I don't pretend it'll be "perfect", but it should let decisions fall (noisily) around what we think of as "fair" (since we're all running that assessment through the same hardware) which is really what we're looking for.