To reduce hiring bias, some tech companies take a cue from symphonies

Photo by Breather.

Most technology companies claim that they want to have a diverse workforce. But relatively few truly achieve it. Today, new software companies promise to help tackle this challenge. Can they deliver?

To answer that question, it’s important to first understand the problems they’re up against. Three separate, but equally problematic mindsets have contributed to the stunted growth of diversity within technology companies:

  • First, a focus on “culture fit” — no longer just a dominant mindset in Silicon Valley, but now across industries. Cultural assessments in hiring can be self-reinforcing. As Northwestern University professor Lauren A. Rivera writes, the use of “culture fit” as a hiring metric “has shifted from systematic analysis of who will thrive in a given workplace to snap judgments by managers about who they’d rather hang out with. In the process, fit has become a catchall used to justify hiring people who are similar to decision makers and rejecting people who are not.”
  • Second, overemphasis on the “pipeline” problem — that there simply aren’t enough diverse candidates to hire. It’s true that the percentage of qualified diverse applicants isn’t yet representative of society at large. But that reality doesn’t account for the remaining, sizable gap between the percentage of minority students graduating college with the requisite degrees and skills and the number hired by technology companies for technical jobs.
  • Third, a focus on a workplace culture of meritocracy — that results and performance are all that matter. Myopic meritocracy may lead to anything but. As researchers at MIT and Indiana University have shown, managers at companies that emphasize a culture of meritocracy disproportionately favor men over women in performance evaluations and monetary awards. Similarly, other researchers recently found that gender impacts the acceptance of submissions on the open source code repository site GitHub. On the whole, their research showed that women’s contributions tended to be accepted more than men. But, the researchers also found that when a woman’s gender was identifiable, their submissions were actually rejected more often, “suggest[ing] that although women on GitHub may be more competent overall, bias against them exists nonetheless.”

Enter companies like GapJumpers, who seek to help other companies overcome these obstacles. GapJumpers capitalizes on already proven techniques to increase diversity in other industries: blind auditions. In the 1970s, symphony orchestras started to use blind auditions to hire new musicians. “Musicians auditioned behind screens so the judges couldn’t see what they looked like, and walked on carpeted floors so the judges couldn’t determine if they were women or men — the women often wore heels,” writes Claire Cain Miller in the New York Times. “Researchers from Harvard and Prince­ton took notice and studied the results; they found that blind auditions increased the likelihood that a woman would be hired by between 25 and 46 percent. In fact, with blind auditions, women became slightly more likely to be hired than men.”

GapJumpers seizes on this insight: Keeping hiring bling as long as possible can decrease the risk of bias. Now, GapJumpers helps tech companies provide the functional equivalent of blind auditions: open-ended challenges for job candidates. Once applicants have solved the skills-based challenges, GapJumpers removes identifying information about the candidate — name, gender, graduation year, college, and address are all stripped away. To be sure, as Miller notes in the New York Times, eventually the jobseeker’s identity is revealed, but “[t]he first piece of information the hiring company sees is applicants’ scores, and, based on those, it selects candidates to interview. Only then does it see their names and résumés.”

And that approach seems to have worked. As Jane Porter reports for Fast Company:

For its first seven months in business, GapJumpers gathered data from nearly 1,200 auditions across 13 companies — attempting to see how the numbers stacked up when the early stages of hiring were done blindly. They found that male applicants raised concerns about having to prove themselves in a blind test more often than women. Once the blind challenge was completed, the gender breakdown of those candidates hired was 58% women, 42% men.

Other companies are taking different “bias interruption” approaches. For example, Textio helps major companies like Microsoft at the beginning of their hiring process: the text of a job posting. Textio uses machine learning and language analysis to analyze companies’ job postings, already finding more than 25,000 phrases that indicate gender bias — phrases like “mission critical,” “top-tier,” and “aggressive,” for example, decrease the portion of women who might apply for a job. Gild takes yet another approach to solving the hiring problem: their software locates candidates based on code they’ve published on sites like GitHub, and then removes the candidate’s biographical information before relaying the candidate to an employer.

At their best, these new hiring systems are helping replace human judgments — often clouded with even unconscious assumptions — with objective criteria. In doing so, they help tech companies take a step in the right direction, challenging the “culture fit” mindset, the supposed “pipeline” problem, and making good on the promise of workplace meritocracy.