Dissertation Talk: Reliable Causal Machine Learning for Public Sector Decisions

Event details Date: Thursday, July 30, 2026 Time: 11:00 AM to 11:00 AM Location: Gateway 4355 Type: Lecture / Workshop Audience: Faculty,Students About this event Machine learning systems are shaping public life and increasingly used to decide who receives public services. The data they rely on is often missing, noisy, or biased, and the outcomes they optimize for are frequently imperfect proxies for what we actually care about. In high-stakes public-sector settings, this erodes the reliability and equity of the decisions that follow. Drawing on causal inference, machine learning, and economics—and on collaborations with nonprofits and government agencies—we develop methods that confront imperfect outcomes across three settings: (1) policy learning from noisy, multidimensional outcomes in cash transfer programs; (2) cost-efficient outcome annotation for treatment-effect estimation in homeless services; (3) and model evaluation under biased proxies and selective labels in healthcare. Across all three, we find that treating measurement as a design problem yields systems that are more reliable and better matched to the social contexts they operate in. Official event details: https://events.berkeley.edu/eecs/event/324922-dissertation-talk-reliable-causal-machine-learning