Procedure Learning

Procedure learning focuses on understanding how agents acquire and execute sequences of actions to achieve goals, a crucial aspect of both human cognition and artificial intelligence. Current research emphasizes developing robust computational models, including graph-based and latent prediction architectures, to learn procedures from diverse data sources like videos (both egocentric and third-person) and sensorimotor interactions, often leveraging unsupervised or self-supervised learning techniques. These advancements are improving the ability of robots and AI systems to perform complex tasks and offer new insights into human skill acquisition, with applications ranging from medical training to assistive robotics. Furthermore, research is exploring how different theoretical frameworks of learning can be reconciled through computational modeling to better understand the underlying cognitive mechanisms.