Slashdot Mirror
Silicon Valley Big Data Startup Palantir Responds To Labor Department's Discrimination Lawsuit (fortune.com)
Silicon Valley's big data startup Palantir, founded by tech billionaire Peter Thiel, has filed a response to the Department of Labor after the agency sued the company for discriminating against Asian job applicants. From a Fortune report: Palantir says that no discrimination took place and that the Labor Department's statistical analysis -- the basis for the recent suit -- is faulty. The suit, according to Palantir's 15-page response, wrongly suggests that the company "should have hired a workforce that matched the racial composition of the group of individuals whose resumes Palantir received, without regard to candidate qualifications." Palantir's response also points out that the suit addressed only three out of 44 job titles for which Palantir hired employees within the 18-month analysis period conducted by the Labor Department. What's more, says the response, 36% of those eventually hired across all the job openings within that timeframe were Asian -- a rate that exceeds the percentage of qualified Asian employees in the external labor market, according to stats from the U.S. Equal Employment Opportunity Commission.
I was surprised to find that's actually explicitly stated in the official regulations. The Code of Federal Regulations actually says "discriminination [against white men] is not discrimination". I thought they just unofficially ignored it, but it's official, written policy.
One might think such discrimination is rare, but it's actually extremely common, just worded in a way that makes it harder to recognize immediately. It looks something like this:
Admittance to the school is competitive. Highest scores get admitted.
Being female adds 15 points
Being black adds 20 points
Being latino adds 10 points
Being asian adds 5 points
Nobody is being discriminated AGAINST, right, only discriminated FOR. Except with competive admissions that's *precisely* the same result as this:
Being male subtracts 15 points
Being white subtracts 20 points
Being latino subtracts 10 points
Being asian subtracts 15 points
In a competitive context, discriminating FOR anyone is discriminating AGAINST everyone else.
The Chinese restaurant down the street has a higher than average ratio of Asians working there. Shit, same deal with the Mexican restaurant. Whole lotta Hispanics working there.
The idea that hiring for any company or job should be statistically representative of "the population" is ludicrous in and of itself. Any such enforcement is a farce.
They allege that from 1160 applicants for positions as Software Engineers only 11 Asian were hired, while 14 non-Asians were hired. Even though 85% of applicants were Asian.
Not sure that this is enough data to prove discrimination, but at first glance those numbers do warrant some suspicion.
The argument is bullshit no matter what "the population" refers to (that's why I put it in quotes). Companies select the best applicants from a pool of applicants, not a random subset of people who meet the requirements. That is almost guaranteed to result in a highly biased selection of applicants when you look at gender or cultural background, even in the absence of any bias.
To see how subtle this can be, consider a large applicant pool of equally qualified men and women (on average). If you pick the 7 best applicants, they will almost all be male, yet there is no gender discrimination. Likewise, if you pick the 7 worst applicants, they will also almost all be male. That's because male and female populations have different variances even when they have the same averages. Another counterintuitive property of these kinds of selections is that the law of large numbers works against you: the larger the applicant pool, the more biased the outcome of the selection. That's why Nobel prize winners are overwhelmingly male, and why prisons are also filled primarily with males.
The DOL statisticians aren't stupid, they understand this, which strengthens the case that this is malicious, politically motivated persecution of a political opponent.