Honestly, this is adding to the problem. We need to be concentrating on hiring the best person for the job. If this ends up being mostly men, non-minorities, etc, then there are other problems that need to be fixed.
Ie: why don't they have the skill sets to get the job?
This reminds me of what many police and fire have been doing to bring in more women: reduce the standards on the physical tests...which just increases the danger for everyone.
Awesome to see a big name company set specific goals for gender and racial diversity. I can't wait for the end of the "almost every meeting is all white dudes" era.
Isn't this wrong at some level? I support the idea of promoting programs like Women In Engg., but hiring based on race/gender is so wrong. Do you lower your hiring bar to get the numbers you've aimed for? And assuming some of these employees turn out to be under-performing, how do you deal with the appraisal process?
I'm not sure if Pinterest posted this article with some other intent (like having more support programs), but in it's current state, it seems discriminatory.
I never said that they will. I'm just wondering how are they gonna do things differently to meet those numbers. I'm sure their current hiring process isn't biased towards a particular race/gender. To make it more balanced, are they planning to have a different hiring bar?
If it is similar to the Rooney Rule, no. The bar is the same, just required to consider a minority candidate.
I personally find tangible monetary value in employees with diverse backgrounds but I don't lower my bar. I just slightly discount sameness and slightly appreciate difference.
@infosample, you are right. You cannot assume minorities and women will under-perform or are under the bar. The issue of people just don't give it enough thought when hiring. At the same time, you should not hire someone solely based on their gender or race. But if there are 2 candidates with very similar skill sets and meet the standards you are looking for, hire the one who will bring you more diverse background.
This is why the discussion can't even happen, and why companies and universities still enforce percentages for underprivileged groups even though it is obvious discrimination (but it's "good" discrimination! Give me a break).
Person 1 says that enforcing a 50 people of type A and 50 people of type B when the number of qualified candidates is 20 and 80 will result in either hiring significantly fewer people in general, or hiring under qualified people of type A. People of type B are being passed over because of their race/gender/ethnicity. This is discrimination.
Person 2 says that implying that anyone of type A could be under qualified is itself discrimination, and this possibility should not be addressed or considered. Nothing can change because the issue cannot be discussed.
It helps to have a discussion when affirmative action and the Rooney Rule aren't conflated.
We have a tendency to assume type A is qualified and type B isn't. That is a natural bias based on evidence. We must actively subdue our biases if we want to overcome them, sometimes with rules.
I think we sometimes forget that the goal isn't to get a qualified employee but a successful employee. Qualifications generally leads to success but not always, so we should constantly question our qualification criteria.
White males have benefited from discrimination for a long time.
White males have been discriminated against sometimes.
This article isn't about discrimination or affirmative action.
It's about setting a goal to be more diverse, not by lowering the bar, or forced hiring, but through awareness, inclusion, and an opportunity to interview.
BTW, everyone's been given a chance they didn't deserve.
And it would be discriminatory to give the job to the women purely based on her gonads in such a scenario. Benevolent sexism is still sexism, but it seems people only care about it when it comes to holding doors rather than holding jobs.
Where I work, if we had 9 men and 1 woman of equal skill applying for a job the woman would get it hands down. I find most good sized companies would rather have that kind of diversity even when it's not official policy. Our team has a few women but still well less than 50%.
For a site whose chief technical accomplishment is to allow women to pin and share cutesy images, I'm not sure having top talent really matters that much.
I'd rather see more top women doctors, scientists, lawyers, civi, biomedical and chemical engineers. Not sure what having more women sit in front of a computer screen writing CRUD apps all day does for equality.
Yes, obvious trolling by a Nazi username gets downvoted. Yay us.
I still say they don't represent the technical community. And it's blatantly obvious they don't represent the HN community, given the downvotes.
I'm not saying there aren't biases, sexism, ageism, or racism in tech, but I'd appreciate it if you didn't point at the obvious troll and say "this right here is the face of tech".
Really? Pinterest is ranked #16 by Alexa in the US for internet traffic[1]. Do you really think running such a site with that much traffic doesn't require top talent?
It's a solved problem in the sense that the 5% of work that yields 95% of benefits have already been done. But because of the massive user base small improvements at great cost can be worthwhile.
I hope that you're just a troll and not serious. Pinterest is one of the most visited sites in the world. It requires huge technical talent to handle that much traffic and data. WTF is wrong with you?
This very common assumption -- which I used to hold -- has it exactly backwards. Let's assume tall people and short people are equally good at cracking eggs. Mathematically, if 90% of your egg-crackers are short, you've already hired under-performing egg-crackers. You've passed over some really good tall people in favor of the short people you hired. If you want the best egg-crackers, you need to interview a bunch of tall people.
This analysis has its own flaw. This assumes that their other opportunities are also equal. But tall people may make better apple pickers. And apple picking pays more than egg cracking. If you want 50% tall people, you are going to need to compete directly against the apple picking market as well. You need to account not just for your own selection criteria, but that of everybody else as well. (Not to say this actually explains tech demographics; it's a point about the stats in your example.)
> Let's assume tall people and short people are equally good at cracking eggs. Mathematically, if 90% of your egg-crackers are short, you've already hired under-performing egg-crackers.
In fact, this depends on the underlying population sizes of tall vs. short people, so this is a faulty argument.
More germane to the situation, however, is your initial assumption, whose analogue is dubious. So much so, in fact, that it is taboo to even allude to it. cf. pg's "What you can't say".
If 90% of all egg-crackers are short, and you want to hire the best egg-crackers, it's very likely that 90% of your egg crackers are going to be short.
There actually are things driving men out of the field. Look at how many head nurses are male. Men in nursing is a problem as well. But this isn't NurseNews, it's HackerNews.
The counter argument that I've heard most is that minorities are less likely to be hired than equally qualified members of the majority because of subconscious bias. Enforcing a quota is a sort of equalizer—formal discrimination in one direction in order to push back against informal discrimination in the other direction.
I'm not completely sold on the method, but the problem of subconscious bias in hiring is very well documented:
* Resumes with stereotypically White names receive 50% more callbacks than identical resumes with traditionally black ones. The addition of honors and special skills had a significant effect on the likelihood of White applications being called, but a statistically insignificant effect for the otherwise-identical African-American ones. http://www.nber.org/papers/w9873.pdf
* Initial subconscious bias is logically justified post-hoc—in the linked experiment, the qualities deemed necessary for a position would actually be redefined by the person reviewing resumes depending on the gender of the applicants so that the male applicant's traits fit the position better. http://www.socialjudgments.com/docs/Uhlmann%20and%20Cohen%20...
There's quite a few more papers on the subject, it's easy to search for them. Depressing read all around.
If those studies are true, then the correct solution would be to:
1) ask people not include their names on their resume. Or don't use resumes at all. Create a web form application that does not require a name and only asks about the crucial skills you actually need.
I don't really buy that subconscious bias is the problem in tech. I've been part of a lot of hiring, we were actually consciously biased toward hiring women since we wanted better numbers. I had an objective interview process, always giving the same questions, and the hiring was still very disproportionately male whasian (the applicant pool was also very disproportionately male whasian, despite doing active reach out to woman-in-tech events).
It sounds like you were trying to correct for an assumed subconscious bias of "deep down, we really want to hire men." In my experience, though, decent people like you don't have that bias. What they do have is a subconscious idea that "good candidates will look like X," where X is a mental image of a male candidate. Both women and men are subject to this bias, and it's a lot harder to correct for. Simply asking the same set of questions to all candidates won't fix it, if the questions themselves have the effect of weeding out women.
Simply asking the same set of questions to all candidates won't fix it, if the questions themselves have the effect of weeding out women.
The questions I asked were representative of things we actually had to do on the job, for instance: "write code to spider a web site." Does that unfairly weed out woman?
1. I agree, this would be preferable and a great step for the initial screener. But that leaves the problem of the actual interview—the studies that I've seen test written resumes, because it's possible to ensure that two resumes are identical except for one variable in ways that you can't ensure that two interviews are identical. But it's highly likely that the bias would carry over to any interaction with the candidates, though I don't have any studies on hand that prove it.
2. Also agree that an objective method would be preferable, thanks for the link. Great read. IIRC, in the study I link in the third bullet point, forcing the person reviewing the resume to create a rubric before reading any of them greatly reduced the biased results. Having an objective process goes a long way. The point in the link about confident interviewers performing better in the traditional hiring process is definitely true, and also tied in a roundabout way to gender, funnily enough (women tend to be less confident than men: http://www.theatlantic.com/features/archive/2014/04/the-conf...).
It's definitely not the only problem—the pipeline is obviously skewed dramatically—but I don't see any reason why tech workers would be exempt from the biases of the general population as they've been studied. Maybe the drive to hire more women would counteract it, like in your example—that's the reasoning under which minority quotas are implemented, after all. The problem of subconscious bias is that it's not consciously recognizable by its very nature. It's just gut feeling. I know a site that uses association to try to test it for you, if you're interested: http://implicit.harvard.edu/.
The fact that you immediately equated "diversity" to "lowering your hiring bar" and "under-performing" is as ignorant as it is insulting.
Women and brown people are not lower quality humans, and it has been shown over and over and over that companies who hire for diversity perform better.
I think ignorance is taking things out of context unnecessarily. I am assuming Pinterest already has an unbiased hiring bar. Yet, due to sheer probability, the number of other races get hired much more. Considering the number of applicants remain the same, how do you balance the race/gender without being biased to one race/cast?
Bar charts should be sorted with the biggest ones at the top. It looks wierd the way it is. And why do all of the Latino parts have a "@" next to them when a "*" is used to reference?
I think that it may be a way of saying "Latino/Latina" that is used sometimes in Spanish. As Latino is the male form and Latina the female one, sometimes a "@" (which is sort of an a and an o) is used to signify inclusion.
Maybe because I'm not American the part that confuses me is the "Hispanic" vs "Latino", are they used to mean different things?
Technically (very technically), Latin-American would include people talking French or Portuguese, while I guess Hispanic will only include Spanish speakers, e.g. not including Brazil. But I don't know if they are used in a different way in the US.
> Bar charts should be sorted with the biggest ones at the top
That would greatly reduce the usability of the charts in this case. The charts are much more useful when they all present the categories in the same order.
This seems like it's pursuing skin-deep diversity. Diversity of thought is much more interesting to me, and probably much more apt to drive profits than this.
Based on their charts, Asians are massively overrepresented and whites are underrepresented compared to US population. So I am curious what they mean when they say "underrepresented", will that include more Asians? If they meant black and Latino, they should just say so instead of vague euphemisms.
Maybe for Pinterest, a sum of diversity is greater than a sum of talent.
The possibility of that being true is certainly unsettling for me because I've always assumed that hiring by merit is intrinsically superior to hiring by quota.
From Steven Pinker's The Blank Slate: The Modern Denial of Human Nature:
-------------------------
But of course the minds of men and women are not identical, and recent reviews of sex differences have converged on some reliable differences. Sometimes the differences are large, with only slight overlap in the bell curves.
...
With some other traits the differences are small on average but can be large at the extremes. That happens for two reasons. When two bell curves partly overlap, the farther out along the tail you go, the larger the discrepancies between the groups. For example, men on average are taller than women, and the discrepancy is greater for more extreme values. At a height of five foot ten, men outnumber women by a ratio of thirty to one; at a height of six feet, men outnumber women by a ratio of two thousand to one. Also, confirming an expectation from evolutionary psychology, for many traits the bell curve for males is flatter and wider than the curve for females. That is, there are proportionally more males at the extremes. Along the left tail of the curve, one finds that boys are far more likely to be dyslexic, learning disabled, attention deficient, emotionally disturbed, and mentally retarded (at least for some types of retardation).
At the right tail, one finds that in a sample of talented students who score above 700 (out of 800) on the mathematics section of the Scholastic Assessment Test, boys outnumber girls by thirteen to one, even though the scores of boys and girls are similar within the bulk of the curve [NOTE this was from the pre-1994 SAT where the math section was harder and not truncated at the top. Today the ratio is a bit less than 2-1 at the top end.]. With still other traits, the average values for the two sexes differ by smaller amounts and in different directions for different traits. Though men, on average, are better at mentally rotating objects and maps, women are better at remembering landmarks and the positions of objects. Men are better throwers; women are more dexterous. Men are better at solving mathematical word problems, women at mathematical calculation. Women are more sensitive to sounds and smells, have better depth perception, match shapes faster, and are much better at reading facial expressions and body language. Women are better spellers, retrieve words more fluently, and have a better memory for verbal material.
...
Nonetheless, discussions of the leaky pipeline in science rarely even mention an alternative to the theory of barriers and bias. One of the rare exceptions was a sidebar to a 2000 story in Science, which quoted from a presentation at the National Academy of Engineering by the social scientist Patti Hausman: "The question of why more women don’t choose careers in engineering has a rather obvious answer: Because they don’t want to. Wherever you go, you will find females far less likely than males to see what is so fascinating about ohms, carburetors, or quarks. Reinventing the curriculum will not make me more interested in learning how my dishwasher works."
An eminent woman engineer in the audience immediately denounced her analysis as “pseudoscience.” But Linda Gottfredson, an expert in the literature on vocational preferences, pointed out that Hausman had the data on her side: “On average, women are more interested in dealing with people and men with things.” Vocational tests also show that boys are more interested in “realistic,” “theoretical,” and “investigative” pursuits, and girls more interested in “artistic” and “social” pursuits.
...
The most dramatic example comes from an analysis by David Lubinski and Camilla Benbow of a sample of mathematically precocious seventh-graders selected in a nationwide talent search. The teenagers were born during the second wave of feminism, were encouraged by their parents to develop their talents (all were sent to summer programs in math and science), and were fully aware of their ability to achieve. But the gifted girls told the research...
1. Their approach seems like it will work, if it does, mostly by increasing the chances that women engineers or women engineering students will choose to work for Pinterest instead of work for someone else.
2. I've frequently read that there is a shortage of tech workers, and companies like Pinterest have trouble finding the people they need.
3. Putting this together, does this mean that much of any increase in diversity they get will come at the expense of reducing the diversity elsewhere (unless, of course, they significantly hire away from other companies that are more diverse than they are)?
A bit of Googling turns up assorted claims on what percent of developers are women, but many seem to claim in the 10-20% range.
It would be interesting to take the data from Pinterest, and from other tech companies where this kind of data is available, and use that to classify them into three groups: those where women are under represented, those where they are over represented, and those where it is neither. The comparison should be to the percentage of women in tech, not the percentage in the general population.
It would then be interesting to see if there is something the companies in each group have in common.
I'd suggest anyone here seriously interested in the effects of Affirmative Action and similar policies read 'Affirmative Action Around the World' by Thomas Sowell. It's only 200pages, but incredibly sourced and well written. Highly recommend it for those wanting a further understanding.
65 comments
[ 7.1 ms ] story [ 197 ms ] threadIe: why don't they have the skill sets to get the job?
This reminds me of what many police and fire have been doing to bring in more women: reduce the standards on the physical tests...which just increases the danger for everyone.
And much respect to Tracy for moving the needle.
Do you work in tech, in which Asians and Indians of all walks dominate, or do you work in "tech" the same way runners work in "hollywood"?
I'm not sure if Pinterest posted this article with some other intent (like having more support programs), but in it's current state, it seems discriminatory.
I personally find tangible monetary value in employees with diverse backgrounds but I don't lower my bar. I just slightly discount sameness and slightly appreciate difference.
https://en.wikipedia.org/wiki/Rooney_Rule
Person 1 says that enforcing a 50 people of type A and 50 people of type B when the number of qualified candidates is 20 and 80 will result in either hiring significantly fewer people in general, or hiring under qualified people of type A. People of type B are being passed over because of their race/gender/ethnicity. This is discrimination.
Person 2 says that implying that anyone of type A could be under qualified is itself discrimination, and this possibility should not be addressed or considered. Nothing can change because the issue cannot be discussed.
We have a tendency to assume type A is qualified and type B isn't. That is a natural bias based on evidence. We must actively subdue our biases if we want to overcome them, sometimes with rules.
I think we sometimes forget that the goal isn't to get a qualified employee but a successful employee. Qualifications generally leads to success but not always, so we should constantly question our qualification criteria.
White males have benefited from discrimination for a long time. White males have been discriminated against sometimes.
This article isn't about discrimination or affirmative action. It's about setting a goal to be more diverse, not by lowering the bar, or forced hiring, but through awareness, inclusion, and an opportunity to interview.
BTW, everyone's been given a chance they didn't deserve.
Your interest in the content they host has zero relevance to the technical challenges they face as a platform and heavily trafficked website.
Unfortunately, biases are rarely so blatant. Many are even non-malicious and entirely unconscious, like how people tend to guess older ages for black youth than white youth. http://www.apa.org/news/press/releases/2014/03/black-boys-ol...
I still say they don't represent the technical community. And it's blatantly obvious they don't represent the HN community, given the downvotes.
I'm not saying there aren't biases, sexism, ageism, or racism in tech, but I'd appreciate it if you didn't point at the obvious troll and say "this right here is the face of tech".
[1] http://www.alexa.com/topsites/countries/US
Umm, in the actual example we are talking about, the evidence is on the side of that assumption being false: https://news.ycombinator.com/item?id=9977751 https://jaymans.wordpress.com/jaymans-race-inheritance-and-i...
In fact, this depends on the underlying population sizes of tall vs. short people, so this is a faulty argument.
More germane to the situation, however, is your initial assumption, whose analogue is dubious. So much so, in fact, that it is taboo to even allude to it. cf. pg's "What you can't say".
There actually are things driving men out of the field. Look at how many head nurses are male. Men in nursing is a problem as well. But this isn't NurseNews, it's HackerNews.
I'm not completely sold on the method, but the problem of subconscious bias in hiring is very well documented:
* Resumes with stereotypically White names receive 50% more callbacks than identical resumes with traditionally black ones. The addition of honors and special skills had a significant effect on the likelihood of White applications being called, but a statistically insignificant effect for the otherwise-identical African-American ones. http://www.nber.org/papers/w9873.pdf
* Applicants with male names are rated as more competent and offered higher starting salaries than identical applications with female names. Identical resumes with male names are called back more often than female ones. http://www.pnas.org/content/109/41/16474.full.pdf+html, http://advance.cornell.edu/documents/ImpactofGender.pdf
* Initial subconscious bias is logically justified post-hoc—in the linked experiment, the qualities deemed necessary for a position would actually be redefined by the person reviewing resumes depending on the gender of the applicants so that the male applicant's traits fit the position better. http://www.socialjudgments.com/docs/Uhlmann%20and%20Cohen%20...
There's quite a few more papers on the subject, it's easy to search for them. Depressing read all around.
1) ask people not include their names on their resume. Or don't use resumes at all. Create a web form application that does not require a name and only asks about the crucial skills you actually need.
2) Create a highly objective hiring process, such as this: http://sockpuppet.org/blog/2015/03/06/the-hiring-post/
I don't really buy that subconscious bias is the problem in tech. I've been part of a lot of hiring, we were actually consciously biased toward hiring women since we wanted better numbers. I had an objective interview process, always giving the same questions, and the hiring was still very disproportionately male whasian (the applicant pool was also very disproportionately male whasian, despite doing active reach out to woman-in-tech events).
The questions I asked were representative of things we actually had to do on the job, for instance: "write code to spider a web site." Does that unfairly weed out woman?
2. Also agree that an objective method would be preferable, thanks for the link. Great read. IIRC, in the study I link in the third bullet point, forcing the person reviewing the resume to create a rubric before reading any of them greatly reduced the biased results. Having an objective process goes a long way. The point in the link about confident interviewers performing better in the traditional hiring process is definitely true, and also tied in a roundabout way to gender, funnily enough (women tend to be less confident than men: http://www.theatlantic.com/features/archive/2014/04/the-conf...).
It's definitely not the only problem—the pipeline is obviously skewed dramatically—but I don't see any reason why tech workers would be exempt from the biases of the general population as they've been studied. Maybe the drive to hire more women would counteract it, like in your example—that's the reasoning under which minority quotas are implemented, after all. The problem of subconscious bias is that it's not consciously recognizable by its very nature. It's just gut feeling. I know a site that uses association to try to test it for you, if you're interested: http://implicit.harvard.edu/.
Women and brown people are not lower quality humans, and it has been shown over and over and over that companies who hire for diversity perform better.
http://lmgtfy.com/?q=do+diverse+companies+perform+better
Maybe because I'm not American the part that confuses me is the "Hispanic" vs "Latino", are they used to mean different things? Technically (very technically), Latin-American would include people talking French or Portuguese, while I guess Hispanic will only include Spanish speakers, e.g. not including Brazil. But I don't know if they are used in a different way in the US.
http://www.npr.org/sections/thetwo-way/2013/01/07/168818064/...
That would greatly reduce the usability of the charts in this case. The charts are much more useful when they all present the categories in the same order.
The possibility of that being true is certainly unsettling for me because I've always assumed that hiring by merit is intrinsically superior to hiring by quota.
-------------------------
But of course the minds of men and women are not identical, and recent reviews of sex differences have converged on some reliable differences. Sometimes the differences are large, with only slight overlap in the bell curves.
...
With some other traits the differences are small on average but can be large at the extremes. That happens for two reasons. When two bell curves partly overlap, the farther out along the tail you go, the larger the discrepancies between the groups. For example, men on average are taller than women, and the discrepancy is greater for more extreme values. At a height of five foot ten, men outnumber women by a ratio of thirty to one; at a height of six feet, men outnumber women by a ratio of two thousand to one. Also, confirming an expectation from evolutionary psychology, for many traits the bell curve for males is flatter and wider than the curve for females. That is, there are proportionally more males at the extremes. Along the left tail of the curve, one finds that boys are far more likely to be dyslexic, learning disabled, attention deficient, emotionally disturbed, and mentally retarded (at least for some types of retardation).
At the right tail, one finds that in a sample of talented students who score above 700 (out of 800) on the mathematics section of the Scholastic Assessment Test, boys outnumber girls by thirteen to one, even though the scores of boys and girls are similar within the bulk of the curve [NOTE this was from the pre-1994 SAT where the math section was harder and not truncated at the top. Today the ratio is a bit less than 2-1 at the top end.]. With still other traits, the average values for the two sexes differ by smaller amounts and in different directions for different traits. Though men, on average, are better at mentally rotating objects and maps, women are better at remembering landmarks and the positions of objects. Men are better throwers; women are more dexterous. Men are better at solving mathematical word problems, women at mathematical calculation. Women are more sensitive to sounds and smells, have better depth perception, match shapes faster, and are much better at reading facial expressions and body language. Women are better spellers, retrieve words more fluently, and have a better memory for verbal material.
...
Nonetheless, discussions of the leaky pipeline in science rarely even mention an alternative to the theory of barriers and bias. One of the rare exceptions was a sidebar to a 2000 story in Science, which quoted from a presentation at the National Academy of Engineering by the social scientist Patti Hausman: "The question of why more women don’t choose careers in engineering has a rather obvious answer: Because they don’t want to. Wherever you go, you will find females far less likely than males to see what is so fascinating about ohms, carburetors, or quarks. Reinventing the curriculum will not make me more interested in learning how my dishwasher works."
An eminent woman engineer in the audience immediately denounced her analysis as “pseudoscience.” But Linda Gottfredson, an expert in the literature on vocational preferences, pointed out that Hausman had the data on her side: “On average, women are more interested in dealing with people and men with things.” Vocational tests also show that boys are more interested in “realistic,” “theoretical,” and “investigative” pursuits, and girls more interested in “artistic” and “social” pursuits.
...
The most dramatic example comes from an analysis by David Lubinski and Camilla Benbow of a sample of mathematically precocious seventh-graders selected in a nationwide talent search. The teenagers were born during the second wave of feminism, were encouraged by their parents to develop their talents (all were sent to summer programs in math and science), and were fully aware of their ability to achieve. But the gifted girls told the research...
If only it was a joke...
1. Their approach seems like it will work, if it does, mostly by increasing the chances that women engineers or women engineering students will choose to work for Pinterest instead of work for someone else.
2. I've frequently read that there is a shortage of tech workers, and companies like Pinterest have trouble finding the people they need.
3. Putting this together, does this mean that much of any increase in diversity they get will come at the expense of reducing the diversity elsewhere (unless, of course, they significantly hire away from other companies that are more diverse than they are)?
A bit of Googling turns up assorted claims on what percent of developers are women, but many seem to claim in the 10-20% range.
It would be interesting to take the data from Pinterest, and from other tech companies where this kind of data is available, and use that to classify them into three groups: those where women are under represented, those where they are over represented, and those where it is neither. The comparison should be to the percentage of women in tech, not the percentage in the general population.
It would then be interesting to see if there is something the companies in each group have in common.