Framing reinforcement learning from human reward: Reward positivity, temporal discounting, episodicity, and performance