Tech Companies Try to Fight Unconscious Bias in Hiring by Making Job Candidates Fight Coding Bots

A startup in Silicon Valley is challenging conventional hiring practices in the white, male-dominated world of tech by giving coders a platform to be tested on their raw talent.

San Francisco-based CodeFights offers a gamified coding platform for job-seeking programmers to reach recruiters by performing well—but also anonymously—on coding challenges. First launched in 2014, the platform offers coders the opportunity to battle each other, or bots, and earn points for accuracy and speed. High performers who reach a “critical point” in the game are then given the option to connect with employers. Last month, the company released its newest mode, a level-by-level solo game in the style of Candy Crush, called Code Arcade.

Codefights CEO Tigran Sloyan says his avid participation in local and international math olympiads in high school gave him the idea to apply game design to math and programming skills. The goal was to create addictive games that help players improve and prove their competencies, without having their race, gender or backgrounds come into play.

“The biggest problem with diversity in tech is that humans are too involved in the skill evaluation process," Sloyan says. “We tend to like people who have a similar background to ours, which created a self-reinforcing cycle. Having people engage in these games provides a platform for discovering talent that’s usually invisible to traditional recruiters who mainly rely on pedigree for finding talent."

Silicon Valley has a well-documented issue with diversity, with its top tech firms lagging behind the rest of the country on this measure. Part of the problem is unconscious bias, which can negatively impact minorities during the recruiting and interviewing process, according to the Anita Borg Institute for Women and Technology.

The divide in the workforce also stems from educational disparities. Even though the likes of Dartmouth University and Harvey Mudd College have started churning out nearly equal numbers of male and female computer science graduates, only 24 percent of the jobs in computing are held by women in the U.S.—a share expected to fall to 22 percent by 2025.

As of 2015, only 11 percent of tech executives in Silicon Valley were women. Blacks and latinos make up less than 5 percent of the workforce at major companies like Facebook and Google. These stats show the pressing need for less-biased recruiting systems.

The focus of some tech companies on insular cultures can also shut some people out, says Karla Monterroso, vice president of programs at nonprofit CODE2040, which supports programming opportunities for black and Latino professionals.

“The concept of ‘cultural fit’—which has been taken as the holy grail for hiring in many tech companies—is often applied incorrectly to filter for potential hires who look or act like the person doing the interviewing,” Monterroso says.

In CodeFights’ Code Arcade mode, players work their way through a map comprising different locations, each revolving around various programming and computer science tasks, including algorithmic, database and front-end challenges like building user interfaces. As with CodeFights’ other modes, once a player’s skills pass a “critical point,” CodeFights asks the player if they’d like to be connected with the company’s 50-plus partners, which include Uber, Dropbox and Quora.

If a players says yes, a member of the company’s “Talent Engineers” team walks the job seeker through some of the companies they work with to determine the best matches. Then, the candidate’s CodeFights results are sent to between five and 10 companies, who can then reach out directly to request interviews. (If a company hires a referred candidate, it pays CodeFight 15 percent of a candidate’s first year salary—the service is free for users.)

Companies can even list specific challenges on CodeFights. In 2015, ride-hailing company Uber invited programmers to participate in a timed, competitive coding challenge called UberBot that tackled problems like finding the optimal route for a car or matching riders for uberPOOL.

“Through CodeFights, we can see how candidates would solve real challenges Uber is facing,” Uber senior product manager Bob Cowherd told Quartz. “Its tools have helped us expand our ability to find great engineering talent we might have otherwise missed.”

Since launching two years ago, CodeFights has built up a community of over 500,000 developers in over 200 countries. The majority of the talent pool—32 percent— is concentrated in the U.S. India is next with 10 percent of the participants. The company says half of the candidates who have expressed interest in being connected with an employer were subsequently hired.

Montessoro believes scalable platforms that help identify candidates’ technical competencies are important. But she worries competitive assessments like CodeFights’ modes might have potential bias built-in, citing research that says timed testing, and competitive formats, have statistically impacted marginalized communities negatively. Men also tend to perform faster in timed conditions, whereas women are more thorough and take more time.

“The emphasis on speed is also misaligned with the technologists I work with regularly, who tell me that deliberate and intentional choices are much more likely to be a part of their work life than breakneck speed,” she said.

CodeFights says it does not function like standardized tests. Not all tests are competitive or timed, and users can take on timed challenges once they are better prepared.

“The stakes are not high at all, unlike when you are doing an SAT test,” Sloyan says. “Most of the user base is using CodeFights to have fun and practice their skills so there is almost zero pressure. The job angle is more like a nice bonus for those who get there.”

While it declined to disclose the demographics of the players on its platform, Codefights also said in female vs. male matches, both genders won nearly half the time, an illustration that the tasks are not skewed to favor either gender.

CodeFights goes some way in eliminating bias, but not all the way. The platform automates technical skill evaluation, but players identities are revealed when they are connected with companies. Biases can still crop up during the hiring process, and gender, race and age can have an outsized impact on employees’ careers once they are in the door.

Some experts also disagree on the utility of blind recruiting, arguing it turns off some prospective talent, lengthens an already-long hiring process and fails to assess whether a person might fit a company’s culture, which can be critical to their success on the job.

Even if it cannot omit all biases, CodeFights hopes focusing the attention of recruiters and hiring managers on competency, before background and pedigree, can go some way to prevent recruiters from stereotyping.

“Any fast-growing company you look at, they are starved for quality engineering talent,” Sloyan said. “On the other hand, the whole country is full of smart and talented engineers that don’t look like it on paper. We have the key to those hidden gems.”