By Laurel Skurko. Algorithmic Justice, a concept that pertains to ensuring that data is processed in a way that is relevant to diverse populations, is increasingly important today as we rely heavily on AI, which is fueled by data, to treat people and populations. An important all-day workshop at UCSF reviewed data collection, processing, and decision making and therefore opened the door to increased awareness of the issue by professional and lay audiences, all of whom are involved in the resulting justice of algorithms.
UCSF has been a focal point for discussions on equity and AI in recent months. On November 7, 2023, Keith Yamamoto, PhD, special adviser to the chancellor for science policy and strategy and director of UCSF’s precision medicine program, invited several teams [1] to host a day-long hybrid workshop, “Toward Algorithmic Justice [2] in Precision Medicine” in Genentech Hall at UCSF. University leaders Chancellor Sam Hawgood and Inaugural UCSF Chief Research Informatics Officer (CRIO) Ida Sim, PhD, MD, joined other leaders in the field [3], including Alondra Nelson, PhD, UC San Diego alum, and former director of the White House Office of Science and Technology Policy, and Monica McLemore, PhD, UCSF alum, to engage in discussions with experts spanning technology, science, policy, medicine, and social equity.
The workshop united speakers and an audience of over 300 people to explore the issue of equity in AI and Precision Medicine. The panel discussions facilitated a valuable exchange of insights, underlining how to approach such a complex topic. It also underscored the transformation institutions like UCSF are undergoing as a result of revolutionary technological advancements today, as well as their critical role in shaping the future of medicine at this juncture.
The risks and why it matters
When talking about AI and genetic data, we are addressing people, their well-being and their lives. The answers we derive from AI are a matter of life and death. While AI provides actionable insights from data, there is massive room for improvement due to the issue of equity, where the need for more data and improved processes are needed to ensure that results are unbiased. Specifically, because their data is under-represented and the way their data is processed is not equitable, marginalized communities have not yet experienced the promise of AI in health care.
Additionally, there is an interdependent relationship between societal health and individual health. As Ida Sim explained, “Precision medicine is precise to our genes, and cells, but also to our humanity.” She continued, “Stress, racial trauma, and interactional trauma are etched in our own bodies.” By advancing science, we can impact society, hopefully repairing wounds that have been passed down across generations as well.
Setting the stage for AI and equity
While acknowledging the absence of simple solutions, the dialogue on November 7th at UCSF will yield recommendations to address equity challenges in data collection, utilization, and interpretation. The emphasis is on ensuring that innovation is strictly tied to positive health outcomes. Conversations like these will be pivotal in shaping successful approaches to managing new technologies like AI in both the short and long term.
The workshop’s rigorous expectations and three areas of focus
Hawgood, in his welcoming remarks, noted the timeliness of the discussion following the recent Executive Order on Safe, Secure, and Trustworthy Artificial Intelligence. Yamamoto set the stage for an interactive workshop, highlighting the expertise of every participant and emphasizing the importance of hearing all voices, while Sim encouraged the audience, saying, “It’s time to reflect, think and to act.”
The workshop organizers identified three key focus areas: (1) advancing transparency and explainability in healthcare algorithms; (2) engaging patients and communities throughout the healthcare algorithm lifecycle to build trustworthiness; and (3) ensuring accountability, equity, and justice in healthcare algorithm outcomes.
Six considerations for managing Risks – Some Insights
While the discussions remained open-ended, several possible guiding principles emerged. These considerations, derived from diverse perspectives gleaned throughout the day, might serve as a broad checklist for addressing the ethical dimensions of AI:
- External Pressure: Advocate for trustworthy AI by mobilizing external pressure on the government.
- Regulatory Frameworks: Establish FDA-like regulations for AI to ensure accountability, similar to devices and pharmaceuticals.
- Trustworthiness through Transparency: Develop AI fact sheets and scorecards to explain model derivation and reliability.
- Diversity in Data and Decision-Makers: Ensure diversity in both data sources and decision-makers, incorporating perspectives from technology, science, policy, medicine, and social equity.
- Community Engagement: Include the lay community up front in the process to ensure acceptance and relevance.
- Critical Thinking: Train current and future generations with the critical thinking skills needed to cross-check AI-derived results.
Gratitude for a day of learning and growth
The workshop organizers and sponsors provided expert facilitation, encouragement of diverse viewpoints, and commitment to positive change. The workshop structure itself, which included a well-designed curriculum, fantastic moderation, a diverse audience, highly informative quick polls, the inclusion of online audiences, and a glossary, all contributed to a productive program fostering collective learning.
What’s next for equity in AI
As Yamamoto remarked, ”What you have accomplished today is notable.” He said that the discussion will feed into recommendations that will emerge from numerous conversations such as these. It was an honor to be included as it will bear witness to the positive change on the horizon.
We look forward to information about similar conversations on equity in AI at other institutions and to promoting upcoming events on the topic. If you have additional information or views to share about this event or others like it, such as those events listed below, please contact the UC Tech News team at UCtech@ucop.edu.
Other recent events, addressing on equity in AI
Ethics in the Age of AI Health Monitoring
Thursday, September 28, 2023, 4 – 8 p.m.
Host: UCSF
Artificial Intelligence Across Biological Scales Symposium
Tuesday, December 5-6, 2023
Organizers: Drs. Nevan Krogan, Bruno Goud, Laura Cantini, Thomas Walter, Kliment Verba, Tanja Kortemme, the Quantitative Biosciences Institute (QBI) at the University of California, San Francisco, Institut Curie, PSL-QLife, and Institut Pasteur,
Location: UCSF
UC Health systemwide grand rounds on data science and AI
Tuesday, December 5, 2023, 12 – 1 p.m
Speakers: Atul Butte, M.D., Ph.D., Chief Data Scientist, UC Health, Cora Han, J.D., Chief Health Data Officer, UC Health, Sara Murray, M.D., M.A.S., Vice President and Chief Health AI Officer, UCSF Health, and Karandeep Singh, M.D., M.M.Sc., Associate Chief Medical Information Officer of AI, Michigan Medicine & Incoming Chief AI Officer, UC San Diego Health
Host: UC Health
Location: Zoom registration
Health AI Symposium – French American Innovation Days
Wednesday, December 6 – Thursday, December 7, 2023
Hosts: The Quantitative Biosciences Institute (QBI) at UCSF and the Office for Science and Technology of the Embassy of France in the United States
State of the University Address
Thursday, December 7, 2023, 12 – 1 p.m.
Speakers: Sam Hawgood UCSF chancellor, Atul Butte, MD, PhD, and Sara Murray, MD, MAS
Host: UCSF
Location: Livestream link
Sponsoring Organizations and Speakers
[1] Sponsoring Organizations
UCSF Research Development Office (RDO)
UCSF Office of the Chief Informatics Officer (CRIO)
UCSF Bakar Computational Health Sciences Institute (BCHSI)
UCSF UC Berkeley Joint Program in Computational Precision Health (CPH)
[2] Definition of algorithmic justice
Algorithmic justice: a concept and movement that focuses on ensuring fairness, equity and accountability in the development and deployment of algorithms and AR systems. It seeks to address and rectify issues related to bait, discrimination, and ethical concerns that Tina arise when algorithms are used to make decisions or automate processes in various aspects of society
[3] Speakers and their topics
- Keith Yamamoto (Introduction)
- Chancellor Sam Hawgood (Welcoming remarks)
- Alondra Nelson (Keynote speaker)
- Ida Sim (Introduction to workshop topics)
- Topic 1: Advance toward transparency and explainability in healthcare algorithms and their use
- Tony Capra (Panel discussion leader, which included those listed below and Alondro Nelson)
- Jianying Hu
- Tatyana Kanzaveli
- Peter Norvig
- Topic 2: Engage patients and communities in all phases of healthcare algorithm lifecycle and earn trustworthiness
- Tung Nguyen (Panel discussion leader)
- Jessica Newman
- Monica McLemore (Speaker)
- Mohana Ravidranath
- Angela Rizk-Jackson (Conclusion of the morning program and introduction to the afternoon session)
- Topic 3: Ensure accountability, equity and justice in outcomes from healthcare algorithms
- Cora Han (Panel discussion leader)
- Hoda Heidari
- Lisa Lehmann
- Margaret Handley
Gretchen Kiser (concluding workshop for online participants)
Related reading
Toward Algorithmic Justice in AI – Video recordings
November 8, 2023
Toward Algorithmic Justice – Final Agenda
November 7, 2023
Embed equity throughout innovation
Science (Authors include Keith Yamamoto, PhD)
September 7. 2023
Groundbreaking computational precision health program appoints 13 new faculty
Berkeley College of Computing, Data Science and Society
January 23, 2023
Year of Open Science
Science.gov
January 1, 2023
Conference explores justice-based model of health data use
UC Health
November 9, 2022
Got Health Data? Moving Toward a Justice-Based Model of Data Use
UC Health
April 19, 2022
NIH Guiding Principles for Ethical Research
NIH
Author
Marketing & Communications
UC Office of the President