One of the most abused headline genres in the tech world uses the phrase “Rocks Silicon Valley.” But today’s news can at least qualify as “Raises Eyebrows”: The United States Office of Federal Contract Compliance Program (OFCCP), part of the Dept. of Labor, has brought a complaint against the fabled Big Data startup Palantir, claiming the firm discriminates against Asian job applicants. Note that the suit will be heard by an Administrative Law judge, not in federal court.
There are a number of interesting aspects here, but first, a needed disclosure. OFCCP approached me about a year ago concerning the H-1B visa issue, and had me as a panelist at their recent conference for investigator training. This was not related to the lawsuit, which I just learned of yesterday, but I do wish to state that I have found the OFCCP leadership to be highly competent and professional, and to have a sincere concern about abuse of the visa. My talk was apparently quite well received.
One question sure to come to many readers minds is “Who qualifies as Asian?” In particular, do international students from China and India, say, who are studying at U.S. universities, count as “Asian”? Since the suit is based on an old federal Executive Order regarding affirmative action, is it limited to just U.S. citizens (USCs) and legal permanent residents (LPRs)? One might guess that it is the latter, since otherwise it would invalidate, for instance, employment-based green card statutes, which require employers to give hiring priority to the USCs and LPRs. But it is quite possible that the investigators simply relied on the applicants’ self-identification of race, including the non-USC/LPR applicants.
In terms of statistical methodology, the complaint seems to oversimplify the situation, with the claims taking the form of “x% [large number] of the applicants were Asian yet only y% [small number] of those hired were Asian.” Such an approach is almost always invalid, because it falsely assumes that all applicants are equally qualified. The complaint does say the investigation considered “equally qualified” applicants, but since they indicate that at least part of their data consists of analyzing applicants rejected at the re’sume’ level, it is doubtful that they were taking the most important criteria into account, as they tend to be the intangibles.
Word on the street is that it is extremely tough to get a job at Palantir, regardless of race. A couple of years ago, for instance, a student of mine (white) was interviewed for an internship and rejected, even though I would rate him as one of the top few undergraduates I’ve ever had over the years. Top employers in general, and especially Palantir, are looking for deep insight, keen intuition, creativity, a noticeable verbal “presence”, quickness and so on.
Note too, by the way, that Palantir is an analysis company, not a software firm like Google and Facebook. They use mathematical analysis, statistical/machine learning methodology and so on. They accomplish these goals via software that either they or others (open source) write, but the software is only a means to an end. Just having good coding skills is far from sufficient.
Thus someone reading through applicant re’sume’s at Palantir will pick up intangibles that OFCCP investigators cannot discern. In addition, given Palantir’s government work, it may set a higher bar for the non-USC/LPR applicants, who are likely overwhelmingly Asian.
Of course, another possibility is that Palantir’s screening criteria, though benign in intent, have discriminatory effects that are unrelated to the jobs being filled. If Palantir does indeed use “verbal ‘presence'” as a criterion, for example, has the company given careful thought to just how important this trait is for effective job performance? Their assessment of the importance of this characteristic presumably is well-founded — the company’s business revolves around interaction with clients, after all — but if not, it would likely have a disproportionately adverse impact on Asian-American applicants.
In other words, this topic is very highly nuanced, and it is puzzling to see that the complaint seems to capture none of those subtleties. It may be that some will emerge as the case moves forward.