Auditing Health Data: Building Better AI for Human Systems
Machine learning systems in health care often fall short of their promise because we lack precise ways to characterize how humans and algorithms behave in real-world settings. Without such understanding, it is difficult to design systems that reliably complement human judgment. In this presentation, Dr. Divya Shanmugam will discuss methods for studying human and algorithmic decision-making in health care, with a focus on how human behavior shapes observed data and how algorithmic performance can be evaluated under real-world constraints. Together, these directions illustrate how better characterization of data, decisions and algorithms can support machine learning systems that more effectively improve care.
About the Speaker:
Dr. Divya Shanmugam is a National Library of Medicine Fellow at Columbia University, where she works to make machine learning systems more equitable and reliable, particularly in health care. Her work has appeared at top machine learning venues including NeurIPS, CVPR and CHI and been featured in The New York Times. Her research was supported by a National Science Foundation Graduate Research Fellowship and she has been recognized as a Rising Star in EECS. She earned her PhD and BS in Computer Science from MIT.
About the Series:
The Big IDEAs About Health Data Speaker Series brings together leading voices from across research, policy, health care and the public sector to explore how administrative data can be used to advance health equity in Canada. Through thought-provoking presentations, speakers examine the responsible use of disaggregated data—including sex and gender, race and ethnicity, disability, income, housing, language and other social determinants of health. They also spotlight emerging research methods and data research practices that embed inclusion, diversity, equity, accessibility and community perspectives into algorithms, distributed analytics, community involvement and equity assessment tools.
Missed a session? Watch recordings of past webinars. External Link. Opens in new window. and explore the ideas shaping the future of equitable health data research.