Papers
7
Total Citations
253
H-Index
6
About
Andrew Hundt is a robotics and AI safety researcher whose work spans human-robot interaction, end-user robot programming, and the ethical implications of machine learning systems. He is perhaps best known for developing CoSTAR (Collaborative System for Task Automation and Recognition), a pioneering platform enabling non-expert users to instruct collaborative robots through behavior trees and vision-based reasoning — a contribution that has garnered over 170 citations and addressed a fundamental barrier to real-world robot deployment. His associated user studies and the CoSTAR Block Stacking Dataset further deepened understanding of task planning and workspace-constrained learning in robotic manipulation. In recent years, Hundt has emerged as a vital critical voice at the intersection of robotics and AI ethics. His influential 2022 paper "Robots Enact Malignant Stereotypes" (50 citations) exposed how machine learning biases embedded in systems like CLIP can translate into discriminatory robot behavior — a sobering finding with significant implications for deployment safety. He has continued this thread with work examining how large language model-driven robots risk enacting discrimination, violence, and unlawful actions, making him one of the field's most important advocates for responsible AI in physical systems. His career reflects a rare combination of engineering innovation and ethical rigor.
Research Focus
Key Achievements
Top Papers
- 1CoSTAR: Instructing collaborative robots with behavior trees and vision170 citations · 2017
- 2Robots Enact Malignant Stereotypes50 citations · 2022
- 3
- 4
- 5
- 6The CoSTAR Block Stacking Dataset: Learning with Workspace Constraints6 citations · 2019
- 7“Good Robot!”: Efficient reinforcement learning for multi-step visual tasks via reward shaping5 citations · 2019