Erasing hiring bias from AI
In the midst of the 100th anniversary of the creation of the U.S Department of Labor’s Women’s Bureau and the passage of the 19th Amendment, I got thinking about how far women’s rights have come, and just how far we have left to go.
Hiring bias is a crisis that causes fundamental societal issues and creates deep inequalities based on implicit and discriminatory factors. It is high time to focus on the merit of the candidates regardless of inherent factors such as gender, race, and sexual orientation. Hiring people based solely on qualifications will not only benefit the economy but our society as a whole.
While I’ve had extremely positive work experiences — all with inclusive and welcoming teams — I have at times been surprised at the lack of representation. In one experience, for example, I was the only female in a sizable team.
The subconscious bias we all hold as humans not only affects how we like to dress or the television we watch, it also affects those with whom we choose to surround ourselves, often with those similar to ourselves. This preconceived bias may seem to have no immediate detrimental effects, but it can often lead to subconscious discrimination, especially when selecting job candidates.
So how can we make sure our inevitable implicit bias doesn’t affect opportunities for others? For starters, we use unbiased Artificial Intelligence (AI) software to make decisions for us. For any company that wants to create an equal company culture, integration and investing in totally unbiased software should be at the top of the to-do list. Sounds simple, right?
Most people don’t immediately think subconscious bias in a company’s culture is a dangerous threat to the wellbeing of our society. Yet, with the unemployment rate reaching an all time high of 14.7 percent in April and a limited number of jobs available, it has become even more crucial for companies to strive for an equal and diverse workplace. If we want to improve the economy, diversifying the workplace is paramount to that goal, given that gender-diverse companies are more likely to perform 15 percent better, and ethnically-diverse companies are more likely to perform 35 percent better.
In theory, it’s an easy fix to this crucial dilemma: a software that could eliminate the risk of implicit bias having negative effects on choosing the right candidate for a job. Yet, one imperative problem remains; the AI currently used in job recruitment has been proven to sometimes be just as biased as humans themselves. The current AI technology may be eliminating diverse candidates before they even get the chance to talk to a human. It’s not such a simple solution after all.
In 2018, Amazon made headlines that the computer programs that had been implemented in 2014 to scrape job candidates had a significant bias against women. The AI was trained based on resumes submitted to the company over the past 10 years, most of them being men. Even when Amazon changed the way the AI responded to women-related key terms, there was no guarantee AI wouldn’t find alternative ways to discriminate against certain candidates, proving more extensive research still needs to be done on how to create a neutral AI tool.
Many conglomerates such as Amazon have tried (and failed) their hand at implementing AI into their hiring systems. Designers of hiring AI need to have social accountability rules for better data sets. Just as Facebook is accountable for content posted on their website, AI developers need to be held accountable for producing more diverse data sets. Because of this lack of good hiring tools, a lot of companies implementing the current technology are just using it for the sake of having it, with no real effect on their hiring process.
There is no denying that the use of AI will soon be integral to recruiting and creating entire work forces. According to a CareerBuilding study, 55 percent of HR managers believe AI will be a regular part of the hiring process. If AI is destined to control HR functions and tasks, an unbiased software is absolutely essential, as the risk of discriminatory gaps within the workplace are at risk to increase.
AI may not be a silver bullet for creating a completely diversified work environment, but if researched and implemented the right way, it will go a long way in eliminating bias in the workplace. While training for humans will remain necessary, companies investing in refining AI will have better results in getting the right mix of individuals through the door. We as humans may never reach a society rid of discrimination and injustice, but with enough development and support, our AI counterparts just might.