Why are measures to promote women in AI failing?

28 Oct 2022

Image: © Wanlee/Stock.adobe.com

The Dublin city lead for Women in AI Ireland says current measures don’t address all the obstacles to progression faced by women – especially parents.

Women are underrepresented in artificial intelligence and technology in both academia and industry. This is not news. The disparity has been discussed for years now. Yet, surprisingly, little progress has been made.

According to the Stanford Institute for Human-Centered AI’s 2021 AI Index Report, just 16pc of tenure-track faculty focused on AI are women across the world’s top universities.

We have taken some positive steps in Ireland, especially since the 2016 review of the HEA’s Expert Group on Gender Equality, after which came measures such as the HEA linking research funding eligibility to progress on gender equality. Institutions in turn have established equality, diversity and inclusion units. There has been improvement, although there is still some way to go in relation to the number of female professors in our universities.

As a woman working in the field, I am often asked about these challenges. Many proposed solutions rightly or wrongly focus on women, encouraging them into the sector.

For example, I do a lot of work demystifying my job for students thinking about a career in STEM. I use my position to support and mentor colleagues, buoying people up as much as possible, often reassuring them that it is okay to say no. I accept invitations to speak on panels and to diverse groups in an effort to encourage more women into this area.

I do this because I believe that greater representation of women in AI and other STEM fields will have positive knock-on effects for other women, by engendering changes in systems and practices. However, while these measures are valuable, it is all for nothing if we don’t meaningfully address the systemic barriers that are in place.

Change the system

If we truly value diversity – and diversity has huge, demonstrable value – we need to change what the system sees, and how the system works.

We have taken some positive steps. Irish Research Council funding applications are now gender blind, for instance. This has directly led to more applications from women being funded.

Diversity is now a requirement for interview panels. As such, I am often asked to sit on such panels. Does my presence improve things? I suppose it does to a degree, but we are still faced with the inherent problem of value.

We data scientists and AI experts tend to like a formula when assessing interview candidates. Formulas are fair. They are dispassionate. They don’t make value judgements based on gender. Right?

I am not so sure. Take the common occurrence of maternity leave. A recent, positive step forward is that in some areas now there is recognition of the need, within a PhD, for allowances for female students who need maternity leave. The European Research Council, for example, has introduced an extension for people who have children to account for the career impact of time dedicated to childcare. For every child, a woman will have an extra year and a half of eligibility to apply for post PhD funding. This is welcome. It is a good move in terms of our interview formula. But it is also incomplete.

While the interview formula now accounts for the extra time needed to catch up on the work and outputs, it does not allow for the fact that someone taking time out for maternity leave or parenting duties will have fewer publications, fewer conference speaking engagements and lower impact numbers. These are the things that interviewers score candidates on. These are the things that are valued.

Often, when it comes to the formula, female candidates do not stack up. In fact, I believe that we have lots of highly skilled women who simply don’t even make it to the interview stage because their CVs are directly compared to their male counterparts, and it takes longer for a women to play catch-up after maternity leave.

Not a level playing field

Essentially, in trying to be fair, we are compounding an unfair process for our female academics. It is not a level playing field. To some extent we can address this through mentoring and training, but we also need to be more creative in our approaches.

If you had asked me at the beginning of my career whether I believed in quotas or female-only posts, I would have said no. It should all be based on merit. I still believe that, but I don’t think that is what we are currently getting.

A recent paper published in Nature’s career guide detailed the experience of the Department of Maths and Statistics in the University of Melbourne. They undertook a limited affirmative action strategy to recruit more women into their faculty as time and patience had done very little to increase gender diversity in 20 years. The approach was temporary and, in one round of recruitment, only women were eligible to apply for jobs in areas where women were underrepresented.

There was a number of interesting facets to this intervention. The positions, for example, were designed to attract a broad cross-section of female applicants, rather than just those in particular sub-disciplines. They were also ongoing teaching and research roles rather than fixed-term contracts which is another issue that arises when examining the attrition of women from academia.

The outcome of this approach was encouraging. The authors of the paper noted that the vacancies attracted top quality candidates, and that the recruitment round resulted in an upward shift in the numbers of female staff in areas where the trends had been stagnant or negative in the previous years. Added to that, the trends have continued upwards, despite the period of affirmative action having passed. The authors state that the period of action sent a positive message to women that this was a faculty that was serious about diversity.

The paper goes on to discuss the challenges and the conversations that need to be had to ensure that people don’t feel disenfranchised by such action and the importance of such action being legal and strongly supported from the top.

A new review of gender diversity in higher-level institutions in Ireland was announced in March of this year. There does seem to be an acknowledgement among the experts about the need to be creative, honest and open in order to achieve not only gender parity, but also a true meritocracy in our system.

We know that women add value. Diversity adds value. We should be striving to bring in any measures that open the gates to those people who are a big part of our society and bring valuable new perspectives. There is no compromise in doing that.

By Dr Georgiana Ifrim

Dr Georgiana Ifrim was recently appointed Dublin city lead for Women in AI Ireland. She is an associate professor at the School of Computer Science in University College Dublin and a Science Foundation Ireland funded investigator with the Insight research centre for data analytics.

10 things you need to know direct to your inbox every weekday. Sign up for the Daily Brief, Silicon Republic’s digest of essential sci-tech news.