How to Keep AI from Becoming “Too Male”

At Facebook, a top tech company, only 15% of their AI researchers were women in 2018. The same year at Google, only 10% were women. Based solely on these statistics, it is evident that there is a gender problem in tech. The lack of gender diversity in AI reveals itself in ugly ways, as evidenced by a recent Amazon scandal regarding their AI recruiting tool. With resumes from the last ten years — most of which came from male applicants — the AI taught itself that male candidates seemed preferable to female candidates. This manifested itself in automatically demoting applications that included the word “women,” so whether the candidate played for a women’s soccer team or graduated from a historically women’s college, they were less likely to be picked. 

Although the Amazon recruiting tool was scrapped upon realizing this bias, it shows that it can be easy to unintentionally introduce discrimination into an AI system. In fact, this incident could be seen as one of the more “innocuous” examples, especially when compared to AI use in fields like healthcare, where Black patients were less likely to be enrolled in programs with significant resources, even when they were just as healthy as their white counterparts. In short, AIs pick up the same biases and discriminations as their human designers, becoming a reflection of societal problems. 

To address this, most of the research surrounding gender and AI is focused on women — trans and nonbinary people in tech aren’t even part of the data set — and why there are fewer women in the field. The most common answer for the disparity is the “pipeline problem,” which posits that fewer women are hired in tech because fewer women have tech degrees and experience. Pipeline researchers cite the stereotypes of tech being a “male” field, less early access to computers, and harassment of women. While these may be valid factors, large companies end up using the pipeline problem as an easy excuse for having very few women on their teams. Many of these companies also invest heavily in programs that try to get young girls interested in tech. These programs have been in existence for quite some time, and there has been no evidence that the programs have increased the number of women in tech. Interestingly enough an AI Now Insitute report notes that “91 large tech companies headquartered in Silicon Valley managed to hire higher percentages of black, Latino, and multiracial employees than Facebook that year,” indicating that the issue may not be solely due to a lack of qualified candidates.

Ultimately, the lack of progress in improving the number of women in tech comes down to larger systemic problems that cause “half the women who go into technology [to] eventually leave the field”. The pipeline problem often implies a lack of self-confidence for women who attempt to enter tech, and puts the reason for discrimination on those being discriminated. Instead, systemic issues like exclusionary hiring practices, unfriendly work environments, and tokenization are reflections of the actual problem. This means there’s a huge opportunity for the tech community to stand up and address the issues and the current environment that encourages discrimination.

_ _ _

Pairity is committed to creating a company culture that embraces diversity and inclusion, addresses diversity in our hiring practices and reaches out to “Women in Tech '' communities when sourcing talent. Diversity needs to be present at all levels of a company. Our COO Regina Amundson makes it her priority to connect with other women in tech. With women making up 20% of our engineering team, a female board member, and sponsoring events such as Women in Consumer & Commercial Finance, Pairity will continue to make inroads in 2020 to increase gender diversity in tech.


Want to stay ahead of the game?


Get the latest from Pairity: new products, updates and other news.