Illustration by Christoph Niemann

This summer, Google, Apple, Facebook, Twitter, and other Silicon Valley superpowers released demographic reports on their workforces. The reports confirmed what everyone already knew: tech is a man’s world. Men make up sixty to seventy per cent of employees at these companies, and, notwithstanding rock stars like Facebook’s Sheryl Sandberg and Yahoo’s Marissa Mayer, senior leadership is even more overwhelmingly male. A recent study by the law firm Fenwick & West found that forty-five per cent of tech companies there didn’t have a single female executive. (The picture is also bleak when it comes to ethnic diversity.) The Valley seems to take the problem seriously—Apple’s Tim Cook recently stated his commitment to “advancing diversity”—but there’s a long way to go.

A familiar explanation for tech’s gender disparity is the so-called pipeline problem: the percentage of female computer-science graduates has almost halved since the nineteen-eighties. But this doesn’t fully explain why there are so few women in senior management or on company boards (where skills other than programming matter). Nor can it explain the high rate of attrition among women in tech. A 2008 study found that more than half of women working in the industry ended up leaving the field. The pipeline isn’t just narrow; it’s tapering.

Tech companies may pride themselves on being meritocracies, but unconscious biases shape the way they hire and promote. Such biases can be tremendously powerful. A 2012 study asked top research scientists to evaluate job candidates with identical résumés. The scientists judged female candidates to be less capable than male ones, and suggested significantly lower starting salaries for them. Even more striking was a 2005 experiment in which participants evaluated applications for a job as chief of police, scanning résumés that indicated varying levels of formal education and on-the-job experience. A male candidate who had less schooling would be credited with street smarts, but a woman with an identical résumé would be dismissed for not having enough education. When the qualifications were flipped, so was the justification for hiring the man. The hirers were really just rationalizing a gut decision about who was right for a traditionally male job.

Techies have been around a lot less long than police chiefs, but they have an entrenched gender profile, too. Joan Williams, a law professor at the University of California, Hastings, and co-author of “What Works for Women at Work,” told me, “Even though it’s a job that basically involves sitting at a typewriter, there’s a real emphasis on the idea that these are jobs that only a real man can do.” The result is a Catch-22: women need to act “masculine” to fit in, but get labelled as difficult and pushy if they do. In a recent study of almost two hundred and fifty performance reviews, the tech entrepreneur Kieran Snyder found that three-quarters of the women were criticized for their personalities—with words like “abrasive”—while only two of the men were.

Subverting these biases requires more than training. Instead, companies should be looking for what Williams calls “bias interrupters”: systems that identify bias and intervene to mitigate it. For instance, until the nineteen-seventies classical-music orchestras were almost entirely male. Once blind auditions were introduced, the percentage of women quintupled. A startup called Unitive is designing software to do similar work. The software forces recruiters to decide what characteristics they value most, and rank blind résumés according to a series of criteria. It also helps companies design interview questions that are carefully tailored to job requirements. The hope is that this will mitigate what Laura Mather, Unitive’s founder, calls the “ ‘I just don’t want to hang out with this person’ problem.” Google, similarly, now requires interviewers to ask every candidate for a given job the same questions, so that bias doesn’t shape what gets asked.

Promoting diversity isn’t, as many techies think, pure do-gooderism. It’s genuinely good for business, since a large body of evidence suggests that making organizations more diverse can also make them perform better. A recent McKinsey study found that the organizations with the most diverse executive teams had dramatically higher returns on equity and earnings performance than those with the least diverse teams. A study of investment clubs by Brooke Harrington, a professor at Copenhagen Business School, revealed what she called a “diversity premium”: groups made up of both men and women outperformed single-sex ones. A worldwide study, published in May, of more than four thousand R. & D. teams found that gender-diverse teams were considerably better at driving “radical innovation.”

The snag is that, while diverse groups perform well, they can be harder to manage and more challenging to work in than homogeneous ones, precisely because diverse perspectives lead to more disagreement and conflict. So even when the results are better people are less happy with the experience. To fix this, you have to consciously reshape how people deal with one another, and that entails some cultural disruption. But it’s a price worth paying. Now that tech companies have evolved into consumer-electronics and social-media giants, their customers come from every demographic group imaginable. It’s time their workers did, too. ♦