Influencing Humans to Conform to Preference Models for RLHF