Decisions are not all equal - Introducing a utility metric based on case-wise raters' perceptions