Facebook’s gender bias goes so deep it’s in the code
Female software engineers found their work was rejected 35 percent more than male engineers, according to a Wall Street Journal report.
A hurricane has been brewing at Facebook.
After years of suspicion, a veteran female Facebook engineer decided to evaluate what if any gaps there were in how female and male engineers’ work was treated.
She did it “so that we can have an insight into how the review process impacts people in various groups,” the Wall Street Journal learned exclusively.
Her analysis, conducted in September, found that female engineers’ work was rejected 35 percent more than their male counterparts based on five years of open code-review data. Women also waited 3.9 percent longer to have their code accepted and got 8.2 percent more questions and comments about their work.
Only 13 percent of Facebook’s engineers are women, 17 percent across all tech roles.
So how do we rule out the idea that the women’s code was worse?
Ah, guess what, that hasn’t even been considered. Odd that, eh?
This might be rubbish, but at least one proper study has done been over in the Open Source world, and made for uncomfortable reading –
http://www.livescience.com/53729-bias-against-female-coders.html
Not sure if this is the same study…
http://m.slashdot.org/story/307061
If people expect code reviews to be approved on the basis of gender, why fucking bother having them?
Easy to check-off do those accepting/rejecting the code know anything about the coder, or do they have to decide on the merits of the code.
If no personal knowledge, then no prejudice.
If there is- then maybe yes maybe no.
The big problem is inexperience and high rates of attrition. Women don’t last long in coding before moving on to say, testing, management, business analysis or just out doing something else.
Code reviews have a number of purposes. Sharing stuff around the rest of the team for knowledge, informing developers about other ways to do things (even though what they’ve done is fine) and gatekeeping bad code. When you have a rejection, it’s nearly always used for inexperienced developers.
@Bob Grahame
“This might be rubbish, but at least one proper study has done been over in the Open Source world, and made for uncomfortable reading ”
This study was analyzed to death when it came out. So here’s where it starts to get wrong:
“To get closer to answering that question, Josh Terrell, a computer scientist at…”
This is a social studies survey conducted by a computer scientist. Any social scientist, whether economist, sociologist, or psychologist spends a large portion of any study discussing possible biases in the sample. Not in this case. This study went the “we make up for it in volume” route of thinking that a large sample size will eliminate bias.
So how does this work? They made a very long list of contributors in github. Many of those used their gmail account to register for github. Every one of those has a Google+ profile, usually empty (because who uses Google+?). But about a third of those have indicated their gender on Google+.
So we now have a shorter list of github contributors for whom we have a declaration of gender in Google+. Is this a representative sample of github contributors? Who knows? The word “bias” does not appear in the study. But we’ve got tens or hundreds of thousands so that makes it OK, no?
Next they wrote an algorithm to split up their github userids in three groups: obviously male (like “BruceJenner”), obviously female (like “CaitlinJenner”) or indeterminate (like “synp”). Wanna bet it mis-classifies “JustinBieberBiggestFan”? But the important issue is that again we don’t account for what might cause one programmer to choose a gender appropriate handle or not.
The results? Women with indeterminate handles had their code accepted more often than men with indeterminate handles. Men with manly handles had a bit less. Women with feminine handles had even less. That’s raw data from a highly-biased survey.
Interpretation #1: Women are better programmers than men, but when code reviewers know they’re women, they actively discriminate.
Interpretation #2: Women are less self-confident on average so they send pull requests only if they’re very sure it’s good. Indeterminate handles win over gendered handles because that’s the kind of handle that a geek chooses.
Interpretation #3: It’s s biased survey. Drawing conclusions from it is a fool’s errand.
When I was doing code testing I had absolutely no idea who had written the code. I tested the code, submitted a report to the reporting system thingy, if it went ‘ping – update ready’ I tested it again and if it failed I would fail it based on whether the code worked, not on a guess as to which random coder may have happened to have done the update.
The updates were tagged with initials, but I never got around to joining initials to humans, let along what sex or gender that human may have had. ‘Update by ERT: resolved mailcount never zeroing’. Never knew who ERT (etc) was. The coders all lived in their den at the other end of the building. There were some women, the coding pool was on the way to the front door.
Transnational progressivism assumes that “victim” groups should be represented in all professions roughly proportionate to their percentage of the population. If not, there is a problem of “underrepresentation.”
https://en.m.wikipedia.org/wiki/Transnational_progressivism
Synp, thanks, you led me to http://slatestarcodex.com/2016/02/12/before-you-get-too-excited-about-that-github-study/
It always seems suspicious to me that a subgroup known to be less people-oriented would somehow be unusually sexist. As opposed to just being unappealing to women.
“Transnational progressivism assumes that “victim” groups should be represented in all professions roughly proportionate to their percentage of the population. If not, there is a problem of “underrepresentation.””
Note the weasel word in that lot – ‘professions’. Some how it ignores entirely the mass of shitty dangerous jobs that just happen to be almost entirely male in their makeup. No great fuss being made that there’s not more female sewer cleaners, or deep sea fishermen, or high voltage cable repair engineers, or bricklayers. Funny that.
Feminism – aiming to get women into all the well paid office jobs while ignoring all the shitty manual ones for over 100 years.
On a business course recently the diversity officer reported with that a decade ago less than 5% of board positions at FTSE 500 companies were filled by females, but now that figure was just under 20%.
He encouraged us all to celebrate this fact.
I asked whether now we had a decent amount of data we should measure whether there was any material difference in business performance at the companies with more female board members.
It was like that scene in American Werewolf in London.
Yokels. Jukebox.
@Jim,May 3, 2017 at 7:18 pm
+1
Coding is mostly creative binary logic.
imho most females are not pre-disposed to logic (map reading!).
QED most females are not good programers.
“imho most females are not pre-disposed to logic”
Strangely, most women don’t get fuzzy logic either!
https://en.m.wikipedia.org/wiki/Fuzzy_logic
‘Female software engineers found their work was rejected 35 percent more than male engineers’
Easy solution. Outsource coding to India. It’s a world market, ladies. You aren’t secure enough to whine about such shit.
On a business course recently the diversity officer reported with that a decade ago less than 5% of board positions at FTSE 500 companies were filled by females, but now that figure was just under 20%.
He encouraged us all to celebrate this fact.
diversity officer…he…
WHAT? I thought they were all female.
Interesting reading this after catching up with the ‘sexism’ drama at Drupal (an open source content management system):
https://reason.com/blog/2017/04/18/drupal-developer-ousted-over-kink
The TL:DR version – one of the project’s leaders lost his position after being outed as a fan of John Norman’s Gor novels (cheezy kinky pulp sci-fi), as well as for having a consensually dominant sexual relationship with his girlfriend.
One thing I noted was how many of the androgynous blue-haired ladies calling for his scalp weren’t actually programmers themselves. They were quick to talk about how his male dominant presence would stop Drupal being a safe space for women coders, but most were ‘diversity and inclusion consultants’, or ‘women in tech evangelists’.
It was also interesting to see how some of them turned on actual women coders who defended him, including some who’d actually worked with that developer, and who challenged the idea that he was a misogynistic abuser. I think the term I saw thrown at them by the ‘diversity’ parasites was “filthy snake traitors” – so warm and inclusive.
Well, yeh.
If Indian coders weren’t reliably more shit than doped up UK coders (of whatever sex, gender, preference or ethnic origin).
Eazy-Peazy-Leeeemon-Squuuueezy.
You just get an all female workforce of coders and reviewers. They are better than men and so you will soon have a company that soon makes lots of money and soon rules the world.
Why hasn’t someone done it?
Especially some disgruntled women working elsewhere, funded by some trade union?
I am personally offended by his cis-het arrogance in defining what a female coder is. Just because someone has a penis, much less a male name, doesn’t make them any less of a woman.
We need to move beyond these binaries. Let’s not assume that any of the programmers in question are confined by the narrow boxes of patriarchal logic categorical oppression.
Kendall – “Interesting reading this after catching up with the ‘sexism’ drama at Drupal (an open source content management system):”
That was f**king awful. I may not have much time for people who are into Gor. In fact I have very little. But because someone did not make a complete 100% secure secret of his sexual preference he is ridden out of town on a rail because he *might* sexually harass someone …? Well it makes me think Gay men should not be scout leaders.
Gor fandom is OK, people often do things in private that they would not do in public.
What next? Sack all Agatha Christie fans because they like reading about murders?
Sack all Warhammer & Warhammer 40k fans because they like re-enacting violence / reading about violence?
I like the Gor novels. No desire to emulate them.
I also like Miss Marple. 🙂
The Gor stuff aside, did you see the ‘problematic’ views that were highlighted as additional reasons to get rid of him?
The worst things they could find him argue were:
-that someone’s sex shouldn’t matter when it comes to their code
-that sex is linked to biology, even if gender is social
-that using ‘master’ in computer terminology really isn’t a big deal
That’s it, all they could find after dredging through a decade of contributions… Of course they still labelled him sexist and racist, and declared his views unacceptable and ‘exclusionary’.
Two other things I found particularly creepy about the whole Drupal drama:
1) The fact that the guy dared to defend himself after being outed, smeared as a misogynistic abuser, and expelled from the project, is itself being treated as an offence.
It was wrong of him to even question what he was accused of. Even if it was all made up to smear him, he should still have quietly taken one for team diversity. Because he made it public, any negative consequences resulting from this kerfuffle are entirely his fault.
That from the same people who rant about ‘victim blaming’!
2) SJW types are using this to argue that existing Codes of Conduct don’t go far enough in making coding a ‘safe space for women’.
In particular, they’ve condemned the fact that the code only covers behaviour related to the project, and not what people do in their private lives. These ‘diversity and inclusion’ liberals are really upset that the Drupal CoC slightly complicated the process of punishing someone for having a sex life they found icky.
Some of these people genuinely want an intrusive investigation into the personal life of anyone who contributes code. Only people who demonstrate properly progressive behaviour at all times would be welcome. Not just sexual behaviour, but thought crimes like enjoying the wrong films/books/music, or making comments in support of freedom of speech, would violate their desired Code of Conduct.
In my opinion, if they actually get what they want they’ll just end up destroying their own communities from within.
Andrew C,
“Eazy-Peazy-Leeeemon-Squuuueezy.
You just get an all female workforce of coders and reviewers. They are better than men and so you will soon have a company that soon makes lots of money and soon rules the world.
Why hasn’t someone done it?
Especially some disgruntled women working elsewhere, funded by some trade union?”
As Timmy has pointed out a couple of times, Dame Stevie did just that and was very successful, until she fell foul of the sex equality laws.
The history of women coders is quite interesting. It seems that some marketer in the ’80s decided that PCs were toys and only men and boys played with those sorts of toys.
“A lot of computing pioneers — the people who programmed the first digital computers — were women. And for decades, the number of women studying computer science was growing faster than the number of men. But in 1984, something changed. The percentage of women in computer science flattened, and then plunged, even as the share of women in other technical and professional fields kept rising.
What happened?
We spent the past few weeks trying to answer this question, and there’s no clear, single answer.
But here’s a good starting place: The share of women in computer science started falling at roughly the same moment when personal computers started showing up in U.S. homes in significant numbers.
Who Studies What? Men, Women And College Majors
PLANET MONEY
Who Studies What? Men, Women, And College Majors
These early personal computers weren’t much more than toys. You could play pong or simple shooting games, maybe do some word processing. And these toys were marketed almost entirely to men and boys.”
http://www.npr.org/sections/money/2014/10/21/357629765/when-women-stopped-coding
I worked with plenty of effective women in IT, including some in senior management positions. They almost all started in ‘data entry’ (some in the days of Hollerith machines, prior to commercial computing), which was seen as an extension of secretarial work and hence 100% female. For a young male trainee to visit this area (or the typing pool) could be quite ‘dangerous’.
BiND,
“What happened?
We spent the past few weeks trying to answer this question, and there’s no clear, single answer.”
Do you know what the real answers are? The non-NPR answer? The answer that isn’t “it’s about the patriarchy”? The answer that is based on my own observations about computer departments over the decades?
The biggest one is that home computers lowered the barrier to entry in the computing market. Huge numbers of people could afford a ZX81 or VIC-20 or get access to a BBC B at school and learn to program. They could find that they had an ability in it and develop it. See, before then, you had to go to college, get on a PDP-11 to learn to program and get an exam in it, and then you’d get a job. That often meant excelling at many exams at A level, which weren’t necessarily signals of being good at programming (maths and physics might be). And university was never *that* rigorous where it mattered. So, companies hired people who’d passed the exams, spent a fortune training them, and then found they weren’t much good. But what do you do? They’re still so-so, and firing them means training someone else up. I didn’t just work with lots of women, I worked with lots of utter donkeys.
And the home PC exacerbated the market effect. Once blokes had their own computers, they started tinkering with stuff in their spare time. And women didn’t. There were 3 of us guys playing around with web stuff in an office in the early 2000s in our spare time. So, who do you think got picked to build a web-based system when the company needed one? Those women who’d need a load of training expenditure, or the blokes who were almost there already?
I’ve worked with women since, and none of them have pet projects. They all do a 9 to 5, and they’re good workers, but they can’t keep up. The guys who are spending an hour building a game or an app are just improving at a far faaster rate.
Oh and the other thing is that there were lots of things in business that used to be “computer department” stuff like business analysis that no longer is. So “women in IT” used to get counted in that, but that’s now more a “business development” area.
If the women’s code is worse, that’s an extraordinary claim that requires extraordinary evidence. The groups in question are large enough to form statistical populations and gain reliable data, so we shouldn’t expect any such difference absent either discrimination or a real difference.
So, on the one hand we have the rather extraordinary proposition that women are unable to code, despite Tim’s favourite Steve Shirley example (among many others) showing precisely the opposite. On the other hand we have the entirely unremarkable proposition that techies can be rather prejudiced, even if not intentionally.
Yeah, it’s no surprise that one explanation is given more weight. That’s because it’s overwhelmingly more likely to be the case.
BiW>
I don’t know why you so utterly reject the common sense that says that the reason for the behaviour you observe is that only one gender (or rather, a small subset of it) was given any reason to dig into the guts of their computers.
BiND gets it spot on in the bits he quotes (despite your nutty rant about ‘the patriarchal’ nonsense): there was nothing being made which would attract even normal mainstream men, let alone women; the majority of the ‘attractions’ of eighties computers and software, such as they were, were attractive only to a small subset of geeky (mostly male) kids.
To the extent that we can generalise about men and women – which isn’t that far – we can say that men and women are both interested in what new tools can do, but it’s mostly only a few men who are interested in the new tools themselves. That’s where computers were in the 80s and 90s.
Dave,
I’m not rejecting that at all. I’m pointing out that the home computer changed who got into it. It was no longer a trade for exam-sitters who did as they were told, but for tinkerers.
You clearly aren’t a programmer because the “attractions” of writing code for a ZX81 are pretty similar to the attractions of writing code for a Windows server. No-one deliberately created computers in that way to help nerdy men.
And computers today are a lot more friendly to non-techies, but most of those people aren’t writing code with them. They’re consuming services or using them to do other creative things It’s still a nightmare getting girls into code, even though the IDEs are far easier today, and there is tons of government and industry money getting poured in.
“You clearly aren’t a programmer”
Well, you should tell that to the people who paid me to write code for them. (Actually, I’m kind of on your side on that one. I told them that too, they still insisted on getting me to do it.)
“the “attractions” of writing code for a ZX81 are pretty similar to the attractions of writing code for a Windows server. ”
Writing code for its own sake is not attractive, except to a very small percentage of the population. The ‘attractions’ are what would lead people to write code – finding things they want to fix, for example. If you’re not seeing anything there to get you on a computer in the first place, you’re not going to start coding.
I think it’s perfectly self-evident, given the take-up of computers among the general population in the last 15 years or so, that when people find something computery they want to use, they buy one. The internet is obviously a killer app in that sense, it’s attractive to almost everyone in one way or another. Before that, though, the things that existed on computers only appealed to a narrow segment of the population.
Does this really need explaining?
P.S.
“It’s still a nightmare getting girls into code”
No, it’s just hard to get them to work as programmers. Which says a lot about the environment they’re asked to work in – no normal person will find the working practices of the majority of IT businesses (in my wide experience of shit employers) tolerable.
For whatever reasons rooted in the past, the IT world has come to be dominated by arseholes. That has now become a self-perpetuating culture, because mostly they employ other arseholes, and many people who aren’t arseholes can’t be bothered with the shit when they could just get a normal job in a normal industry.
I moved up several pay grades the day I stopped tolerating arseholery and started insisting on being treated like the professional I was, rather than a naughty schoolboy. But having to fight that fight almost daily was needless stress. I retired from doing IT stuff and started doing something else stress-free for the same/more money.
Dave,
“For whatever reasons rooted in the past, the IT world has come to be dominated by arseholes. That has now become a self-perpetuating culture, because mostly they employ other arseholes, and many people who aren’t arseholes can’t be bothered with the shit when they could just get a normal job in a normal industry.”
No, it hasn’t. IT management used to be awful. I worked for project managers who had no idea about software, as in, the thing that needed to be shipped. These are people who insisted on me filling out timesheets and wear suits even though it was clear that it was purely for ceremony rather than useful to doing the job. Or they were pissheads who basically didn’t care, but as there was almost no real competition in old internet IT, nothing happened to them.
Management today are light years better than then.
“there was nothing being made which would attract even normal mainstream men, let alone women; the majority of the ‘attractions’ of eighties computers and software, such as they were, were attractive only to a small subset of geeky (mostly male) kids.”
Yes and those kids grew up into adults with an interest in computers. Of course a 35 yo in 1980 probably wasn’t going to be that interested in computers (but some were, they were often the ones running the shops selling the computer gear), but then older people are rarely the first into new tech.
I grew up in the 1980-1990 period, every boy I knew was into computers, had one. Most for playing games admittedly, not all were into programming them, though most will have had a go. Computers WERE mainstream for that age group, 100%. So there was absolutely no reason why girls shouldn’t have been equally interested, unless there was some biological difference between the genders in the degree of interest in such things. There was no cultural pressure against girls being into computing, it was a brand new scene.
You’re also mixing up two things – everyone (men and women equally) are interested in IT to the extent they can use it for something else (web browsing, sending emails, online shopping,going on Facebook etc), but only men by and large are interested in IT for and of itself, and end up tinkering under the hood. Rather in the way that while both men and women have cars equally for the job of getting from A to B, it tends to be men who tinker about with bits of oily metal in workshops building kit cars, and restoring old classics.
This can be sorted without offending anyone…
1) Assume that women coders are exactly as good as men with the same spread of ability.
2) Assume that Facebook pays well enough to be able to pick and choose from job seekers.
3) Note that 13% of Facebook coders are female.
4) [this is the big one] If more than 13% of coders in the market are female then the facebook female coders should on average be better than the males, because the female recruitment hurdle is higher for some reason.
5) [Likewise] If fewer than 13% of females in the market are female, then males faced higher recruitment hurdles to get into Facebook, and the males should be better on average.
6) See also Patten reforms to policing in Northern Ireland…
“It’s still a nightmare getting girls into code”
Why would they want to get into coding? If they’re so technically inclined they probably realise they’ll get a better job doing something else and if not there’s no point fighting against the current.
BiW>
I’ve had enough different jobs – hell, different careers – that I’ve seen plenty of different ways of working. In any big office/company – say, >50 – you’ll get at least one arsehole unless you’re very lucky. In any IT office/dept with more than three or four people, guaranteed at least one of them will be a weapons-grade arsehole, and one or more of them will have serious personal hygiene issues. And I only gave up on computery stuff about five years ago, so I doubt it’s changed that radically since then.
Seriously, I used to work with SWMBO for several different IT companies. The behaviour of a wide variety of cow-orkers beggared belief. The best of the bunch simply couldn’t speak to her without stammering, the worst would do things that were outright criminal – uninvited touching, she’s quite capable of defending herself, they felt much pain – or just obviously stupid, like swapping desks with me while I was out of the office, making up some computery reason, hoping that by sitting next to her they’d end up going home with her instead.
And just to be clear, we were introduced to them as a married couple. We’ve never had anything like that amount of trouble in any other field. IT is the only field where I’ve had to explain to multiple layers of management a) what sexual harassment was, and b) what the potential repercussions for their business were, before disciplinary action was taken over things that would get you barred from any pub.
NDReader>
“If more than 13% of coders in the market are female then the facebook female coders should on average be better than the males, because the female recruitment hurdle is higher for some reason.”
You’ve assumed your conclusion there.
Dave: “You’ve assumed your conclusion there.”
That’s why I gave item 5 as the alternative to 4. I’m offering up a diagnostic test, not a conclusion.
BiW,
” I worked for project managers who had no idea about software, as in, the thing that needed to be shipped. These are people who insisted on me filling out timesheets and wear suits even though it was clear that it was purely for ceremony rather than useful to doing the job. Or they were pissheads who basically didn’t care, but as there was almost no real competition in old internet IT, nothing happened to them.”
The project manager is a project manager is a project manager crowd who think that giving someone a Prince 2 course means they can manage any project in any discipline really winds me up. They think all they have to do is shout Gantt or PERT chart, demand a risk and issues register and getting a one page PowerPoint slide filled out every Friday is all that is needed.
The number of civil servants who thought that they were brilliant PMs by doing just that really wound me up. Most of the time they are the cause of projects slipping as people run round answering their damn fool requests and producing pointless reports and slides that no one understands and therefore they have to be explained to “Senior Managers”, wasting even more time, so that can walk around being knowledgeable and self important with each other.
It’s 2017. How personal computers were marketed in 1981 has absolutely fuck all to do with how many women do or do not want to write code now. I’d like to see the workings on that assumption, please.
“If more than 13% of coders in the market are female then the facebook female coders should on average be better than the males, because the female recruitment hurdle is higher for some reason”
I would assume there is Affirmative Action and all other things being equal. A woman has less competition for the same position as a man and is less likely to be as skilled at the start.