Rachel
Freedman

Photo of Rachel Freedman
PhD candidate in AI
UC Berkeley · CHAI

I'm a PhD candidate in AI at UC Berkeley, advised by Stuart Russell at the Center for Human-Compatible AI. I will graduate in 2027 and am currently on the job market — here is my CV.

Research

I work on robustly aligning AI systems with human values given noisy and pluralistic human feedback. Active Teacher Selection (TMLR, 2026) learns to identify the most informative feedback sources and efficiently spend limited query budget to learn robust reward models. Adaptive Pluralistic Alignment (ongoing) uses social choice theory to aggregate judgements from personalized reward models, and efficiently adapts the aggregation process over time to prevent value lock-in.

Previously, I adapted a kidney exchange algorithm to prioritize kidney recipients based on human moral judgements (Artificial Intelligence Journal, 2020), applied social choice theory to learning from human diversity (ICML 2024), and contributed to a taxonomy of limitations of modern RLHF methods (ICLR 2025).

Service & support

Outside my own research, I mentor early-career researchers through the CHAI internship and the PRISM fellowship, serve as a grant advisor for Schmidt Sciences, and serve on program committees for the AAAI alignment track and AI safety workshops. My PhD has been supported by CAIF, Foresight, LTFF, and Berkeley EECS fellowships.