Can an algorithm remove gender bias?

female students
Credit: Francisco Osorio/Flickr

The Holberton School shows increased female admissions due to an automated selection process.

Diversity (or lack of it) in the tech industry is one of today's hottest topics. And many reports suggest that human intervention leads to bias towards males. Can machines do better? Holberton School thinks so.

The San Francisco. Calif.-based Holberton School has as its mission "to train the best software engineers of their generation." The school offers a two-year program for students to become full-stack software engineers. And, according to its website, the school's "selection process is based only on talent and motivation, with no consideration given to gender, nationality, ethnicity or social status."

Holberton shared some details about the results of its automated, software-driven admissions process. The school reported that because of the automated admissions process, 40% of accepted applicants for its next class of students are female. At the same time, the school reported an acceptance rate of 3% (by comparison, Harvard's 2016 acceptance rate was 5.9%).

“By using automated processes, we've selected the most motivated and talented individuals, and those who best fit with our problem-oriented curriculum,” said Julien Barbier, co-founder of Holberton School. “The funny thing is, we're also finding that the automated processes have dramatically increased the number of female and minority students.”

I asked Barbier if the process to enroll for other courses is driven by manual selection? I mean if someone wants to enroll for RHCE, why would Redhat want to lose a paid student based on gender? Barbier said, “Yes, the vast majority of selection processes of colleges / universities are manual. Not sure we can compare to RH, which is targeting working software engineers. But companies are also facing diversity issues.”

He added that while the industry is trying to solve the gender diversity problem, most organizations are taking shortcuts such as specifically selecting women based on their gender, having forced quotas, or offering a discounted price for women. “At Holberton, all women went through the exact same selection process, and they earned their place as much as men. And this is very important. Because if it is not the case, you put them in a position that you are trying to solve: men will think they are here not because they are good, but because they are women. And that is really bad. This is a cultural problem. And it will take a lot of time before it changes. And I believe the tech industry is really working hard to make it happen. I am confident that we (as the tech industry) will succeed.”

Barbier also pointed of that the vast majority of companies and schools actually do want more diversity, but one big issue that we all face is unconscious bias.

When asked about the gender makeup of its faculty, Barbier said 20% of its 100 mentors are female. Barbier added that “it’s good to note that only 12% of software engineers in the valley are women. Women in tech earn less than men for the same job. So if you think about it, they comparatively have less free time for volunteering at Holberton or any other initiatives. Also, there is unfortunately more pressure for women in tech. In our industry, they face higher expectations and are criticized much more easily, and much harder.”

Holberton's algorithm-based admissions process is an intriguing solution to the gender bias problem, particularly in the context of the emergence of automation, machine learning and AI. What do you think? Can we remove gender bias by taking humans out of the equation?

This article is published as part of the IDG Contributor Network. Want to Join?

Drexel and announce Analytics 50 award winners
View Comments
Join the discussion
Be the first to comment on this article. Our Commenting Policies