Katherine Shih

Papers

Total Citations

H-Index

About

Katherine Shih is a rising researcher at the intersection of human-robot interaction and preference-based learning, with a focus on making robotic systems more intuitive and accessible. Her work addresses a critical gap in robotics: how to optimize robot behavior when traditional metric-based feedback is unavailable. In her most-cited paper, "Optimizing Algorithms from Pairwise User Preferences" (2023, 6 citations), Shih pioneered methods for learning robot control policies directly from human preference comparisons, enabling developers to train robots without ground-truth scores—a breakthrough for human-centric applications where subjective user feedback is the only reliable signal. This work has already influenced how researchers approach preference-based reinforcement learning in assistive robotics. Her more recent study, "A Multi-Method Investigation of Guide Robot Characteristics for Blind and Low-Vision Users" (2025, 3 citations), demonstrates her commitment to inclusive design. Working with 16 blind participants, Shih identified key behavioral features that make guide robots trustworthy and effective for independent travel. Her research uniquely combines algorithmic innovation with rigorous user studies, positioning her as a leading voice in developing robots that truly serve diverse human needs.