Reuters reports that Amazon tried to build an AI tool for recruiting, but after it showed bias against women, it had to be scrapped.
Amazon had been building the software since 2014 to make narrowing down top candidates more efficient. The tool would give scores from one to five stars to each of the candidates without a human looking at the applications.
Think of this as the Google of recruiting, algorithms rank the candidates and create a list, then the hiring manager looks at the top 5. Sounds ideal – the recruitment process can be expensive and time-consuming, what could possibly go wrong?
In 2015, the company discovered that the AI was showing bias towards male candidates for their technical jobs. The reason behind this was that the computer model was trained to observe patterns using resumes sent to the company in the past 10 years. Most of the resumes happened to belong to men, not surprising considering the male dominance in the industry. Not only that, but resumes that included the word “women’s” and graduates of all-women’s colleges were penalized.
Other than the gender bias issue, unqualified candidates were still passing the AI software and being recommended for jobs where they would not be a good fit. After recognizing these problems, Amazon decided to shut down the project.
There is a gender imbalance across the tech industry, with the majority of Amazon, Facebook, Apple, Google, and Microsoft employees being men. Amazon, interestingly enough is the most gender-balanced of the tech giants with women making up 40% of their workforce. However, numbers were not available for how many of their employees in technical roles are women. Based on stats for the four other companies, the numbers look bleak ranging from 19-22% of their technical workers being women.
Here’s the question: Why are fewer women applying to work at Amazon? Is it Amazon’s fault that most of their applicants for the tech positions have been men? At the university level, men outnumber women in computer science degrees, so it is to be expected that male applicants would outnumber female applicants in tech jobs.
Why are fewer women studying and working in computer science today? Computer science was not always predominantly male. According to Smithsonian Magazine, male programmers wanted their jobs to be taken more seriously, and so they created professional associations and discouraged the hiring of women.
However, these biases didn’t disappear after anti-discrimination legislation was passed. In preliminary recruiting sessions, women often feel discouraged from even applying for STEM degrees. Stanford researchers documented that women felt unwelcome because of sexist jokes, a lack of women engineers, female attendees not being introduced to recruiters, and male students dominating Q&A sessions.
Could the reason fewer women study computer science begin while a girl is in primary or secondary school? A Guardian article mentions that fewer girls than boys choose to study STEM subjects at secondary school. It also mentions a study of about 1,300 Swedish secondary school students and the reasons why they pick the subjects they do. That study found that teenagers felt they would fit in better where their gender was more represented and because of that they felt they had a better chance of succeeding.
Unfortunately, gender equality can’t be reached in computer science overnight. It’s going to take at least a generation to get more girls in STEM majors and therefore more women applying for STEM jobs and working in the field. But how do we get there? The solution isn’t to force women to study science or put in quotas, but to encourage all students to study what they want and achieve their goals and that their gender doesn’t affect their science or math abilities. Schools and companies should also take any accusations of gender bias seriously.
As for the AI software side of the situation, Human Resources has human in there for a reason. It doesn’t make sense to automate it entirely. Humans are far from perfect and they have their biases, but machines are rigid, and there’s not much room for discretion, which is an essential part of the hiring process. It’s not just about qualifications, personality is key.
That being said, AI in recruitment is nothing new, and it’s looking like this is the future of Human Resources. Undercover Recruiter reports that AI is predicted to replace 16% of HR jobs in the next decade.
Apply to any large company and you’ll be directed to a portal where you fill out forms, maybe even a personality test, and attach documents which are scanned for keywords relevant to the job. After you submit the application, it gets processed through the software and gets sorted into different categories: to interview or to reject. Sometimes candidates get rejections within hours or even instantaneously. Ouch.
Hiring is something that is best done with a human evaluating all applications and making the decision. To make it more equal and prevent bias, one possibility is to give applications a number and hide the name of the applicant. People like when companies treat them more personally, and AI doesn’t have the same personal touch. If Amazon’s failed experiment has taught us anything, it’s that AI for hiring needs a lot more work.