The explore/exploit dilemma in human reinforcement learning: Computation, behavior, and neural substrates 03-30-2006