Path-dependent program induction under resource constraints explains human sequence learning

The article presents a hierarchical Adaptor Grammar (HAG) model that integrates rate-distortion theory with program induction to explain how humans learn sequences under cognitive resource constraints. HAG utilizes distinct local and global libraries to improve rate-distortion trade-offs and generalization, outperforming traditional methods in simulations and aligning with human recall behaviors in sequence-learning experiments. This work is significant for practitioners as it provides a framework for understanding structured learning processes, which can inform the development of more efficient AI models that mimic human cognitive strategies.