Can Companies Use Big Data to Fight Racist and Sexist Hiring Practices?

Humans are fallible, biased decision-makers, and even the most well-intentioned hiring managers tend to favor candidates who “look like me” or “act like me.” These unintended prejudices—whether based on race, gender, or socioeconomic background—are precisely the shortcomings that a growing number of big-data firms hope to address through large-scale analytics.

By mining vast amounts of personal and professional data, these companies claim they can not only match employers with high-performing candidates, but also help close long-standing diversity gaps in the workforce.

“Big data in the workplace poses some new risks, but it may yet turn out to be good news for traditionally disadvantaged job applicants,” said David Robinson, a principal at Robinson + Yu, a consulting group focused on the intersection of social justice and technology.

Concerns remain, however. Earlier this year, the White House released a landmark report on big data warning that, intentionally or not, companies could use data-driven systems to discriminate against minorities and low-income populations. The Federal Trade Commission has echoed these concerns, recently holding a workshop on “discrimination by algorithm.”

“Big data can have consequences,” said FTC Chairwoman Edith Ramirez. “Those consequences can be either enormously beneficial to individuals and society, or deeply detrimental.”

Despite these warnings, companies working in HR analytics often strike a more optimistic tone.

“We think there’s a huge upside,” said Kyle Paice, vice president of marketing at Entelo, a San Francisco–based startup that uses predictive analytics to help companies identify talent likely to be open to new opportunities.

Paice noted that many companies want to build more inclusive workforces—particularly as research shows diverse teams are more innovative, more profitable, and experience lower turnover. Yet organizations often struggle to attract diverse applicant pools, such as women in technical roles, where candidates may face persistent negative bias. Entelo aims to widen the net, surfacing qualified candidates who might otherwise never encounter a given opportunity.

Still, data itself is not inherently neutral. Just as hiring managers may be unaware of their own biases, data can encode blind spots and historical inequities, said Solon Barocas, a postdoctoral researcher at Princeton’s Center for Information Technology Policy and coauthor of a study on big data’s disparate impact.

“Data mining works by learning from patterns in the past,” Barocas explained. “How you assemble those examples determines the models you build—and the mistakes you make.”

That dynamic can easily perpetuate bias. “If a company hasn’t historically hired women or minorities, the model may conclude those groups are less qualified,” he said. “Even efforts to minimize turnover can end up systematically excluding certain populations.”

Big-data firms argue that such outcomes are neither inevitable nor acceptable. Evolv, a San Francisco–based workforce analytics company, says it actively adjusts its scoring models to avoid discriminatory effects.

For example, Evolv found that employees who live farther from work are more likely to quit. However, the company excludes this variable from its hiring algorithms because of its potential correlation with socioeconomic status and race, said Michael Housman, Evolv’s chief analytics officer.

“Distance from work is something we don’t use because it’s a can of worms,” Housman said. “We don’t want to go near anything that could discriminate against disadvantaged communities.”

While compliance with federal equal-employment laws is part of the motivation, Evolv also follows an internal big-data code of ethics intended to minimize unintended harm, Housman added.

Barocas cautions that good intentions alone are not enough. Preventing bias requires careful model design, contextualized data, and a deliberate commitment to diversity as a goal in itself—not merely a byproduct of optimization.

Done correctly, however, big data “can be an enormous force for good,” Barocas said. “The key is not assuming that data-driven automatically means fair.”

At minimum, data-driven approaches may be an improvement over traditional hiring practices, said Daniel Castro, a senior analyst at the Information Technology and Innovation Foundation.

“All data is biased. All data is interpreted,” Castro said. “The difference is that with data, we have tools to identify and mitigate those biases—and that can move us closer to a fairer system than the one we have today.”

To access the original article, click here.