Andrea Procopio
BSc Computer Science & Economics @ Bocconi University
Research Fellow @ Harvard University
Research Fellow @ Harvard University
Abstract
My main interest is investigating how resource constraints shape the development of human-like representations in AI systems, particularly in visual intelligence and physical reasoning. I believe understanding these constraint-driven mechanisms could be essential for the informed design of human-like AGI systems.
Previously at the RIOT Lab (UNIVPM), I built tiny CNNs for wearables in the context of movement detection for Patients with Parkinson's Disease. Now at Harvard's CoCoDev Lab, I focus my efforts on investigating shared mechanisms of physical reasoning between computer vision models and humans.
Selected Publications
Under review at CVPR 2026 — Main track, Explainable Computer Vision
2025
Accepted at AAAI 2026 — XAI4Science Workshop
2025
Accepted at AAAI 2026 — AI for Time Series Analysis workshop
2025
Research Interests
Cognitive Science
Computer Vision
Intuitive Physics
Machine Learning
Neural Networks
Representation Learning
Human-AI alignment
Wearables