UvA-DARE (Digital Academic Repository) Towards learning reward functions from user interactions