Henry Arthur
M.Sc. Student, Cognitive Science — University of Trento (CIMeC) & Ev's Language Lab, MIT
I am broadly interested in language and reasoning in human and artificial intelligence. How can we measure model-human alignment, and what sort of representations are needed to capture meaning computationally? I am also curious how these questions inform the way we regulate and treat AI in our daily lives.
Currently I work with Andrea de Varda, Hope Kean, and Ev Fedorenko on model-human reasoning alignment and arithmetic.
I spend as much time as I can climbing rock/ice, and making furniture.
Background
Previously, I spent time in NASA's Human Research Program under Mona Matar, building AI tools to prepare for future Mars missions; before that, in NASA's geophysics lab; and before that, at Fermilab on the Mu2e experiment with Kamal Benslama. As an undergrad, I studied Computer Science and Philosophy.
Publications
- (2026). Thinking effort aligns between humans and reasoning models in abductive reasoning. To appear in Findings of EMNLP 2026. arXiv:2609.01867.