This challenges the dominant paradigm of pluralistic alignment as a simple problem of data aggregation. It reframes it as a dynamic, interactional failure, suggesting current methods are building systems that are fundamentally broken at the conversational level, not just under-representative in their training data.