Andrés Palladino

Independent researcher working on AI corrigibility, effective human control, and corrective reachability in agentic systems. My research examines whether human intervention remains causally effective after an AI agent’s earlier actions have reduced available options. I am developing The Corrigibility Window through conceptual work, controlled experiments, real-world case analyses, and bounded agent evaluation. Two foundational preprints are publicly available.

Personal Website

https://andrespalladino.com/corrigibility

Projects

No grants recorded.