Saeyoung Rho is a computer scientist developing scalable and privacy-preserving methods for personalized causal inference to support the design of policies, algorithms, and AI services. She is a final-year PhD candidate at Columbia University, co-advised by Professors Vishal Misra and Rachel Cummings. During her PhD, she gained industry experience at Amazon AWS and Google X. Prior to her PhD, she studied at Massachusetts Institute of Technology (MIT) and Korea Advanced Institute of Science and Technology (KAIST), earning a Master’s in Electrical Engineering and Computer Science and a Master’s in Technology and Policy from MIT, as well as both a Master’s and a Bachelor’s in Industrial Design from KAIST.
Research Summary
TL;DR: I strive to view the world through a causal lens and develop the right tools to answer “why” in various situations, while also exploring the ethical issues that arise from it.
My central research question is: How can we derive individualized causal insights from data to support the tailored design of policies, algorithms, and AI services? We are surrounded by abundant individual-level time-series data—from Apple Watch sensor logs to global weather records—but we lack a reliable tool to gain personalized causal insights at this scale. Estimating “individual” treatment effect is known to be a hard problem, but I aim to tackle it borrowing the ideas from machine learning and the power from more data.
My PhD dissertation focuses on advancing synthetic control methods to enable personalized causal inference. Synthetic control is a powerful observational method using time-series panel data to estimate counterfactual outcomes. Yet, its classical form faces challenges with large, noisy, and sensitive individual-level datasets. To address these issues, I initiated three following studies. My first paper introduced the first differentially private synthetic control algorithm to ensure privacy while preserving utility. In my second paper, I introduced a data pre-processing step for synthetic control that mitigates high-dimensional overfitting for large-scale individual data. My third paper proposed using a state-space model with a constant trend to improve predictive accuracy in extended time horizon.
Building on these foundations, my future research will focus on developing theory-driven tools for personalized causal inference at scale, emphasizing ethical and methodological issues in working with human data. Drawing on my background in technology policy and industrial design, I plan to broaden this agenda toward human-centered AI design, integrating causal reasoning into the design and evaluation of systems interacting with humans and society.
My long-term goal is to complete the iterative design cycle by integrating qualitative and quantitative methods to design, evaluate, and update complex, evolving systems.
Saeyoung Rho 