Acting successfully in dynamic environments requires learning supported by two systems that differ in computational demand: a fast, model-free system that repeats rewarded actions, and a more ...