Lucia Huo is a Master’s student in the Technology Policy Program (TPP). She works in the Healthy ML Lab and is interested in understanding how limited health data availability affects modeling outcomes, particularly for marginalized and minority groups. Prior to TPP, Lucia worked on mitigation for emergent AI harms on Microsoft’s Responsible AI team.
What is the focus of your research? What sort of knowledge and disciplines does it bring together? How will it make an impact?
My research aims to understand how we can extract meaningful, causal insights from population health data using traditional epidemiological modeling techniques. Specifically, my research studies how natural population processes (e.g. population distribution patterns across neighborhoods) lead to confounding in causal inference with observational health data; as well as how such confounding might invalidate or undermine conclusions drawn from such health modeling. My work also disaggregates outcomes according to demographic subgroups to assess the fairness of computational approaches. This research integrates causal inference and epidemiological modeling while drawing on perspectives from ethics and the social sciences. By examiningwhere and for whom population health knowledge generation is weakly supported by real data, this work contributes to improving the validity of health evidence used in policymaking and clinical practice.
Last summer you interned with Calla Lilly Clinical Care. Who did you work with and what did you do?
Last summer, I interned with Calla Lily Clinical Care, a women’s health startup in London, UK, focused on improving drug delivery for women. I worked closely with Vinh-Thang Vo-Ta ’98, the startup founder and MIT alum, to develop an AI and machine learning implementation plan to support vaginal drug delivery research and development. I also supported the team by reviewing recent population health literature on under-investigated women’s health issues, such as how bacterial vaginosis symptomatology varies by ethnicity. In parallel, I advised the team on the responsible and ethical use of AI tools in a startup context.
How does the internship connect to your current research and future plans?
This experience connects to my future plans by deepening my understanding of the challenges inherent in women’s health research—even in settings that actively invest in minority health innovation and have highly skilled scientific teams. It highlighted the limitations of applying machine learning in data-sparse domains where data is systematically missing, as well as the difficulty of developing models that can meaningfully guide real-world clinical decision-making in highly specific and sensitive areas like women’s health.
