Fair AI needs interdisciplinary effort
Interdisciplinary research is cutting edge, exciting and - as you will read below - growing in importance. But not having a “pure” discipline can pose challenges for postdocs too. Dr. Susan Leavy, recently appointed Assistant Professor at UCD School of Communication and Information Studies, explains how UCD Institute for Discovery helped her on her career path.
Dr. Susan Leavy describes her career as “a long and winding road” but having recently taken up an Assistant Professorship she reckons it has “finally made sense”.
A Degree in Philosophy and English, a Masters in Artificial Intelligence, an MPhil in Gender and Women’s Studies, a PhD in Computer Science - and ten years spent in investment banking designing and managing trading platforms in between - hers is a diverse portfolio of talents and interests.
Her research now looks at “Fair AI” and developing a platform to mitigate gender and racial bias in data sets for machine learning. She is also preparing to teach AI in UCD, “a course that’s going to be accessible to people who’ve never written a line of code”.
Though it is “exciting to work in an emerging field”, Dr. Leavy points out that “it can make career advancement more challenging”.
It is a familiar story to UCD Institute for Discovery Manager Tara Byrne, who first met Dr. Leavy in 2018 at a Cyber Lab networking event organised by Dr. Colm McMahon, Senior Manager at UCD Research and Innovation.
“We just got chatting. Susan was working as a postdoc for both the UCD School of English and the UCD Insight Centre for Data Analytics. She talked about how difficult it is to be a really interdisciplinary postdoc because you’re up against people in one pure discipline who might have published a lot more in top journals. I knew she was a very good fit for Discovery.”
The Institute for Discovery specifically supports interdisciplinary researchers like Dr. Leavy.
“Susan mentioned that she was thinking of doing a workshop on ethics and AI. I talked to her about our Global Visiting Fellowship bursary and about how we could support her workshop by bringing in a keynote speaker. That’s how we got started.”
Dr Leavy already had an amazingly interdisciplinary keynote speaker in mind: Patrick Lin, director of the Ethics + Emerging Sciences Group at California Polytechnic State University, where he is also a philosophy lecturer. She invited him informally when they met at the Paris Global Forum for AI and Humanity in 2019 - and afterwards UCD Discovery issued the formal invitation, which he accepted. Dr Leavy’s Fair AI event - now postponed until Spring 2021 - will bring together thought leaders from academia, industry and government to explore this pressing issue. It will also be another networking opportunity that might lead to more interesting collaborations.
“When I met Tara at the Cyber Lab event I found out about UCD Discovery and what they do. I found Colm McMahon, Tara Byrne and Elva O’Sullivan [UCD Discovery Engagement Manager] a fantastic help and very encouraging,” says Dr. Leavy. “It was amazing to discover that these people are there to help you out. You’re so used to academia being a solitary enterprise; you’re kind of off on your own, forging your path and it’s such a complex world. It’s really hard to understand funding, where it comes from, how to apply. It’s wonderful to be able to meet with people like Tara, to get encouragement and advice. They’re definitely key people in my career.”
Cross-disciplinary efforts will become more prevalent as we try to stay ahead of artificial intelligence.
“You can’t keep loading more responsibility for ethics and data governance on machine learning engineers,” explains Dr. Leavy. “They can’t do it all and we can’t educate them to do it all. The systems are just so complex now but at the same time to impart real intelligence to systems you need people from different backgrounds and genuinely diverse teams.”
Biases can seep into these algorithms unless they are compressively critiqued and evaluated. Companies risk reputational damage and legal action if they build systems that discriminate.
“AI teams will need people with expertise in social science, law, humanities, statistics, maths, computer science and more.”
Shortly before the Covid-19 lockdown Dr. Leavy took up her new position at the School of Information and Communication Studies.
“I was extremely lucky because it is a genuinely multidisciplinary school and possibly one of the only ones. Interdisciplinary research wouldn’t suit everyone but I enjoy it,” she says. “It’s where you can find new perspectives and bringing people together from different backgrounds can lead to very exciting projects and ideas.”
This article was brought to you by the UCD Institute for Discovery - fuelling interdisciplinary research collaborations in UCD.