Addressing people’s current and future states in a reinforcement learning algorithm for persuading to quit smoking and to be physically active