On March 7, 2022 more than 10,000 people from 99 countries participated in the Women in Data Science (WiDS) hybrid conference held in-person at Stanford University, and online in a virtual event platform and via livestream. The in-person conference included keynote addresses, technical vision talks, panel discussions, and a fireside chat. The online audience could view the in-person conference broadcast, while the virtual event platform attendees also attended workshops, meet-the-speaker sessions, a career expo with our sponsors, and one-on-one networking.
Leading data scientists from industry, academia, government, and non-profit organizations discussed a wide range of topics, from human centered AI to healthcare to human rights to cybersecurity, and beyond. In several of the talks and panels, one common thread emerged: using data science to create more equitable solutions, across domains.
During the Data Science in Healthcare: Opportunities & Challenges panel discussion, Dr. Jinoos Yazdany, Chief of Rheumatology at Zuckerberg San Francisco General Hospital, explained that when she was doing her medical training working in several hospitals, she noticed that the care that patients received was uneven. Yazdany explained, “…this was the sort of thing that I wanted to work on and solve and my career has really been about putting together the pieces in my field to make care more equitable and more standardized…”. She described a project that changed the way that electronic health records were used, standardizing the way that patient outcomes were recorded, creating a registry with 2.5 million patient records that could be studied to assess the quality of care received. That has allowed for healthcare disparities to be identified, allowing an opportunity to provide feedback on how to close the gaps in care.
In her technical vision talk, Beyond Bias: Algorithmic Unfairness, Infrastructure, and Genealogies of Data, Dr. Alex Hanna, Research Director at The DAIR Institute, addressed her research in the genealogy of data, studying how data is collected and used to create datasets for research and applications. Research done by Hanna and her colleagues found that only 12 benchmark datasets constitute 50% of the benchmark data. That signals an issue since some of these benchmark datasets have issues with biased and even offensive classification, sometimes due to crowd sourcing. To solve for more algorithmic fairness, researchers need to have a thoughtful and methodical approach to dataset creation.
Dr. Tierra Bills, Assistant Professor at UCLA, delivered a technical vision talk, Confronting Data Bias in Travel Demand Modeling. During her talk, Bills posits that, “…inequality jeopardizes our national economy by making society weaker and poorly equipped in times of emergency…”, citing the U.S. response to COVID-19 and subsequent recovery as evidence. As a transportation engineer, Bills studies transportation inequities, noting that there are historical inequities like minorities displaced to produce the US highway systems, but also current inequities, such as the availability and quality of service by Uber/Lyft for black travelers, largely due to ride cancellations. She goes on to explain some techniques for how to confront and mitigate the bias in transportation in travel demand modeling.
Bills’ talk was followed by a panel that she participated in, the Algorithms and Data for Equity panel where Dr. Ling Jin, Research Scientist at Berkeley Lab, discussed some of her work to mitigate the environmental injustice of air pollution. Jin explained that there are twice as many deaths from particulate matter pollution than from car accidents, a health outcome that disproportionately affects lower income and racial minority groups. Some of her recent work combines air quality monitoring tools with atmospheric sensitivity analysis to come up with pollution control strategies that can benefit the disadvantaged communities in California’s Central Valley.
Dr. Jessica Granderson, Director for Building Technology, White House Council on Environmental Quality, talked about how she and her colleagues are looking through a different equity lens, anticipating who will participate in the new clean air economy. Granderson asks, “…who is it that is able to reap the benefits of rooftop solar, of the behind the meter battery storage of the smart and connected equipment and appliances…”. The entire panel continued the discussion, referring to specific challenges and opportunities to create more equitable solutions in the environmental and energy sectors.
The topic of better equity through inclusion and representation was discussed by Dr. Nadia Fawaz, Senior Staff Applied Research Scientist at Pinterest in her Inclusive Search and Recommendations technical vision talk. Fawaz talked about how she and her team implemented skin tone ranges and hair pattern searches, allowing pinners to narrow down their beauty searches based on their own skin and hair characteristics. Another inclusive project was described in a short video by Andrea Gagliano, Head of Data Science and AI/DS at Getty Images. She and her team updated a search algorithm that allowed more images of people from underrepresented ethnicities to be seen and downloaded more easily.
The final technical vision talk, Data Science for Equitable Recovery and Resilience, was delivered by the first female Chief Data Scientist of the United States, Denice Ross. She wanted the audience to know, “…how essential data science is for an equitable recovery from the pandemic and economic crisis as well as resilience for future shocks and stressors like climate change”. Ross described a newly formed equitable data working group that is tasked with identifying the inequities in the US government’s data collection infrastructure, laying out a strategy to improve data practices in the federal government. She encouraged participants to “scrub in” using the website build.gov guidebook and associated spreadsheets, to help underserved communities to apply, for example, for infrastructure funds to ensure that bridges and storm water management helps to mitigate the effects of climate change. Those starting their careers were also encouraged to apply for good government jobs at usajobs.gov as a way to serve their communities and their country.
Participants also learned about topics such as chest radiology AI, data science that enables drones to generate 3D models, a mathematician’s view of machine learning, and more. These and all the WiDS 2022 talks are now available on our YouTube channel. In the coming months, and throughout the year, we invite you to attend upcoming WiDS regional conferences, participate in new WiDS Workshops, and listen to new WiDS podcast episodes.