Computational Reinforcement Learning Using Rewards from Human Feedback