I am writing this article while my baby daughter sleeps. Like all new parents, her dad and I have spent the last few months in a joy-filled, sleepy haze of getting to know her and imagining what her future might look like. This brings a new intensity, and a little more trepidation, to my role advising on the future of work. What will work look like for this generation of young women, especially as more and more of our roles are being automated — or even replaced — by artificial intelligence (AI)? And how can leaders ensure that AI does not lead to gender bias in their organizations? Recent research is beginning to answer these questions, and the outlook is mixed: on the one hand, women may be spared from the job disruptions men will face in the longer-term. On the other, the lack of gender diversity in AI-related jobs could be reflected in the tools that are created, affecting whether women are hired or promoted.
First, the impact of AI on work will be influenced by the distribution of women and men in particular jobs. While an AI tool may not be designed to replace the tasks of women or men in particular, many occupations are so skewed in their current distribution that waves of automation may be felt more by women, or by men, at particular times. Bureau of Labor Statistics data show that there’s an unbalanced gender distribution among the most common jobs in the U.S. today. Jobs such as elementary and middle school teachers, registered nurses, and secretaries and administrative assistants each comprise at least 80% women; while jobs such as truck drivers and construction laborers employ more than 90% men.
Because AI tools will tend to automate tasks, rather than whole jobs, many occupations will be affected unequally. While the gender distribution of occupations may shift over time, PwC has estimated that more women than men will be affected by job changes between now and the late 2020s. This disproportionate impact on women is based largely on the high number of women employed in clerical occupations: in the U.S., for example, 94% of secretaries and administrative assistants are women. These kinds of roles are being disproportionally affected by technological developments like automated assistants, and smarter email, calendar, and financial software.
This picture changes over the medium-term. As new AI capabilities develop, such as self-driving technologies, more men than women will be affected by job changes between the late 2020s and the mid 2030s. During those years, automation is predicted to lead to job losses in what are currently male-heavy industries, such as construction and transportation. Employers should be thinking about this job redistribution in advance, to help ensure that a wave of redundancies following technological change does not lead to a sudden worsening in organizational gender balance. This could mean slowing down job losses to enable the organization to adjust. Aiming for gender parity in those areas in which jobs are more secure, such as management roles, becomes all the more important.
Second, consider that women’s current representation in jobs related to AI is unequivocally poor. According to 2018 data from the World Economic Forum and LinkedIn, only 22% of jobs in artificial intelligence are held by women, with even fewer holding the most senior roles. This is an important disparity, because those who learn about, experiment with, and implement AI technologies will be creating the tools that organizations use on a day-to-day basis — and any unconscious biases baked into their decisions they make could have serious consequences. For example, more and more HR departments are using algorithms to help sift through resumes, conduct interviews, determine pay, and spot performance problems. These tools are often intended to be more objective than human decision-making, but they can easily go awry. For example, Amazon abandoned its AI recruitment tool after discovering that it showed preference for male over female candidates.
Leaders of organizations using AI tools can help prevent the use of gender-biased tools by encouraging diverse technical teams wherever possible. Having more women developing tools may help teams spot unintentional gender biases, like training an algorithm on historic data that reflects gender inequality in who is hired or promoted. Leaders should also regularly check the completeness of tests used to detect gender bias. That’s because a resulting tool can still produce different outcomes for women and men even when an algorithm has been trained without using gender as a data parameter. In the case of resumes, a gap between jobs or a longer period without promotions may be treated by an algorithm as negative indicators, but could be for reasons unrelated to work, such as a mother spending more time at home around the birth of children. A tool that gives fair advice about hiring, performance, promotion or pay based on resumes should provide the same answers about men and women of equal competence, without assuming that male and female resumes will always look the same.
What does this all mean for girls like my daughter, who will be entering the workforce in two decades or so? There are substantial risks to navigate in the coming years, especially when women are judged using tools built on data from the world as it is, rather than the world as it should be. Leaders should do their own checks to ensure that the AI tools that their organizations are using are helping to reveal female talent, rather than accidentally overlooking it.
At the same time, the under-representation of women in science and technology roles is occurring alongside an over-representation of women in the kinds of roles that require emotional intelligence and advanced communication skills, such as speech pathologists, preschool teachers, or occupational therapists, to name a few. As skills such as empathy and collaboration are among those that are hardest to recreate in AI tools, many of these occupations are likely to be safer from technological disruption. Looking ahead, one happy possibility from the rise of AI is that people’s ability to understand one another and work together may become more valued as technological tools overtake us in other areas. My optimism also has me wondering whether, as workers gravitate towards the safest roles, there may be greater gender balance in jobs that have traditionally been dominated by men or by women. If so, this opens a greater variety of choices — and the possibility of greater job satisfaction — for both our sons and our daughters.
Emma Martinho-Truswell is the co-founder and Chief Operating Officer of Oxford Insights, which advises organizations on the strategic, cultural, and leadership opportunities from digital transformation and artificial intelligence.
Read more on hbr.org