Nesta, the innovation foundation has today published a large-scale analysis of gender diversity in AI research. The report found that:-

• There is a serious gender diversity crisis in AI research with just 13.8% of the authors being women, compared to 15.5% for STEM subjects, with numbers decreasing over the last 10 years.
• The proportion of AI papers co-authored by at least one woman has not, in relative terms, improved since the 1990s.
• Apart from the University of Washington, none of the top 35 institutions for AI research have more than 25% listed as being authored by women.
• Only 11.3% of Google’s employees who have published their AI research on arXiv are women, 11.95% of Microsoft employees and 15.66% of IBM employees.
• The UK lags behind the international community for percentage of females authoring AI research, coming 23rd out of the top 34 countries publishing on arXiv, behind countries such as Norway, Turkey and Malaysia.
• Women are more likely to consider societal, ethical and political matters in their work on AI.

Artificial Intelligence (AI) technology is changing our world, but those producing these cutting-edge systems are predominantly male. Not only does the report show a severe gender diversity gap in AI research, it also shows that, in relative terms, the proportion of AI papers co-authored by at least one woman has not improved since the 1990s.

The report, which examined publications on arXiv, a repository with more than 1.5M preprints widely used by the AI community, also revealed just 18% of Oxford's researchers with AI publications on arXiv are women, this falls to 15.6% for Cambridge.

Joysy John, Director of Education at Nesta said “This is not simply an issue because of the lost talent of capable women; it is also a much wider problem. Future technology will not be able to meet the needs of a diverse population if it is being shaped by a small section of society with a singular worldview. AI is a powerful tool that can be used for good but also has the potential to be misused. Whilst it could be used to improve healthcare, the education system and public services, it also has the potential to be used for mass surveillance, online propaganda and entrenching pre-existing biases and stereotypes.”

The report shows that women are more likely to consider societal, ethical and political matters in their work on AI. The analysis reveals that papers in the UK in 2012 and 2015 with at least one female co-author tend to be more applied and socially aware, with terms such as fairness, human mobility, mental, health, gender and personality being among the most salient ones.

Joysy John continued, “This work from Nesta aims to expand the evidence on gender diversity in AI research, and create a baseline with which to interrogate the impact of current and future policies and interventions. By engaging with key figures in the industry, we also hope to develop a greater understanding of the cultural and institutional factors that determine gender diversity in AI research, share best practice from schemes that work and raise the profile of leading female figures.”

Mihaela Van Der Schaar, the female AI researcher with the most number of publications in the UK on arXiv, and the John Humphrey Plummer Professor of Machine Learning, Artificial Intelligence and Medicine at the University of Cambridge said, “'My original training was in Electrical Engineering, a field that was, at the time, almost completely populated by men. (I was the only woman among 200 men in my undergraduate years.) I began working in AI in 2003 when I moved from industry to academia, but my presence in the field and much of my early work in this area have only been recognized recently. I think that part of the reason for this is because I am a woman, and the experience of (the few) other women in AI in the same period has been similar.

“This disparity of recognition between men and women is slowly changing but there’s a lot more that needs to be done, as this report shows. We need to highlight that AI is now transforming all industries, not just engineering and computer sciences but everything from health to education. Publicising the interdisciplinary scope of possibilities and career paths that studying AI can lead to will help to inspire a more diverse group of people to pursue it. And, on other side, the industry will benefit from a pipeline of people who are motivated by combining a variety of ideas and applying them across domains.”

Sir Alan Wilson, Executive Chair of the Ada Lovelace Institute and Director, Special Projects of The Alan Turing Institute, where he was CEO between 2016-2018, said, “From my experience, it’s precise and targeted interventions that work in reaching underrepresented groups. Unless a scheme is focused on increasing diversity, it won’t achieve it. There needs to be further research into the kinds of interventions that can inspire young women and minority groups to pursue study all subjects that lead to careers in AI - computer science, but also maths, statistics and engineering. Research projects like this are important to benchmark these pivotal fields of work and more data sources should be made available.”

The University of Washington was the only one out of the top 35 institutions for AI research on arXiv to have more than 25% listed as being authored by women.

Eve Riskin, the associate dean of diversity and access in the University of Washington’s College of Engineering said, “I’ve often heard women say they feel they don’t belong when studying male dominated subjects. This is commonly known as ‘imposter syndrome.’ Other undergrads will tell them they only got onto the course because they’re ‘a girl.’This is simply a toxic environment. There need to be schemes to encourage both women on the faculty and more female students. This is what we’ve been working on at University of Washington through schemes such as working with our department chairs on cultural change while providing professional development to our female faculty in STEM. Research has shown that female undergrads achieve more under female faculty, and so you need a two-pronged approach. What are the main barriers? Funding and time. Cultural change is a lot of work and will need high level, committed staff to apply for funding and drive change.”

Ed Lazowska, the Bill and Melinda Gates chair in computer science and engineering at University of Washington said, “Unfortunately, there’s no silver bullet to achieve diversity. It takes lots of little things. You have to create an environment that’s friendly for women, actively encourage women to apply for positions on the faculty and reach out to potential students. You also have to suppress all kinds of harassment. Sexual coercion and unwanted sexual attention of course, but also gender harassment. This can include talking down to women or questioning their qualifications. Even if done subtly, or without the perpetrator even realising they’re doing it, this can be very harmful. You have to transform the whole culture of an organisation and this needs to be driven from the top, with leadership embracing it as a priority.”

-ENDS-

For more information contact Juliet Grant in Nesta’s press office on 020 7438 2668 or 07866 949047, [email protected]

Notes to editors:

About Nesta
Nesta is an innovation foundation. For us, innovation means turning bold ideas into reality and changing lives for the better. We use our expertise, skills and funding in areas where there are big challenges facing society. We've spent over 20 years working out the best ways to make change happen through research and experimenting, and we've applied that to our work in innovation policy, health, education, government innovation and the creative economy and arts. Nesta is based in the UK and supported by a financial endowment. We work with partners around the globe to bring bold ideas to life to change the world for good.

Report Methodology
We conduct a large-scale analysis of gender diversity in AI research using publications from arXiv, a widely-used preprints (a version of a scholarly or scientific paper that precedes formal peer review) repository where we have identified AI papers through an expanded keyword analysis, and predicted author gender using a name-to-gender inference service. We use Google’s cloud services to geocode academic institutions and we developed a machine learning system to identify semantic differences in a collection of documents. Lastly, we conduct a correlation analysis to examine if there is a link between the number of paper citations and female co-authorship.

We only present results for publications where the name-to-gender inference service was more than 80% certain in the name matching. Inferred genderisation assumes that gender identity is both a fixed and binary concept. We acknowledge that this limitation restricts the scope of our analysis to binary genders.