Why AI will come for women’s jobs first

As the AI frenzy continues to upset the job market, women are more likely to be affected by AI-induced disruptions and set to lose their jobs, according to a government policy paper.

Presented by the UK Parliament of Science and Technology, the rapid adoption of AI could disproportionately affect female workers. The report highlights that 79% of working women are employed in occupations deemed most susceptible to AI automation.

The healthcare and education industries are the most likely to be impacted. In these sectors, 76% and 67% of the workforce respectively is made up of women.

With no current legislation dedicated to the regulation of AI in the UK, experts have recommended that the government should prioritise retraining and upskilling workers to ensure inequalities are not exacerbated.

Mind the AI gender gap

The emergence of new technologies can kickstart a cycle that eliminates jobs, only to pave the way for fresh ones. Yet, the issue with the AI gender gap is that women could then struggle to be picked for the roles AI will create.

Evidence suggests that stubborn structural inequalities persist in the data science and AI fields. Here, women are more likely than men to occupy jobs associated with lower status and pay.

According to figures from the Alan Turing Institute, women make up half of the UK population, yet only 22% of AI and data professionals are presently females.

The issue persists when zooming into turnover and attrition rates, as women tend to display higher frequency for both in comparison to their male counterparts.

This trend is also hurting the UK’s standing, as its tech sector ranked fifth place in the Women in Technology Index for the G7.

The risk of exacerbating inequalities worsens when factoring in automated hiring – the fewer the number of women employed within the sector, the greater the potential for the algorithm to reinforce gender biases.

The AI gendered feedback loop

With AI boardrooms increasingly becoming more unrepresentative of the population, experts warn there is a risk that biases and their replication will become ingrained in the datasets used to train generative AI.

As stated by the European Commission, “Technology reflects the values of its developers. It is clear that having more diverse teams working in the development of such technologies might help in identifying biases and prevent them.”

Several AI products have made headlines for their discriminatory outcomes due to algorithms and datasets that can be susceptible to bias.

The image-generation algorithm of ChatGPT and Google’s SimCLR were found to be more likely to autocomplete a cropped photo of a man with a suit, but a woman with a bikini. Similarly, marketing algorithms have disproportionately shown scientific job advertisements to men.

While this dilemma is not unheard of in any boardroom, the Alan Turing institute notes that technical bias mitigation and fairness metrics are by no means sufficient to correct discrimination.

This is because fairness cannot be mathematically defined, and is rather a sociological issue. The task of mitigating biases can become the responsibility of developers themselves, who in turn are at risk of unconscious bias in a field where representation has been shown to be far from even.

An AI ego gender gap?

The AI gender gap goes beyond the confines of the workplace, and rather, starts in the classroom. According to the Higher Education Statistics Agency, women comprised 23% and 20% of higher education students on computing and engineering courses respectively in the 2021-2022 academic year. This disparity is also reflected in the faculty, as men constituted an average of 80% of AI professors.

Beyond the quieter presence of females in AI courses and boardrooms, statistics hint there might be an emerging gender ego gap.

According to the same study by the Alan Turing institute, women working in AI and data science have higher formal educational levels than men across all industries, including tech. In fact, 59% of women working in those fields hold a graduate degree, compared to 55% of men.

However, men routinely self-report having more skills than women on LinkedIn. This suggests women might have lower confidence in shouting about their technical abilities despite being highly qualified.

The future of women in AI

While multiple structural barriers are complicating gender parity in AI, the UK government has a strategic interest to champion equality to become a leader in the tech sector.

According to research by Ipsos Mori of 118 UK public and private sector organisations using AI or developing AI-led products, 62% of respondents cannot meet their goals because job applicants and existing staff lack the skills needed to work with AI.

On the education front, a 2021 study estimated that the supply of data scientists from UK universities was unlikely to exceed 10,000 per year. However, there are potentially at least 178,000 data specialist roles vacant in the UK.

If more women can make it into AI courses and not feel daunted by the prospect, it’ll be easier to meet the UK’s AI skills demands and nurture workplaces that wave the flag of gender parity.