That Google memo business has got me thinking.
The author got himself into trouble for a bunch of reasons (too much logos and not enough pathos for a start) but I thought the core of his argument, restated to be more positive, could be appealing to many of the people who condemned it.
I’m going where angels fear to tread. Wish me luck!
Diversity is good for software companies. It helps us make better products.
I work at a little software startup just up the road from Google. Men are a minority at my company and I like it that way. It makes for a more pleasant work environment. I’ve worked on too many all-male teams and they tend [stereotype alert!] to be much more competitive and less collaborative. With women around, the men try — but sometimes fail — to be on their best behaviour.
Less selfishly, I think that our diversity makes us a better team. The women [stereotype alert!] really do see things that the men don’t see (and vice versa).
Men and Women are Different
Although there are differences in interests and traits between men and women at the group level, we should judge individuals as individuals, not as members of their group.
I don’t think either point is controversial, though the second might be hard to accept if your blood pressure is already high from reading the first. There’s a long history of people using statistics like this for ill purposes but I hope and expect that we can avoid the mistakes of the past with a little goodwill and a generous reading of the evidence.
There’s no need to posit genetic explanations for the differences. Cultural history can explain it adequately well.
There is chronic sex discrimination in the hiring process.
I have no doubt about this and we should certainly fix it. But bias in the hiring process and discrimination in the workplace are inadequate explanations for the gender imbalance in software engineering.
Creating awareness of implicit bias during the interview process is certainly helpful but it’s not enough to fix the whole of the problem. If we want to get more women into software development (we do! we do!), we need to look further. We need to consider other solutions too.
Vive la différence!
Men are taller than women on average but, if I need tall people for a job, I would be foolish to go out to hire a bunch of men. I can just measure the candidates. If they are tall enough, I can hire them without considering their gender.
The analogy doesn’t quite work because an objective quality like height is easy to measure. Subjective qualities of the kind we might consider when hiring a software engineer can be harder to untangle from our biases and our stereotypes but we can try.
Once we’ve decided that our candidate is good enough (or tall enough) for the job, it’s not helpful to consider whether they have the right configuration of sex chromosomes (individuals, not groups remember?). But maybe it’s useful to think about group differences during the interview processes itself?
The evidence from psychology seems to suggest that women and men are quite different, on average, on certain traits like neuroticism and agreeableness even if there is considerable overlap between the sexes. Could those differences make the interview process more treacherous (on average) for women? I think it does, especially when you consider that most of the interviewers are men.
I don’t know of any work that looks at agreeableness as a liability in interviews but there was an interesting experiment recently at interview.io that tried to understand why women are less likely to succeed in interviews.
Make interviews more social?
interview.io is a website that gives software engineers an opportunity to practice interviewing in a safe environment. They noticed large differences in outcomes between male and female candidates.
we’ve amassed data from thousands of technical interviews, and in this blog, we routinely share some of the surprising stuff we’ve learned. In this post, I’ll talk about what happened when we built real-time voice masking to investigate the magnitude of bias against women in technical interviews. In short, we made men sound like women and women sound like men and looked at how that affected their interview performance. We also looked at what happened when women did poorly in interviews, how drastically that differed from men’s behavior, and why that difference matters for the thorny issue of the gender gap in tech.
Here’s what they knew before the experiment:
Specifically, men were getting advanced to the next round 1.4 times more often than women.
They designed the experiment expecting to confirm a bias against female interviews but…
Contrary to what we expected (and probably contrary to what you expected as well!), masking gender had no effect on interview performance with respect to any of the scoring criteria (would advance to next round, technical ability, problem solving ability).
They didn’t really come up with a satisfactory explanation for the failure to uncover bias but an author can speculate…
Anecdotally, it seemed like women were leaving the platform a lot more often than men. So I ran the numbers.
What I learned was pretty shocking. As it happens, women leave interviewing.io roughly 7 times as often as men after they do badly in an interview.
If women drop out so much more during the interview process, might we see the same behaviour throughout a woman’s fledgling career in software engineering?
The author speculates some more…
If that’s true, then we need 3 times as many women studying computer science than men to get to the same number in our pipelines. Note that that’s 3 times more than men, not 3 times more than there are now. … [snip]… to get to pipeline parity, we actually have to increase the number of women studying computer science by an entire order of magnitude.
Maybe the author of the Google memo was on to something when he wondered if neuroticism was significant when we look at the differences in the number of male versus female engineers? Just not in the way you thought.
Armed with this new information, perhaps we can change the way we interview engineers?
I’ve interviewed and been interviewed perhaps hundreds of times over my career. I enjoy being interviewed. But I can honestly say that my interview at Google was the least pleasant that I have ever endured. The interviewers were completely uninterested in my hopes and dreams and were actively hostile when I tried to engage them in conversation.
Maybe Google has got better at interviewing since then. If they haven’t, maybe they could try? No need to lower the bar on quality. They could start by acknowledging that there’s more to being a great software engineer than solving Big O problems.
Maybe we can make software engineering more social?
The author of the interviewing.io study suggested that we can’t fix the gender imbalance by better interviewing alone. Is there anything about software engineering itself that is off-putting to women?
Let’s first acknowledge that the predominance of men off-putting to women. Being the odd one out on a team is a high cultural hurdle to clear before we even think about differences in traits and interests. The evidence from other disciplines suggests that the hurdle is not insurmountable though.
Women now progress to PhDs as often as men in science, maths…
…and engineering and women now dominate health professions and life sciences.
If the life sciences can do it, maybe software engineering can too.
Software engineering has a poor image in the outside world. Most people think of a software engineer as a lonely introvert locked away in a a cubicle and typing code into a computer all day. Maybe Google is like that, but I haven’t worked in such an environment for over fifteen years.
Since I discovered Kent Beck’s crazy way of making software, I’ve come to value interactions and collaboration much more than the old process of turning specifications into code. I spend most of my day collaborating with other people.
If young women knew that software engineering was highly collaborative, might they be more inclined to give it a try?
Isn’t that just what Damore argued in his memo?
I read four or five rebuttals of the Google memo before I read the actual memo itself and I was quite surprised, when I finally read it, to find that the memo was neither as crazy nor as hostile to women as I had been led to believe.
Rather than refuting his points, most of the rebuttals were reiterating what Damore himself had said in his memo.
Now, granted, my argument is not exactly the same as Damore’s but it’s not far off. Damore would’ve done better to have saved his rant about liberal bias at Google for another day. And the rant about extra support for women and minorities would have sounded better if he had shouted it at the clouds.
He got a lot of things right though.
In conclusion, based on the meta-analyses we reviewed above, Damore seems to be correct that there are “population level differences in distributions” of traits that are likely to be relevant for understanding gender gaps at Google and other tech firms. The differences are much larger and more consistent for traits related to interest and enjoyment, rather than ability.
Population differences in interest may be part of the explanation for why there are fewer women in the applicant pool, but the women who choose to enter the pool are just as capable as the larger number of men in the pool. This conclusion does not deny that various forms of bias, harassment, and discouragement exist and contribute to outcome disparities, nor does it imply that the differences in interest are biologically fixed and cannot be changed in future generations.
If our three conclusions are correct then Damore was drawing attention to empirical findings that seem to have been previously unknown or ignored at Google, and which might be helpful to the company as it tries to improve its diversity policies and outcomes.
Most of the first round of criticism missed his point entirely. Gizmodo called it an anti-diversity screed .
I value diversity and inclusion, am not denying that sexism exists, and don’t endorse using stereotypes. When addressing the gap in representation in the population, we need to look at population level differences in distributions. If we can’t have an honest discussion about this, then we can never truly solve the problem.
Conor Friedersdorf at The Atlantic summarizes the misleading coverage well.
To me, the Google memo is an outlier—I cannot remember the last time so many outlets and observers mischaracterized so many aspects of a text everyone possessed.
Casually perusing “anti-diversity” headlines without reading the memo might mislead readers into thinking a Google employee had assigned a negative value to gender diversity, when in fact he assigned a positive value to gender diversity, but objected to some ways it was being pursued and tradeoffs others would make to maximize it.
Sabine Hossenfelder at Nautilus wonders about the wider implications of his memo.
Damore’s memo strikes me as a pamphlet produced by a well-meaning, but also utterly clueless, young white man. He didn’t deserve to get fired for this. He deserved maybe a slap on the too-quickly typing fingers. But in his world, asking for discussion is apparently enough to get fired.
If you remove biology from Damore’s notion of “population level differences”, his critique is still nearly as powerful. And his question is still valid: “If we can’t have an honest discussion about this, then we can never truly solve the problem.”
Damore was fired, basically, for making a well-meant, if amateurish, attempt at institutional design, based on woefully incomplete information he picked from published research studies. But however imperfect his attempt, he was fired, in short, for thinking on his own. And what example does that set?
Bring your daughters to work. And keep them there.
After the last round of “Silicon Valley Is Sexist” outrage, I asked each of the very smart, very capable women of my acquaintance why they did not become software engineers. They all answered with some variant of “I can’t think of anything less appealing than the idea of working with a machine all day” or “I would hate to work in an environment where I was the only woman”.
I also asked all my male friends why their daughters were not interested in computer science. They all sighed and said “I just can’t get her interested”.
Only 18% of CS graduates are women. If we want to attract more women into software engineering (and we should – it’s a fun job; very social and pays well!), we have to find a way to get our daughters interested. We are not gonna fix the whole industry by making interviewers less biased.