit, but they have also made the math test a lot easier, far more testers get an 800 math than get an 800 verbal). Then when looking at who gets a 100+ on the American Mathematics Competition tests, men dominated by about a 10 to 1 ratio.
3. I have observed throughout life that men seem a lot more likely to "geek out" on stuff -- whether that be coding all night for fun, editing Wikipedia articles, taking apart a mechanical device, trying to beat all the quests in an RPG, etc. This is inline with studies showing men to be much more system oriented. I have a hard time believing this is due to culture messaging because for a lot of these things men receive enormous cultural messaging that they are losers for being such geeks, yet they do it anyway because it is such fun.
4. Men tend to have higher variance in almost all things. More men at the top, more men at the bottom.
5. This Stanford article notes "Men, on average, can more easily juggle items in working memory. " http://stanmed.stanford.edu/2017spring/how-mens-and-womens-b... I think that skill becomes more important the harder a math test becomes, or the harder a programming project is.
I think the 1970's style computer punch card programming, which was very corporate, and did not involve as much keeping large complicated systems in working memory, was less prone to creating a gender imbalance. It more favored conscientiousness.
Whe computers became available in every household, men were much more prone to geek out and become expert programmers on their own time. This means the pipeline of programmers ends up being much more male dominated.
Furthermore, the skills required to be a computer programmer at a top company like Google, require someone from the top percentiles in being able to juggle items in their head, and there are a lot more men in those percentiles than women. I think programming at Google is much more akin to doing well on the AMC than it is to answering an average SAT-level math problem. When we not the disparity in math ability is small, it obscures that the disparity at the very high ends is much larger.
I read all the slides and I think the commenter raises a good point. If math doesn't matter a lot when it comes to CS, then why bother talking about it.
Show me how many standard deviations apart men and women are when it comes to skills that lead to success in CS.
Rather than looking at the math ability distribution as a proxy, why not look at coding competition ability? They are pretty objective since an autograder doesn't know your gender and they are predictive of whether you can pass a tech interview since it's the same format.
You see the same thing at the college level too (for example topcoder where they isn't even any possibility for a sexist team selection bias since anyone can just register and compete).
In some sense, it makes it not tech industry's fault. It's a failure of our education/training pipeline where we are not training competitive women even from an early age.
> Testosterone has effects on competition, so there could still be bias there
Oh. And that would be a biological, innate difference in behaviour between men and women, wouldn't it?
And even if women are as good as men at maths or computer programming- which wouldn't surprise me much- can't they be just less interested in it, just in the same way they're less interested (as you seem to imply) in competition?
And could I be as good as the average woman in teaching or as a nurse? Possible. Am I interested in it? No.
Because, do you think sitting eight hours a day behind a desk trying to find a way to instruct a machine to do something you already know how to do, is a particularly appealing job? Can you imagine the amount of fulfilling human interactions a nurse has every day in the hours you spend looking at that damn screen? I mean, maybe they don't do it just because they can actually do something better.
Many people would disagree, and some would have you fired for saying that.
As far as the presentation, it accurately notes that as far as we can tell, the average difference in ability between men and women is very small or zero. But there is compelling evidence that men and women do naturally differ in interests (the "people vs things" thing), which if true would directly lead to unequal representation in various fields.
> Many people would disagree, and some would have you fired for saying that.
This mentality is exactly why the Google enginner made the post. The company and many others like it are forming echo chambers where anyone with any opinions that go against the status quo is fired. That's ridiculous! Especially with something as nuanced as sex and gender.
It is a known fact that there are physical[1], emotional, and psychological[2] differences in people around the world. If there are psychological and emotional differences between the genders, then why is it so surprising to think that they would overall have different career preferences? The fact that someone would fire an employee instead of let them question the boundary between biology and cultural influences is absurd.
M = male / female
X1 = math ability
X2, X3 ... Xn = other factors (analytical ability, ability to work in teams, ability to understand real problems, etc, also bias of interviewers against women)
Q = performance on interview questions
W = performance in real life work
The slides, and most peoples argument is based on the model where there is only one factor math ability, which is why people go 'oh but that doesn't account for it / the disparity is so small'.
The important question is how do all the factors ADD UP?
Now,
If Q does not correlate with W, fix your interview questions first.
Then, if you want more women to do well in the interview, well, ...
-- the wrong approach is to compromise the integrity of the interview, which compromises the company's business, and which is demeaning to women, although you could use this as a proxy to make the interview more comprehensive
-- the correct approach is to investigate X1, Xn factors and train to remove these discrepancies earlier on way before the interview itself, if we decide that we want to. biases in the interview processes are only biases when removing them improves the correlation between Q and W
Another thing to do is to change the nature of the work itself. In which case the factors and interview questions will change accordingly. Play to peoples' strengths!
The nature of the work must include the performance of the team including the individuals as there may be a benefit to representation when it comes to solving problems of a crowd with varied people.
Interestingly, the Google article covers a lot of this, and even suggests some changes that can be made.
A large number of factors can quickly add up even if the individual factors as small.
I welcome comments from the more knowledgable, but I feel that a lot of knee jerk reactions here are just taking individual statements from the argument and loudly saying NO, or saying well there's no difference here.
Programming is one of the most difficult branches of applied mathematics; the poorer mathematicians had better remain pure mathematicians. Edsger W.Dijkstra, EWD498
https://www.slideshare.net/mobile/slideshow/embed_code/key/3...
Of course men and women are different on average, but are they different enough to justify the massive disparity in tech? No.