The lack of representation of women and minorities in STEM fields maintains gender segregation at a later stage in education and the labour market

Furthermore, the exclusion of education from the scope of Directive 2004/113/EC is problematic in light of the under-representation of women in STEM fields and curricula related to IT and software development, which is specifically discussed in section 3.2.6. The lack of representation of women in these fields of education maintains gender segregation at a later stage in the labour market and leads to a lack of diversity in software developers and programmers, which in turn means that algorithms fail to reflect a variety of gender perspectives. The lack of EU legal guarantees against discrimination on grounds of age, disability, sexual orientation and religion is problematic for the same reasons. This lack of voice of women and minority groups in algorithmic design has clear repercussions in terms of biased algorithmic design leading to discrimination. For example, the case of Dr Selby illustrates how harmful gender stereotypes crept into the design of a piece of commercial software giving automated access to changing rooms in a fitness studio.206 Because Dr Selby’s title was ‘Dr’ and not ‘Ms’, she had been classified as a man and could not enter the women’s changing rooms. This type of mistake in the design of algorithms could be corrected through increasing the diversity of the workforce, and thus the representation of various minority perspectives, in relevant areas of the labour market. This shows that the sources of algorithmic discrimination lie as much in the functioning of algorithms as in their human design.

3.2.6 A gender digital gap in European countries Finally, specific diversity problems have been reported in nearly all countries studied in this report both in STEM-related educational curricula and in STEM-related professional communities. In particular, the IT sector struggles to attract and keep a female workforce. This reflects the wider context: in 2017, only 12 % of AI researchers in the world were female.369 Several interacting explanations for this can be identified, including gender stereotypes, gender segregation in working life, the exclusion of women from IT jobs through discrimination and the lack of role models for female IT workers. The statistics and figures below offer a picture of the situation in the EU and EEA countries to date. In Austria, STEM curricula had a 34 % share of female students in 2016 against, on average, 61 % in other curricula.370 In particular, only 16 % of IT students were female. After their education, the integration in the labour market of IT graduates also differs, with a 94 % figure for males and a lower figure of 88 % for females, as well as a 7 % difference in male and female salaries. Statistics show almost no change in female participation in STEM education since 2007/2008. In Croatia, the share of women employed in the ICT sector is 13 %, which is below the EU average of 17.2 %.371 In Denmark, women make up only 24 % of IT employees overall.372 In the Netherlands, a study found the number of female students in STEM subjects in 2019 slowly increasing, while the number of women in technical positions increased by 24 % between 2013 and 2019.373 In Estonia, in 2019, women represented only 22 % of the ICT sector workforce.374 The expert for Estonia has reported that there is a low share of women among programmers due to widespread acceptance of the myth that programming is a man’s field.375 Similar trends are noticed in Finland, where concerns have been expressed about the lack of attractiveness of IT education and professions for women and the fact that educational choices are highly gendered.376 Germany also faces a significant underrepresentation of women in IT education as less than 20 % of all IT students are female.377 Researchers also report a digital gender gap in Greece and note that digital exclusion is reinforced when gender combines with other exclusionary factors such as disability, age, race and socioeconomic background.378 This problem was noted by the Greek Minister of Education in 2020 in an article entitled ‘Women in science: A bet that we have to win’, in which she attributed the underrepresentation of women in STEM worldwide to gender stereotypes and segregation in the job market, which affect women’s career choices and chances, as well as the pay gap, unequal access to funding and unequal family responsibilities.379 Similar concerns have been expressed in Luxembourg, which has been reported as having one of the lowest rates of women’s participation in IT and STEM sectors.380 The number of female Polish IT students is increasing but remains low (14.6 % in 2018- 2019). A breakdown of this statistic indicates that women represent 40 % of the student body in IT fields pertaining to data analysis and processing, 32.1 % of the total number of students in computer science and econometrics, and only 8.9 % in industrial computer science studies.381 The proportion of women studying IT majors has increased in Slovakia from 3-5 % to 10-12 % in the last five years, but IT professions remain insufficiently attractive to women.382 In Spain, data shows an important gender gap in ICT related professions and in women’s access to ICT studies.383 For example, only 15.6 % of ICT professionals were women in 2017, and this number is decreasing, while the number of male ICT professionals is growing.384 In other words, women in ICT account for only 2 % of total female employment in Spain.385 At the source of this problem lies the persistent absence of women in ICT-related education and training. Numbers have been stagnating, with 33 % of women in ICT-related training in 1999 compared to 37.4 % in 2017.386 In technological university and non-university degrees in Spain, the gender gap has also been increasing, reaching 12.6 % in 2017. Finally, in the United Kingdom, women made up only 16 % of entrants to undergraduate courses in engineering, technology and computer science in 2018-2019, compared to 56 % of total female undergraduate entrants.387 The House of Commons Science and Technology Committee and the House of Lords Committee on Artificial Intelligence have recognised the problem and acknowledged that diversity among professionals developing algorithms is a key tool in tackling algorithmic discrimination.388 The underrepresentation of women and minorities in STEM educational and professional communities has been widely recognised by scholars as problematic. Practical examples provided in the literature include driving machines or weapons that are trained only by men’s voices and thus perform less well in recognising female voices, and assistant chatbots called after female names and welcoming/serving robots with feminine appearances, which can perpetuate harmful stereotypical perceptions of women’s roles in society and in particular the typical association between female and caring/assisting activities.389 It has also been pointed out by scholars that a lack of diversity in data and design leads to systems biased against underrepresented groups.390