Gabriel Poesia

The ability to reason, to construct logical chains from hypotheses to conclusions, is a central cognitive skill developed throughout formal education. Yet, reasoning still largely eludes modern artificial intelligence systems. My research centers mathematical reasoning from three angles: What does it take to build flexible systems that reliably solve a range of mathematical problems? How do people learn to reason mathematically? How can we leverage computational agents for mathematics to assist students in their education? My ongoing PhD work towards answering these questions bridges artificial intelligence, cognitive science, and education. My grounding goal is to build tools capable of understanding and giving verifiable feedback on students’ mathematical work. Unlike human teachers, popular educational platforms can still only check final answers, severely limiting what students can learn from the system. Providing feedback on students’ reasoning is a challenge requiring innovations in AI, with a potentially broad impact in and beyond academia.