recentpopularlog in

diversity

« earlier   
A MongoDB Engineering Response to the Anti-Diversity-Effort Manifesto
The following is an email I sent to the MongoDB engineering team this past
Friday. At the behest of the team, I am now making it public. —Eliot



You’ve likely heard of the 10-page memo that was published by a (now
former) Google employee regarding Google’s diversity efforts. Parts of the
diversity 
yesterday by pietvanzoen
Why Meritocracy Doesn’t Work
Reading the mystery Googler’s anti-diversity manifesto really pissed me off. The least of the problems was his terrible use of footnotes.

Inclusive Diverse Team

It’s easy to have a knee-jerk reaction, so I’m not going to go into all the bad arguments because frankly there are too many. If someone can’t identify the issues with citing biological differences as a basis for discrimination, it’s not likely that logic or research is going to sway their mind.

I would like to have a serious discussion about diversity, but I’m not going to treat racist and sexist claims as if they qualify for serious discussion.

So if we’re going to seriously discuss the actual intellectual arguments that genuinely smart people seem to actually believe in, I’ll try and dust off my philosophy brain. I strongly encourage anyone to find the holes in this argument and then build it up to make it better as I’m going to skip over or simplify some things.

This may invite a lot of trolls, but so be it. The one upside to the recent sexual harassment scandals in venture capital and the mystery Googler’s anti-diversity manifesto is that now we can see the issue in broad daylight.

I would rather see the issue and speak out against it than be in the dark or be silent.

I will focus on one claim that seems to be the general undercurrent of all anti-diversity arguments in the valley: Meritocracy is the best form of governance.

Faith in Meritocracy

There is an absolute insistence and faith in meritocracy from many, many people in Silicon Valley.[1] In case you’re not familiar, a meritocracy is a political system where power is allocated to those with the most ability. In the case of Silicon Valley, this equates to “the best ideas win and the best people get promoted.”

Best ideas win, best people get promoted

The idea that Silicon Valley is a functioning meritocracy is then used as a basis for saying that policies like affirmative action and diversity quotas are bad. I am not agreeing with this argument, just making sure the argument against diversity policies is clear.

The basic argument is:

This is a meritocracy.
Meritocracy is good.
In a meritocracy, only ability is considered for advancement.
Diversity policies require factors outside of ability to be considered for advancement.
Therefore diversity policies undermine meritocracy.
Therefore diversity policies are bad.
Diversity at Euclid Co.

To make this simpler and more concrete, let’s say we have a company called Euclid Co. with a population of 60 blue squares and 40 red triangles. We have 10 job openings for senior polygon, and only 10 percent of the population is qualified for this position. This means 6 blue squares qualify and 4 red triangles qualify. Let’s assume that the population outside of Euclid Co. is 50 percent blue squares and 50 percent red triangles.6 blues to 4 reds

The meritocracy argument says that any diversity quota that dictates that the jobs be given to a 50-50 split to match the population would be unfair. This is because only 4 red triangles qualify, so the sixth blue square is going to be shut out of its dream job because a red triangle is going to be randomly allocated to the senior polygon position that they don’t deserve.

The blue square is sad.

Sad Blue Square

How Meritocracies Really Work

This argument equates any diversity policy designed to combat discrimination with discrimination against the majority. It flips the anti-discrimination argument like this:

Discrimination is bad.
Diversity policies are discriminatory.
Therefore diversity policies are bad.
Essentially it challenges anyone campaigning against discrimination to justify why their diversity policy isn’t discriminatory and regards any such policy as an undeserved handout.

To put it even simpler, the meritocracy argument says that diversity policies tilt the level playing field in favor of the undeserving.

Bias in Big Company Hierarchies

Let’s reset the scenario at Euclid Co. and see what happens when we introduce bias. We’ll make the company bigger so the math is easier. There are 60,000 blue squares and 40,000 red triangles. There are 10,000 senior polygon positions open, and 10 percent of the population is qualified: 6,000 blue squares and 4,000 red triangles.

Unfortunately, there is a small bias in the promotion process. In a 100 question skill assessment, one of the questions is, “Why are 90 degree angles the most perfect type of angle?”

Red triangles, of course, don’t understand the question because they enjoy all sorts of angles. They are triangles, after all, and come in all sorts of angles, whereas squares only have 90 degree angles. So the triangles get the question wrong, and the squares get it right.

Confused Red Triangle from 1 percent bias

However, believing that a 90 degree angle is perfect doesn’t have much impact on the job performance of a senior polygon and there are lots of other questions that don’t predict job performance. So let’s imagine that this question or another similar factor has introduced a minor 1 percent bias.

This small 1 percent bias means that 11 percent of blue squares are considered for promotion and only 9 percent of red triangles are considered. That’s 6,600 “qualified” blue squares and 3,600 “qualified” red triangles —10,200 candidates total.

So if there are only 10,000 positions open and they are divided evenly among the pool of candidates, we will get 6,471 senior blue squares and 3,529 senior red triangles.

This small 1 percent bias has already resulted in a loss of representation. From 40 percent to 35.29 percent.

Continuing this math with the same small 1 percent bias, we will have only 30.86 percent red triangles represented at the VP level, 26.75 percent represented at the senior VP level, and down to 23 percent at the C level.

More levels of seniority will compound biased representation. In other words, hierarchies amplify any bias in the system.



(For those wondering, even a “small” company of only 2000 people can have 9 levels of seniority. Also, yes…the lack of diversity in the U.S. Congress probably has something to do with this phenomenon.)

The Bias of a Growing Company

Some will argue, “We don’t have hierarchies. We don’t even have titles!”

First, get real. Even when we remove titles, people sort themselves into a pecking order. People use other ways of indicating social status, including who talks first and last in meetings. But okay, let’s just assume that Euclid Co. has no hierarchies.

Will this lack of hierarchy eliminate the problem? Maybe, but only if that company doesn’t hire anyone.

Let’s say that Euclid Co. starts with a small population of 5 red triangles and 5 blue squares, and there are no promotions. Euclid Co. is going to grow from 10 people to about 10,000 — every Silicon Valley company’s dream.

As Euclid Co. starts to grow, there is a slight 1 percent bias in the hiring process that gives new blue square hires an 11 percent chance of getting the job and red triangles a 9 percent chance of getting the job.

In the beginning, this makes very little difference. By the time Euclid Co. is ~20 people it will be ~51.58 percent blue and ~48.42 percent red. But things quickly escalate.

By the time the company is ~10,000 people, it will be 69.69 percent blue and 30.31 percent red. Again a pretty radical problem in representation introduced by a relatively small bias over time just by a bad hiring process.

It’s true that we don’t hire fractional people so the math here is off. However, rounding the hiring bias makes the situation worse. With standard rounding, the company will be 76.19 percent blue at just 21 people and 99.95 percent blue at ~10,000 people.Time Clock

Time will amplify any bias inherent in the system.



Meritocracy by Democracy

Of course, some will say, “Our hiring process has no biases.” The variant of this that I have heard is, “Our hiring process has no biases because it’s peer-to-peer and we vote.”

This is a kind of magical “democracy beats racism” in a meritocracy argument. Unfortunately, in a democracy, discrimination can go viral.

Let’s say blue squares are really biased and only hire other blue squares. Red triangles hire only red triangles. Clearly no problem. Everyone is equally biased and the representation will remain proportional.

However, if only one part of the population is biased, problems show up quickly. Let’s say that blue squares only hire blue squares while red squares hire equally.

If the population starts at 50 red and 50 blue, in just one round of hiring, blue will hire another 50 blues. Red will meanwhile hire 25 reds and 25 blues. So in just one round, there will be 62.50 percent blue and 37.5 percent red.

Democracy Can Actually Help

It’s worth digging out a spreadsheet and playing with the numbers in many of these scenarios because a relatively fair-minded population can erase a lot of biases over a sufficient length of time by voting. If you’re more technically minded, try playing with genetic algorithms to see the same thing.

It gets very interesting when constraints are put on the population, such as a limited number of senior polygon positions, limited venture capital funding, limited housing, and so forth. But that’s too complicated for a Sunday night essay.

The end result is that if blue is only slightly biased, this effect may be relatively insignificant and the bias in representation can approach a limit. However, that’s the point. If people make great decisions, democracy is great.

This greatness requires two things:

All people are perfectly rational.
All people have access to perfect information.
Neither of these are true.

Even if the counterargument is, “In our company we make perfectly rational decisions,” it doesn’t matter. There just has to be some bias somewhere in the system for things … [more]
Meritocracy  Googlememo  Google  Diversity  db 
yesterday by walt74
In Search of a Good Emperor - The New York Times
A new paper from the economists Oded Galor and Marc Klemp finds a strong correlation between diversity and autocracy in pre-colonial societies, with a legacy that extends to today’s institutions as well.
hierarchy  diversity  DEMOCRACY  rossdouthat 
yesterday by Walpole
I’m An Ex-Google Woman Tech Leader And I’m Sick Of Our Approach To Diversity!
I ran cross-functional engineering projects managing 20–30 men (engineers mostly) at Google. Most of the time, I was the only woman. Men who reported to me liked and respected me. Men who managed me…
google-memo  gender  diversity  programming 
yesterday by pmigdal
Suppressing debate will not help women at work
August 11, 2017 | Financial Times | Anne-Marie Slaughter.

I have argued extensively that women at the top will need men supporting them just as men at the top need women supporting them; hence one important way to increase the number of women in leadership positions is to change the way society values men just as we have changed how society values women. It is a pity that Mr Damore will not be able to lead that conversation at Google.

The company should be committed to diversity and equality as a first principle, to get the best from all its staff: this is a business priority as well as a moral imperative. Yet one of the major values of diversity and equality is to encourage people to challenge orthodoxies of many different kinds, even if others are offended. And both of these things can be true at the same time.
Anne-Marie_Slaughter  women  diversity  Google  firings  orthodoxies  stereotypes  cultural_stereotypes  Sundar_Pichai  equality 
2 days ago by jerryking

Copy this bookmark:





to read