Machine Learning: Mind The Gender Gap

Just 15 per cent of UK technology roles are held by women

D/SRUPTION speaks to Reshama Shaikh about the organisations that are championing change

Despite the prominent role of women in the early years of computer programming, today’s technology gender gap is vast. Data collected by the Women’s Engineering Society in 2017 found that women made up just 15 per cent of STEM employees in the UK. At a senior level, the number was even lower at five per cent.

This dearth of female talent is worrying for various reasons. A particular area for concern is AI and machine learning. If programmes are written primarily by men, then the systems they power could exhibit bias. As artificial intelligence seeps into everyday life, balanced and unbiased code is imperative.

Welcoming women

The lack of females in machine learning, coding, and technology roles in general is down to many different factors. For Reshama Shaikh, data scientist and statistician, one of the main obstacles occurs in formal education where there is shockingly still a belief that STEM subjects are ‘for boys’.

“There’s definitely no doubt that computer science, statistics, maths and all of those quantitative education fields could be more welcoming to people at all levels,” says Shaikh. “I see that from when I attend the ‘regular’ events and conferences, it’s primarily men.”

In 2017, PwC carried out a survey of over 2,000 students in the UK. While 27 per cent of female respondents said they would consider a career in technology, more than a quarter were discouraged by male domination in the sector. Luckily, the growth of organisations and events committed to bringing more women into technology has encouraged gender diversity. Shaikh joined Women in Machine Learning and Data Science (WiMLDS) in 2015 to help to build local communities of female coders and programmers called ‘chapters’, and to organise a range of events geared towards female inclusion.

“We also do workshops, weekend workshops, open source sprints, networking events, and careers events. Over the last four years we’ve seen tremendous growth in the number of chapters of WiMLDS. It was inspired by the International Conference on Machine Learning (ICML) and the Conference on Neural Information Processing Systems (NeurIPS). They have workshops for women that happen within and alongside the conferences but they aren’t accessible to those people who don’t attend the conference.”

As well as WiMLDS, Shaikh is also a committee member of the Machine Learning Conference (MLconf), and the Diversity In Scientific Computing (DISC) community set up by open source project sponsor NumFOCUS. DISC aims to increase diversity in open source through a number of initiatives including the Discover Cookbook, which offers guidelines on how to make events and conferences more inclusive.

Sexist tech

Increasing gender diversity in programming is so important is because, in Shaikh’s words, everyone will consume technology and AI, but the producers of the technology do not always come from different backgrounds.

“Having diverse people involved in creating them will lead to less bias,” she says. “There is also comprehensive research that shows that diversity increases return on investment as well as creating a more supported and healthy society.”

So, gender diversity isn’t simply ‘the right thing to do’. It’s profitable, makes good business sense, and reduces the chance of artificially intelligent systems making ethically questionable conclusions.

Looking for inspiration

In PwC’s Women in Tech survey, 78 per cent of respondents could name a famous male in the technology sphere. In contrast, just 22 per cent were able to name a famous woman. It would seem that part of solving the gender gap in machine learning could be to celebrate female role models.

“There are many women in AI like Rachel Thomas who is the co founder of Fast AI. Another is Erin LeDell, who is actually the founder of WiMLDS. She is the Chief Machine Learning Officer at H2O and she is an organiser for the Bay Area chapter. There is Hanna Wallach, who works in AI for Microsoft and is one of the programme chairs for NeurIPS. There is also Timnit Gebru who is doing critical research on AI, and Anima Anandkumar, a professor at Caltech.

While there are many female role models in the world of AI and machine learning, they don’t get the same recognition as men. This is something that MLconf aims to address by featuring a diverse list of speakers. For Shaikh, finding female speakers is possible if organisers are willing to look outside of their networks and connect with communities.

From gender diversity to general diversity

The diversity dilemma, of course, goes beyond gender. As well as bringing more women into the machine learning arena, underrepresented groups (URGs) also need to be given a voice.

“It started with increasing women, but looking back it maybe should have been something like URGs,” says Shaikh. “There are all different sorts of diversity. There’s gender, sexual orientation, race, age, disability, location… All of those dimensions deserve attention.”

One of the most important considerations when discussing URGs, Shaikh explains, is using the right terminology. Doing so widens the scope for inclusivity, and reaches traditionally underrepresented groups. In a world where machine learning and AI affects all demographics and social pockets, then they too should be given the opportunity to contribute to its development.

For more information about the work of WiMLDS, visit

Read more about the female tech movement in our weekly newsletter.