Artificial intelligence: made by men for men?

AI might seem neutral and technical but it poses a differential challenge to female jobs and can be imbued with insidious gender biases.

Digitalisation is encroaching further and further into all our lives in ways of which we are scarcely aware. Algorithms and artificial intelligence guide a vast range of society’s choices and opportunities.

Trade unions are alarmed that AI poses a threat to gender equality, by undermining women’s job opportunities and reinforcing stereotypes in the workplace and society, incorporating biases and prejudices into algorithms and programmes. The European Trade Union Confederation is calling for tough measures to ensure women are not second-class citizens in the digital revolution.

There are two main areas of concern, covering women’s jobs and the wider impact of algorithms on their lives.

Challenging for women

Recent research for the International Monetary Fund finds greater automation will be especially challenging for women in the labour market. Hard-won progress towards improving the place of women in the workforce and winning equal pay may be quickly eroded if they work predominantly in sectors and occupations likely to be automated.

The IMF estimates that 26 million women’s jobs in 30 countries (mainly members of the Organisation for Economic Co-operation and Development), have at least a 70 per cent risk of being replaced by technology within the next 20 years. Around 11 per cent of female workers could lose their jobs, compared with 9 per cent of men.

Older women and those in low-skilled clerical, service and sales positions are most likely to be affected. There is a danger too that employers will use the threat of automation to further force down wages and working conditions, compelling workers to compete in a race to the bottom they cannot win.

At the same time, women are under-represented in expanding fields in science, technology, engineering and mathematics—the so-called STEM disciplines—where cultural and social influences have contributed to their exclusion. Globally, women make up 36 per cent of STEM university graduates, yet only 25 per cent of STEM workers (see figure) and just 9 per cent of the decision-makers.

Employees in technical roles (blue male, red female)

According to the IMF, much more needs to be done to boost women’s STEM skills and bridge the digital gender divide, by investment in access and infrastructure, and to close gender gaps in leadership positions and facilitate career progression. For example, attention is needed to childcare and working-time arrangements, as also to the needs of older and low-skilled female workers.

Biased systems

The shortage of female AI professionals carries a further risk—that the digital future will be made by men for men. Biased systems based on flawed data can now influence who gets a job interview or a loan, a promotion or a pay rise. If the people creating the algorithms do not reflect the world’s diverse population, that is a problem, and more female coders, inventors and investors are needed to develop innovations that cater to female needs.

Because existing algorithms tend to treat men as the default and women as atypical, bias and discrimination are in danger of becoming the norm. Women pay the price in time and money—even with their lives. For example, women and men show different symptoms for heart attack, but systems based on male experiences mean that women’s conditions may be incorrectly diagnosed.

There are numerous examples of AI systems behaving badly. Reuters revealed that after Amazon had spent years working on an automated recruitment system it realised that women were being excluded from software developer and other technical posts. Amazon’s computer models used data submitted over the previous ten years, most of it coming from men and reflecting male dominance in the tech industry. In effect, Amazon’s system taught itself to select male candidates and mark down curricula vitae that included the word ‘women’s’—including graduates from some women’s colleges.

Last year Apple faced allegations that its new credit card offered lower credit limits to women than men. Similar problems have dogged Google, IBM and Microsoft AI systems.

In facial-analysis programmes, some algorithms find men’s faces easier to identify than women’s, especially if they are dark-skinned—the computer scientist Joy Buolamwini discovered this discrimination through personal experience. Another headline-grabbing example came from a recent UNESCO report, which found female-gendered voice assistants, such as Alexa, reinforced stereotypes through their submissive, servile responses.

Growing injustice

The use of AI requires ethical and legal frameworks that will counter the risks of gender bias. Without regulation, society will reap the harvest of growing injustice.

All workers need to become ‘AI literate’—understanding its impact and anticipating how it will transform their roles, being able to process and manipulate data and to identify and solve problems. This requires a joint effort, with school systems backed up by employers, being sure to include women, minorities and older people and including AI literacy in negotiations with trade unions.

The digital age should eradicate inequalities at the workplace and bring better-quality jobs and healthier working conditions for all. It should anticipate skills needs and launch reskilling and upskilling programmes as an investment in the future, not a perceived cost. Workers and trade unions should be involved in planning all moves towards digitalisation.

Change in governance

The ETUC insists that AI should be designed to improve the quality of life, safety and prosperity of all Europeans citizens, including women. This requires a change in governance, to involve workers and trade unions in planning, investment and control of technology.

Workers and their unions should be informed and consulted and should participate in the whole process of implementing AI systems, with a guaranteed ‘right of explanation’ when AI is used in recruitment, promotion or dismissal procedures, and a right to appeal against decisions. Facial recognition and tools to monitor performance present risks to women workers in particular and should only be employed with adequate safeguards guaranteed in collective agreements. Workers should have the right to access, and to manage, their own AI-generated personal data.

Tech companies should recruit more women and education systems should encourage more women into STEM subjects. Human beings must have the right to supervise and control AI systems and this must be established as a fundamental legal and ethical principle for the regulation of robotics and artificial intelligence, including in the world of work.

Ultimately, companies can gain a competitive advantage by developing and using AI systems that are trustworthy, ethically and socially responsible, and which respect European values, gender equality and labour rights.