Sex differences
darrendahly.github.io
darrendahly.github.io
https://www.youtube.com/watch?v=9PaL5FW5src
The key takeaway for the Google Memo, is that James Damore was fired for referencing mainstream science, then asking if that was relevant to Google's hiring practices. What's more, many of his coworkers and many of my fellow college classmates were more than willing to throw him under the bus and impute all sorts of unsavory thoughts and emotions to Mr. Damore with no evidence, while the Google management was willing to publicly do the same while re-disseminating his memo in the worst way possible to encourage misreadings of it. See also Christina Hoff Summers' video on the subject.
https://www.youtube.com/watch?v=tu4tB9W3xFo
Steven Pinker identifies as a feminist, as does Christina Hoff Summers. Diane Halpern certainly isn't a misogynist.
The extreme left which ascribes to Postmodernist political theories is coming for the scientists, rationalist nerds, and Aspergers neuro-atypicals. They want to cow everyone into agreeing to their ideologically driven Lysenkoist-like censored science. They represent a new authoritarianism operating through a fear of social ostracism and economic disenfranchisement. (You don't have to take my word for it -- you can take theirs. Just pay attention to what campus activists have been saying since about 2014.) If you believe in rational discourse, free speech, and meritocracy, you need to courageously speak out against these authoritarians.
(First they came for the social scientists and evolutionary biologists...)
I have generally high regard for your comments. But in this instance, I think your claim about what happened to Damore is facially flawed. Damore didn't simply "reference mainstream science". Whether you agree with him or not, I don't think it's possible to defend the claim that he was fired for citations.
I'd also urge you to step back from the Damore framing of this post. I know why it's hard to escape. But I think the author is just using Damore as an opportunity to make a more general statistical point. Certainly, at no point in this post does the author even mention Damore by name, let alone advocate for his termination.
What I say it establishes. That James Damore was referencing mainstream science. One might reasonably conclude that his conclusions are wrong. Here is an example: https://www.youtube.com/watch?v=BIRvtA2JIIA
I think all of those who concluded that he was being a misogynist have no leg to stand on. The accusation is levied just to silence views that are not liked.
I'd also urge you to step back from the Damore framing of this post. I know why it's hard to escape. But I think the author is just using Damore as an opportunity to make a more general statistical point.
I don't have a problem with the article's statistical point. I think there are all sorts of opportunities to make points here.
Here, though, I just see a professional biostatistician making a basic point about how genetically-influenced traits do and don't influence real-world outcomes. It may seem basic, but if it is, it's not obvious to the typical HN thread!
Later edit
I mean, doesn't it seem apparent just from the responses on this thread?
Not at all. Also, for the record, as far as I can tell, James Damore is not a member of the Alt Right. (I identify as a Center Left Liberal.) I have noted that mainstream news people have tried to stick that on him.
If your argument involves hazily imputing motives without evidence, or if you are using a "No true Scotsman" fallacy involving his position -- then I'm afraid you're just using a fallacy.
If you watch Steven Pinker's presentation, you will see a striking resemblance to the science cited in the Damore memo. Until recent years, there was nothing pointedly ideological about the mainstream science referenced in either place. I would urge you to read Pinker's _The Blank Slate_ for a scientist's take on the pointedly ideological and unsupported absolutist doctrine of the human Blank Slate.
I strenuously agree with your overarching point. But I have a more charitable perspective about the accusations of misogyny. I think that many, many women have had awful and in many cases traumatizing experiences in the world of tech. A lot of people associate opinions like Damore's with the bad behavior they have seen and experienced. They see these kinds of arguments as being part of the same phenomenon; men who hold opinions like these are the same kind of men who would talk over women in meetings, dismiss the achievements of women, be less likely to promote women, pay them less, harass them, etc.
I think a lot of people (women especially) felt an extreme and visceral reaction to the memo, because they intuitively made this association that these sentiments were related to their bad experiences. And while I strongly believe that the accusations of misogyny were unfair and misguided, I interpret them charitably as an attempt to push back against what they interpret as sexist behavior. They are trying to support the women around them.
You don't get a pass just because your emotions ran over your rationalism.
Exactly this. Both sides of the political spectrum could use a bit of this thinking in my opinion.
I don't necessarily believe he's a misogynist, but I do think he's in denial about the impact of sexism on gender gaps. To quote his letter:
"We need to stop assuming that gender gaps imply sexism."
The fact is that we don't have to assume. It's fairly unlikely such gaps are entirely the result of sexism. But to suggest that sexism does not play a role in gender gaps is absurd.
Edit: To respond to similar points in the replies, yes, I'm quite aware that you can read this sentence in the strictest logical sense as saying a gender gap, in and of itself, does not imply sexism. This is obviously and trivially true. But studies on evaluatory bias tell us that biases, such as and including sexism, are in fact the norm. It is so fantastically unlikely that they are not present among Google's interviewers / hiring committees that the sentence, taken in the strictest sense, is making an irrelevant point.
Rhetorically, however, it serves the effect of suggesting we ought to consider it plausible that perhaps sexism isn't at all at play in Google's hiring. Which, again, is absurd.[1] It would be far more likely that Google's hiring tilts sexist against men than sexism not being a factor at all.
[1] Note that this has nothing to do with Google specifically, but rather with humans generally.
It is not a statement about a specific case, but a general point on rationality.
And it is undeniably true in the general sense.
This means that if someone wants to convince us there is sexism in a specific case, they must demonstrate that specifically, not just note a lopsided gender ratio and assume the mechanism.
To me, part of the trickiness of the entire "equality" discussions is that we don't have clear, "correct" distributions of variable X for demographic Y.
Pure egalitarianism is iffy, because given any distribution, we can manufacture a demographic that is "underrepresented", so it's not at all clear that flat equality across "reasonable" demographics is even logically sound.
I kind of piggy backed off your comment there and ran in a bit of a different direction. Hope you don't mind.
I have written before that I would love to see statistical data that can predict gender segregated as a factor based on the number of reported cases of sexism.
What I have seen is that the causes of segregation differ massively based on if it is male or female dominated. If it is male dominated profession, sexism is always and early included as the suggested cause. If its female dominated it is rarely included. There seems to be a strong sex difference in studies studying gender segregation, so the thing I find most interesting is the few areas where they do converge, which is:
a) Work culture, which tend to incorporate gender identity in seemingly similar way in both heavily male dominated and female dominated professions.
b) Social pressure, especially for students, by peers of the same gender and age.
What I would like to see is meta studies on gender segregation that can incorporate a large number of different professions and make a unified theory to explain why only 12.5% (data from Sweden) of the population work in a profession which is not gender segregated (defined as having less than 60% of a single gender).
I think his public behavior since his firing allows me to reasonably draw conclusions as to what motivated his "citations" of "mainstream science".
Facts don't fit your narrative. The percentage of female software engineers at Google matches the percentage of female software engineers that graduate with CS degrees.
If sexism is driving women away from CS, then it's driving them away way earlier than when they apply for jobs at Google.
That said, Damore did himself and everybody else no favors by writing the document in the style he did. A much simpler memo, with a simple thesis, would have been much more effective.
Any principle of charity one could have had towards Damore was soundly negated by his actions afterwards.
---
"The pseudo-scientific ideas of Lysenkoism assumed the heritability of acquired characteristics. Lysenko's theory rejected Mendelian inheritance and the concept of the "gene"; it departed from Darwinian evolutionary theory by rejecting natural selection."
Had to look up Lysenkoism. Thanks for the new pejorative. I've added it to my filter bubble.
Okay, so upon hearing about a major chapter in the history of science, where ideology crippled science in a major world power for many years, the best you can do is to say, "I'll add it to my filter bubble!?" I'm sorry, but that's politicized anti-intellectualism of the highest degree! Shouldn't we be trying to learn from history, not denying it?)
Lysenko primarily held his theory because it fit Marxist ideology. The Blank Slate is the new Lysenkoism.
Any principle of charity one could have had towards Damore was soundly negated by his actions afterwards.
Which actions are those? In a situation like that, why wouldn't one sue? In a situation like that, why wouldn't one appear on news outlets that would actually let you speak, and not try to turn your interview into a hit piece?
The tone of this introduction doesn't inspire confidence that the author is truth-seeking without agenda.
> Remember, a 1 SD shift is a pretty big one (e.g. a difference of 3 inches for height), so not suprisingly, the evidence seems to suggest that the 10% increase in the mean is the more realistic scenario for most variables where differences seem to exist.
Several times he references the Stevens/Haidt results, but doesn't seem to acknowledge this finding from Stevens/Haidt (https://heterodoxacademy.org/2017/08/10/the-google-memo-what...):
> Gender differences in interest and enjoyment of math, coding, and highly “systemizing” activities are large. The difference on traits related to preferences for “people vs. things” is found consistently and is very large, with some effect sizes exceeding 1.0. (See especially the meta-analyses by Su and her colleagues, and also see this review paper by Ceci & Williams, 2015).
Excellent citation. I think it overreaches a little in attributing the position that women don't perform as well as men at software engineering, ie. they label many claims red which people have attributed to Damore, but he didn't make those claims; I believe Damore said the exact opposite actually, that women absolutely can perform at the same level, but that women simply aren't interested in CS careers. Your cite covers this in the conclusions, so it's pretty fair overall.
As your citation explains, in societies with greater gender equality, women have so many other choice of careers that CS just isn't an attractive option. Perhaps it's not attractive for reasons that can be changed, but those reasons should be investigated, not simply assumed to be driven by sexism.
IMO, the latter is mostly true. Less males are attending Western Universities than females, yet there is absolutely no push what-so-ever to fix this. Nobody cares about the lack of men in subjects such as Art or English Literature and certainly no initiative exists to even begin to correct this. Nobody is trying to push women into the lower 5% of jobs. Brick layers, plumbers, electricians - all well paying jobs - where is the push to get women into those?
There are fundamental differences in the sexes which favour men for some jobs. For example, millions of years of evolution means that when you take a random pack of dogs, a group of apes or a gathering of humans and allow them to discover natural social order - almost always the leader will be male. Whether we want to accept it or not, we are still just highly functioning animals.
Probably the most concerning, despicable, disgusting act of them all is a "recognition of women in... award" (rarely recognition of men of course), or the manipulation of wages to equal the "pay gap", or biasing employment, etc. To me, it says "we know women can't compete, therefore we're biasing the end result to make the outcome equal". It's like first-wave feminism meant nothing, what was the fight for equal opportunity even about if you only go on to bias towards equal outcome anyway?
The point: Equal opportunity > Equal outcome
Because these jobs don't address the power imbalance between men and women in our society. Lawyers and doctors are highly respected careers that influence the direction of society. Lots of women pushed into this field. Politics also needs more women, as should be obvious given how much states and Congress tries to control women's health care options.
Computer science is increasingly pervading every facet of our existence, and so having gender parity ensures female concerns are equally represented as this field evolves.
The "power imbalance" should be represented in all areas, surely? You can't just have the icing on the cake. And I think you've hit the nail on the head, this isn't about equality, this is about control, particularly money and power.
>Lawyers and doctors are highly respected careers that influence the direction of society.
And have begun to dominate these fields now (although I can't find a source of that I would be happy to quote).
>Lots of women pushed into this field.
I'm not sure there has been a significant struggle for a long time.
>Politics also needs more women, as should be obvious given how much states and Congress tries to control women's health care options.
I wouldn't say the main problem with Congress is that there is a lack of women, I think it's very existence is it's largest issue.
>having gender parity ensures female concerns are equally represented as this field evolves
"female concerns" sounds like a terribly generic term, you'll have to be more specific and explain how these concerns override the need for women representation (not necessarily women by the way, just a person who fairly represents them) in politics. My thinking is that tech for the most part is completely genderless, it's simply the application of Science. Most control is tech can be or has been completely overridden by law.
But in general, I think (without being rude and wanting a genuine reply) that you've missed my point with regards to the lower 5% of jobs. My point is that instead of making STEM fields artificially more difficult to get into for men, another option is to make other fields more attractive. Social care for example is horrendously dominated by women, yet of massive importance to society - they are struggling to recruit due to social stigma, yet little is being done to address this.
I guess that's the disagreement I have with people is that it's hard for me to see engineering as open and meritocratic when I've seen situations where it definitely is not (even if those situations are just anecdotes)
That said, the completely incorrect way of handling this is to bias the outcomes, a.k.a purposely hiring genders (or any group) to meet quotas, balancing pay checks despite job descriptions, etc. That just puts incentives for unhappy people to remain where they are, nobody wins that way.
I would want to mention that the definition of leader don't perfectly match between alpha males in animals and human leaders.
I recall this troop of baboons. As usually the alpha male leaders the group in the morning to the feeding ground, but a new male had recently joined and won the alpha spot and he didn't know where it was. The flock followed for a while, but at a point the older female baboons simply decided to break to the right place and the rest of the flock followed. The alpha continued walking forward, oblivious that the flock behind him had gone elsewhere and was quite surprised when he finally noticed that the flock was gone.
There were no punishment for the older female baboons. No fight when they stopped following the alpha. The alpha is permitted to lead, but only to a point.
Of course not, but it's "perfect" enough to be evolutionary advantageous.
>There were no punishment for the older female baboons. No fight when they stopped following the alpha. The alpha is permitted to lead, but only to a point.
Of course in not all cases will the alpha be the best leader, I won't even remotely argue that.
But I've seen time and time again, the best leaders are often the most bold. Sometimes it's simply not clear what decision should be made and instead of allowing that indecisiveness to spread throughout the group, a decision is quickly made for better or worse and it's confidently executed. Your better leaders will be able to make better decisions from incomplete data that maximize opportunity and reduce risk.
There are a few takeaways from your anecdote:
1. Leader doesn't always know what leader is doing, but does it anyway. Sometimes looking the part is more important than being the part.
2. Better a heard of apes lost than a feeding ape alone - predators be watching.
3. Pull on the skills of a larger group - the old and wise female baboons for their apparent localization, the male baboons for fighting off a predator, the females for directly looking after the interests young, etc.
The people fulfilling the roles may change, but the roles themselves remain.
I don't think it's that females aren't good at or don't have natural abilities equal to men when it comes to programming but for what ever reason (that I don't care to speculate on) programming is not an attractive career option to many of them.
So I think this point is a little off
>>At this point, anyone already familiar with the normal distribution is probably having a sensible chuckle. This is because while Google might be selective in their hiring practices, they aren’t +4 SD selective.
Because there is just a much smaller number (at least in my experience) of female developers. So if one company is 50-50 male to female they would have a disproportionate number of the total population of female developers.
This isn't a google problem, no need to talk about variances.
http://splinternews.com/survey-says-92-percent-of-software-d...
http://www.exploringcs.org/resources/cs-statistics (scroll down to 2013)
Whatever the cause, it seems like we should look upstream and leave Google out of it.
This says it's an international number but it also puts the % of female CS grads in the U.S. at 12%.
Other ways it might be off include that women who get CS degrees enter non-development positions at higher rates(Project managers, business analysts, etc) and drop out of the industry at higher rates(I've heard).
https://blog.hackerrank.com/which-countries-have-the-most-sk...
No it doesn't. "Mean" isn't used in this way even once in the whole memo. The only indirect reference to means differences are two sample graphs that Damore only uses to explicitly label judgement based on means as fundamentally incorrect because it ignores significant overlap between groups. The greater male variability hypothesis is the only correct way to interpret the memo.
Further, this article's conclusion doesn't follow from the facts as presented. It's literally asserting, "I can't fathom the existence a gender-specific difference of 4SD, therefore it doesn't exist and so such a difference can't explain our observations". Didn't you miss a step where you have to actually demonstrate that this difference actually doesn't exist before you can dismiss it?
Certainly we can debate how reasonable such a distribution might be given what we know about gender-specific variability of other traits, but we absolutely can't dismiss it out of hand.
Consider, for instance, that "hiring at Google" isn't the first time this variability comes into play, but instead the Nth time, where N is fairly large. If people tend to pursue things they're naturally good at, and such variability will obviously influence what genders will be naturally inclined to pursue, on average, then this acts as a compounding filter over a decade where students are first exposed to computer science.
Consider the distribution characterizing "people's net worth". The variability in people's ability to make and save money isn't that broad, but including compounding effects of interest on debt and investments, even these small differences will yield huge variability in your net worth on your death bed.
The average male is about 6 inches taller than the average female. If a >1SD difference is possible in height, why not other areas? Or am I missing something here?
> Starting again from the standard normal, adding 0.10 SD to the male mean would lead to a 1.2 to 1 male to female ratio (at ≥1.5SD); but adding a full SD would give almost 5:1 [emphasis mine]
If you list jobs that have large social components like teacher, social work, etc... more women will say they sound pleasurable than men.
When you list jobs with smaller social components like carpenter, electronics technician, forester, car mechanic more men will say they are pleasurable.
Is this difference entirely cultural, women being told they should prefer people jobs, and men being told they should prefer thing jobs? No entirely.
It looks like biology plays a significant role too. Women(XY) with congenital adrenal hyperplasia are raised as women, but tend to show occupational choices that are less people oriented. And the degree to which these women have been exposed to prenatal testosterone predicts how much they'll prefer thing oriented jobs.
Just to reiterate though, I don't think this is a valid reason to give up on getting more women into tech. It just means it'll be an uphill battle, and maybe we should expand our toolset to include things we otherwise wouldn't have tired.
When I think about how we get to a profession with much better gender parity, I imagine the world that has changed how it sees developers. Right now the world sees great developers as technical geniuses sitting in a dark basement staring at a command line solving garbage collection bugs and doing memory dumps. But instead we thought about a great developer as someone who used their eloquent communication skills, amazing empathy and a generous helping bad ass technical skills to delight their end-users with features they didn't even know they needed. I think that's a job more 16 year old girls could get excited about.
P.S. this doesn't mean we shouldn't stamp out sexism where ever it rears its ugly head.
I'm highly suspicious of any claim that biology affects typically gendered jobs, whether because of suitability or just because of preference, because it seems like gender roles in employment ought to be a lot more consistent if that were true.
The history of computing is soaked in historical context.
And I don't buy this explanation about the war. Women were significantly more present in computing as late as the 1980s. For example, the percentage of women majoring in computer science was over 35% in 1985, versus under 20% today.
Effects from wars are neither eternal nor instant. In economy people are still saying that we are seeing ripples from the world war 2, but at some point those ripple effects will be too small to be relevant. It is relevant to explain the economy during the 1960, somewhat for 1980, but less so for 2017.
To claim that a profession born out of war isn't effect by war politics seems odd.
And the defense that the larger number of women in CS in the 80s was a long effect from the war still doesn't make sense, given that the number had been going up for quite some time at that point.
There was one interesting thing happening 1975 however. The US stopped sending millions of young men away from the work force and fight in Vietnam, often to never return. After 1975 the school system also had a massive intake of male students, both old and new. Young men was not sent to die, and old men who survived the forced military draft was given an opportunity to study.
Maybe a coincident, but that match quite well the census data. When there was wars, women had a upwards trend in the computing industry. When there was peace there is a slight delay for education to catch up and then more men enters the work force.
I don't see how the end of US involvement in the Vietnam War would explain the percentage of female CS majors peaking a decade later.
Why are you so insistent that there must be some historical aspect behind the numbers of the past, but not the numbers of the present? What makes you think that the way things are today is the "natural" state of things, when it was never the case before?
Individuals with congenital adrenal hyperplasia look exactly like women and were raised as women.(think Jamie Lee Curtis) but they are genetically male. And they found that these individuals had significantly more male patterned likes and dislikes. This is evidence that at least half of the preference is biological.
How does the hypothesis that women are biologically drawn to certain types of jobs, regardless of stereotyped roles, explain the fact that this stuff constantly changes?
https://www.bls.gov/cps/cpsaat18.htm
Logging 3.2% Female
Child day care services 94.4% Female
Being a software developer at all is ~2.5% of the US population (and even less internationally).
Moreover, if the variability hypothesis is true then the entire point is that 1SD for men will be larger in absolute terms than 1SD for women.
Second, what you're pointing out is orthogonal to the question. The premise of the question is that there aren't many women software developers; the dispute between the memo and Dahly is why that is.
Third, if you bring international labor into this, you muddy the issue even more, because female participation in the software development market is sharply higher in some major overseas markets.
I don't follow your second sentence.
~2.5% is the percentage of the US employed workforce.
> Second, what you're pointing out is orthogonal to the question. The premise of the question is that there aren't many women software developers; the dispute between the memo and Dahly is why that is.
The argument is that Google would not be excluding >95% of people, but 97.5% aren't even software developers.
> Third, if you bring international labor into this, you muddy the issue even more, because female participation in the software development market is sharply higher in some major overseas markets.
But why would a US employer hire foreign employees from those markets and not the world market overall?
> I don't follow your second sentence.
The complaint is that four standard deviations is a lot, but it's not actually four because the size of a standard deviation is not the same.
Taleb's book The Black Swan talks about this phenomenon, but I can't recommend the book--it's basically one chapter's worth of material stretched over 400 pages.
It sounds like you're saying that it's more important to know the identity of the person making the argument than the content of the argument itself?
I don't think the tail of the techiness distribution would be an issue either, because this tail isn't that fat. Techiness is most likely constrained to a definite variance. I mean, Googlers are part of that tail, yet none of them are some sort of post-singularitarian cyborg, beyond the comprehension of their own colleagues. It might sound stupid, but this is the sort of thing we expect to see in the fat-tailed distributions.
That's a really strong assumption, and one that even the author of the "Diversity memo" didn't make, IIRC. In fact it would make sense to assume the exact opposite: Whatever the differences of "initial state" may be, they will compound over time in a "feedback loop".
Another issue: "techiness" isn't the only or even the main factor considered when hiring employees. If you consider this vaguely defined variable a function of several other variables, some not related to technology at all (e.g. cooperation, willingness to work at a major corporation, to work long hours, alma mater, other job opportunities, etc) you're left with basically no way of estimating if the standard deviation differences between males and females are, say, 1.5x or 4x.
The author also didn't combine simultaneously changes in mean and standard deviation on his analysis, and didn't consider the possibility that the traits might fit a different curve (e.g. Tracy-Widom)
Damore's hypothesis is not about ability or qualification. It's about interest. The paper claims that women are underrepresented in the tech sector simply because they're less interested in tech than men, not because they're less qualified or capable.
I disagree with that hypothesis, sure (quite a bit of anecdotal evidence among coworkers and family/friends actually suggests the opposite, and I don't think Damore adequately isolated supposed biological factors from potential socioeconomic factors), but I at least bothered to read the paper before blabbing about it on the Internet.
Does anyone have any idea of what the selectivity at google hiring is? How many people do they reject for each person they hire? Also using this number would have to roughly take into account that the distributions are supposed to be "over all people" whereas the hiring pool is self-selected. Simply by being able to get python to print "hello world" you're probably already in the top 15% of "all people".
I am also curious about modeling this where the median of population A is LOWER than population B but variance differences and selectivity pushes selection for population A higher. (to wit: I think, and the scientific evidence supports me, that men are generally dumber than women)
and given that the programming is a pretty dumb job, a blue-collar job of the 21st century ('S' in CS isn't a real science), i think we have the explanation.
That is such a crude way to go about things.
It's been suggested before that there are larger influences from society on how men and women should act and live their lives. I still believe that. I don't think we've made a lot of progress on gender issues. A lot of them still lurk under the surface and the PC culture shields them and prevents discovery and resolution.
Except it's only implausible if you assume the conclusion. The implausibility is literally summoned from the aether, because it's certainly not based on any empirical facts.
Maybe there is something I don't understand but 4 sigma is .006334% so > 4SD would be 0.003167%.
"This percentage equates to just 273 New Yorkers, 30 people from San Francisco, or just 11 thousand people in the entire United States, the 3rd most populous country on earth."
0.000032% of 320 million would be 102 people.
It would make more sense to look at average companies that don't have the resources to hire from the right side of the normal distribution. If they have a 4 to 1 ratio where a 1.5 to 1 ratio was expected that means the differing distribution is one of the less significant possible factors out of many.