Using Options to Accelerate Learning of New Tasks According to Human Preferences