This is an interesting idea. Maybe we should be looking at this as an optimization problem where drivers have a limited budget for attention and we want to concentrate it in the parts where it's most useful by varying the quality of UI assists?