Thinking fast or slow? A reinforcement-learning approach