Improving User Experience in Preference-Based Optimization of Reward Functions for Assistive Robots