I’m broadly interested in ensuring human-AI systems work as intended and can be beneficial for the people involved: specifically, I’ve worked on improving the quality and robustness of agents trained to collaborate with humans, and am interested in the effects of recommender systems on users’ preferences and beliefs.
I’m very thankful to have grown up in Livorno, a lively city on the coast of Italy – it is to Livorno’s character that I owe much of my development and formative experiences.
Evaluating the Robustness of Collaborative Agents. Paul Knott, Micah Carroll, Sam Devlin, Kamil Ciosek, Katja Hofmann, Anca Dragan, Rohin Shah. AAMAS 2021.
On the Utility of Learning about Humans for Human-AI Coordination.Micah Carroll, Rohin Shah, Mark Ho, Tom Griffiths, Sanjit Seshia, Pieter Abbeel, Anca Dragan. Imitation, Intent, and Interaction Workshop, ICML 2019.
Overview of Current AI Alignment Approaches.Micah Carroll. Final Project for AI Safety and Control (CS294-149 at UC Berkeley), 2018.