Where do automated metrics come from, and how meaningful are they? I’m interested in understanding the underlying measurements rather than taking a score or algorithmic label at face value.