EWSC: Context in AI Research: Focus on Healthcare

EWSC-MIT EECS Joint Colloquium Series
Presented by Eric and Wendy Schmidt Center
September 9, 2025
Broad Institute of MIT and Harvard

Katherine Heller
Research Scientist, leading Context in AI Research (CAIR) team
Google

This colloquium is part of an ongoing series that is jointly hosted by the Eric and Wendy Schmidt Center at the Broad Institute and AI+D within the Department of Electrical Engineering and Computer Science at MIT.

The series features speakers who will share with us how their work drives novel insights into the most pressing biomedical questions of our time — and how biomedical questions are spurring foundational advances in machine learning.

Bio
Katherine Heller is a Research Scientist at Google leading the Context in AI Research (CAIR) team. She works on methods for identifying and mitigating robustness and fairness issues in medical and creativity contexts. She has previously worked on developing and integrating multiple machine learning systems into hospitals and clinical care including: a sepsis detection system which has been integrated into the Duke University Hospital Emergency Departments, a system for detecting the likelihood of complications resulting from surgery, and a nationally released mobile study on Multiple Sclerosis. She is interested in the inclusion of all people in the development of medical, and general, AI technology. Before joining Google, she was at Duke University in Statistical Science, Neurobiology, Neurology, Computer Science, and Electrical and Computer Engineering. She was the recipient of an NSF CAREER award and a first round BRAIN initiative award.

Talk Abstract
In this talk we will discuss reasons to value specificity in AI research, and related human-centered attributes, paying particular attention to AI research in the health space. Work on the construction of improved evaluations, out of distribution learning, and the benefits of causal methods for alleviating harms to underrepresented groups is highlighted. This methodology is used as the underpinnings of globally sensitive data collection, which is then used to improve understanding of these contexts in generative AI systems. We also explore the development of methods for the analysis of time series wearable data for prediction of mood disorders in maternal health.

This work is done in collaboration with the (Context in AI Research) CAIR team at Google.


For more information visit: https://www.ericandwendyschmidtcenter...
and https://www.broadinstitute.org/ewsc

Copyright Broad Institute, 2025. All rights reserved.