In honor of Women's History Month and International Women’s Day, the UC Womxn in Technology Committee (UC WIT) hosted a speaker series to address a hot topic on all of our minds, the world of Artificial Intelligence (AI). Far too often, the field is dominated by male voices. This series aims to showcase the significant contributions these women are making to AI development, while also underscoring the challenges that AI has presented (and continues to present) for women, and other underrepresented groups.
Cutting Through the AI Hype
Brandie Nonnecke, PhD., kicked off the speaker series on March 4 with her talk, “Beyond the Hype: Unraveling the Myths, Realities & Governance of AI.” Dr. Nonnecke aimed to “cut through the hype” surrounding artificial intelligence. She pointed out that while we’re hyperfocused on generative AI (like chatGPT), there are other forms of machine learning currently in use, and have been in use for decades - even at UC, which affect the way faculty, staff, and students engage with the campus community.
Dr. Nonnecke explained, “I’ve been in several meetings lately where it becomes clear about ten or fifteen minutes in that when I’m talking about AI, someone else may interpret it as only talking about generative AI tools (like ChatGPT). But there’s a continuum of different types of artificial intelligence, and I actually prefer to refer to it as machine learning, rather than artificial intelligence. AI is not new, it’s been around since the 1950s. Its basis is in algorithms and statistics. If you remember back to some of your statistics courses on linear and logistical regression, a lot of the machine learning (that we use today to make decisions in high-consequential areas) is based on that. So essentially, we have all sorts of machine learning algorithms that are being used, that are essentially surreptitiously making decisions that are affecting our lives every day. If any of you hopped onto YouTube this morning, or Netflix, and it fed you up a recommendation, guess what - that’s machine learning! It’s a recommender system.”
Dr. Nonnecke is no stranger to this subject matter, as she also hosts the video and podcast series, Tech Hype: Debunking Emerging Tech. In the series, she sits down with experts to debunk misunderstandings around emerging technologies, debate the real benefits/risks, and talk about the technical and policy interventions we can implement to better ensure we can harness the technology for good.
AI’s Bias Problem
Dr. Lauren F. Klein was the final guest speaker, on March 7. Her talk was titled, “Data Feminism: A Feminist Lens on Ethics, Diversity & Justice in AI & Data Science.” Dr. Klein highlighted a well-documented issue, bias in AI. She explained that the real problem lies with the data, “All AI, algorithms, and large language models…are ultimately and first powered by data. Many of the biases and inequalities that we see these models reproducing are actually directly connected.”
“Generally speaking, minoritized groups do not have the power to decide what data is collected,” Dr. Klein explained. “This is why…a feminist approach to data and AI needs to begin with this analysis of power. Because far too often, the datasets that we can access, it’s because they’re the ones that have been decided to be collected, and then in turn the models they’re used to train, then the research questions we can ask then, the real-world applications…all of these questions, all of these decisions, have already been over-determined and sort of narrowly scoped, by the imbalance of power in the world.”
“When you hear people talking about the ‘brittleness of AI,’” Dr. Klein added, “They mean the systems that have been optimized for only certain situations, or for only certain groups of users, and not others. But the basic idea is that systems (not technical systems, but all kinds of systems) are optimized for certain groups and not others. This is not news to feminists…the idea that biases are pretty much baked into the cultural record, again - not just the scientific record, not just into algorithmic decision-making, but just sort of baked into the way that we record information about our cultures. This is the ground truth of so much of feminist work.”
Dr. Klein also shared a realization that inspired her book, Data Feminism (co-authored with Catherine D’Ignazio), that data is power, and that power was “wielded by a small and homogenous group of corporations and other well-resourced institutions who had the resources to design and deploy these data systems, mostly for their own profit, and usually at the expense of everyone else. Feminism and intersectional feminism, in particular, has been focused on precisely this issue - imbalances of power and the structural forces that cause them.”
Dr. Brandie Nonnecke explained that the bias is built-in, “We often hear people say AI creates bias. AI does not create bias, All machine learning, AI, does is replicates it and scales it. And for me, that’s actually been sort of a hidden blessing…because you can’t argue against it now. You have the proof that the bias was there. So we can see it now, and we can address it. Now, can we develop a machine learning algorithm that is free from bias? No. Not ever. Never….it’s never about eliminating bias. Even NIST (The National Institute for Standards and Technology) pushes back on using the term ‘mitigating,’ Instead they say, ‘Manage. Managing bias.’ Bias is always present. It’s more about what are the biases that are most important for your application area and what are you doing to address that bias?”
Watch the recordings
The UC Womxn in Tech Speaker Series recordings (and future UC WIT presentations) can be found on the new UC WIT YouTube Channel! Subscribe to our channel, or watch the recordings below.
 |
Beyond the Hype: Unraveling the Myths, Realities & Governance of AI with Brandie M. Nonnecke, Ph.D. director of the CITRIS Policy Lab and associate research professor at UC Berkeley’s’ Goldman School of Public Policy. Watch the recording. |
 |
Data Feminism: A Feminist Lens on Ethics, Diversity & Justice in AI & Data Science, with Dr. Lauren F. Klein, Winship Distinguished Research Professor and Associate Professor in the Quantitative Theory & Methods and English departments at Emory University, where she also directs the Digital Humanities Lab. Watch the recording. |
A big thank you to co-sponsor Van Williams, VP of ITS and CIO of UC whose generous support and partnership helped make this speaker series possible. This speaker series is in celebration of Women’s History Month, a time to honor women’s achievements throughout American history and celebrate women's contributions across various fields - including IT. Throughout the remaining days in March, we encourage you to learn more about the pioneering women who have shaped the tech industry we know today.
Stay Connected
There are several ways you can stay connected to and engage with UC WIT:
-
Reach out to your campus UC WIT representatives to share your story, raise a concern, promote an event, and more.
-
Send us your submissions directly for pertinent news, spotlights on women in tech doing awesome things at your UC location, events that are open UC-wide, podcasts, and videos to add to our website.
-
Join #women_in_tech in Slack, this channel is open to anyone at UC.
Follow our UC WIT LinkedIn page to keep up with news, spotlights, and more related to women in tech across the University of California system.