I was sent this paper (by Amanda B. Diekman, Elizabeth R. Brown, Amanda M. Johnston and Emily K. Clark; behind a paywall-- sorry) in response to my earlier post about research into the gender disparity in computer science. This particular paper more broadly considers all of STEM (science, technology, engineering, and mathematics), but seems applicable to CS in specific. It can very much be seen as a complement to the Cheryan et al. paper we discussed before. That paper focused on a mechanism which women are prevented from joining CS. This paper focuses on one possible underlying reason this mechanism works. As I know that was of interest to some of my readers, and it will also give me a platform on which to explain why I care about this issue so much, let's see what this paper says.

Briefly put, the paper concludes that:

  1. Women (as a population) prize communal goals (working together, or helping people) more than men (as a population).
  2. STEM careers are perceived as less likely than other careers to fulfill or further these goals.
  3. Therefore, women (as a population) are less likely to choose STEM careers then men (as a population). QED.

I don't think anyone disputes that women (I'm going to stop saying 'as a population' now, but please understand that it is still there) choose STEM careers less than men, so the interesting part of that last statement is the 'therefore.' Showing causality is hard, and I'm not sure that they actually achieved it, but they did manage to show something called 'mediation.' The actual experiment went as follows:

First, Shanghai a few hundred (333) 18-to-19 year old undergrad psych students (plus 27 paid volunteers from STEM majors) into taking a survey. The survey listed a bunch of careers (including STEM careers, non-STEM male-stereotypic careers, and female-stereotypic careers) and a bunch of goals (communal and individual; listed below). Given this set-up, the survey asked:

  • For each career, what do you estimate the percentage of women to be?
  • How much will each career fulfill each goal?
  • How interested are you in each career?
  • How important to you is each goal?

Next, fire up SPASS. The first question allowed the researchers to validate their assignment of careers to male-stereotypic and female-stereotypic, and STEM careers. In what I am sure will be a surprise to no one, the STEM careers were considered to be more 'male' than things like 'dentist' and 'architect', and waaaay more male than things like pre-school teacher and HR manager. (The careers were chosen on the basis of actual gender ratios, but this allowed the researchers to validate that these careers were actually perceived as being gendered, too.)

But that was just warm up. The real meat of the analysis was statistically verifying three things:

  1. STEM careers are believed to impede communal goals.
  2. Communal-goal endorsement negatively predicts STEM interest.
  3. Communal goals mediate gender differences in STEM interest.

As with the previous paper, I'm not quite sure what 'mediate' means in this context. I think that it means 'is a likely layer of indirection between gender difference and STEM interest.' If so, I think that this result means that

  • They compared two models:

    1. The gender of the responder directly predicts their interest in STEM, and
    2. The gender of the responder directly predicts their endorsement of communal goals, and endorsement of communal goals directly predicts their interest in STEM careers.
    3. And that they found that their data is more consistent with the second than with the first.

Oh, and I promised to list the goals, didn't I? Here they are:

  • Communal goals: Helping others, serving humanity, serving community, working with people, connection with others, attending to others, caring for others, intimacy, spiritual rewards.
  • Agentic (which means 'individual,' I think) goals: power, recognition, achievement, mastery, self-promotion, independence, individualism, status, focus on the self, success, financial rewards, self-direction, demonstrating skill or competence, competition.

And for good measure, here are the careers:

  • STEM: Mechanical engineer, computer scientist, aerospace engineer, environmental scientist.
  • Non-STEM male-stereotypic: Lawyer, architect,dentist, physician.
  • Female-stereotypic: preschool or kindergarten teacher, human resources manager, social worker, education administrator, registered nurse.

So, what to make of all this? I, personally, find it very frustrating. Not because of any social justice issue-- I don't actually care about social justice. I don't think that a CS career is some sort of reward or privilege that needs to be distributed fairly. I don't think that a gender disparity in CS is inherently wrong or harmful, or that an ideal world requires gender parity to be ideal. That's not why I care about this issue. The reason I care about this issue is that I want my kids to grow up in a world without cancer. And with cold fusion. And jetpacks.

See, it turns out that computation is actually a bottleneck hindering technological progress. Consider genome sequencing: From what I understand, we don't have technology that actually sequences an entire genome as a biological process. Instead, current technology breaks a 3-billion DNA strand into billions of shorter strands and sequences those. How do you use these tiny little fragments to reconstruct the actual full DNA strand? You have to find overlaps in the sequenced portions and combine them back together like a huge, one-dimensional jigsaw puzzle. And to do that, you need computation. Lots of it. (And I didn't even mention that sometimes the biological process makes errors in the sequencing...)

Or, as another example, take protein folding. Once you get these DNA sequences, you can deduce what proteins will be built. That is, you know the chemical structure of this molecule, but what actually matters is the way this molecule folds up in three-dimensional space. And to figure that out, you need to simulate the motion of the molecule as it folds up. To do that, you need to calculate the force each atom exerts on each other atom, find the cumulative force on each atom, turn that into the motion of direction for that atom, move that atom a teeny-tiny bit in that direction, and repeat. And the best computer we have for doing this takes 100 days to simulate a millisecond of this folding.

I'm told that there are similar examples from physics, fluid dynamics, and so on. In many fields, computation is a bottleneck. But it doesn't have to be this way. In many cases, there could be some clever way to speed these computations up, just waiting to be found. These problems could conceivably be solved by the right person having the right insight. And the best strategy I can imagine for finding these insights is to get as many people as possible thinking about these problems.

This is why I care. I don't care that women are underrepresented in CS because women are underrepresented in CS. I like women, note, and enjoy mixed-gender environments better than single-gender ones. And let's be clear: harassment is wrong and should be eliminated. (Okay, that I do care about for its own sake.) But the real reason I get upset by evidence that women are leaving or avoiding CS is that it is evidence that people are leaving or avoiding CS. And right now, I think that human progress needs as many people in CS as we can get.

...Which leads back to why I find the results of this paper so frustrating: it tells me that a class of people (women) are more likely to avoid CS careers than would be otherwise the case because they perceive it as impeding communal goals, yet it is exactly such a goal (serving humanity) which make CS such an important field.

(Of course, I still don't have any good ideas for what to do about this issue, but that's for another post.)