This paper introduces Dialogic Artificial Intelligence (DAI)—a conceptual and computational framework for modeling how AI systems engage with human epistemic preferences in processes of knowledge retrieval and semantic representation. Grounded in Wittgenstein’s theory of language games, DAI facilitates context-sensitive interpretation of user intent through dialogical interaction. Using multiverse hypergraphs and a mechanism called Dispute Learning (DL), the framework maps diverse queries onto epistemic subgraphs that reflect overlapping but distinct knowledge communities. Unlike existing preference-matching models, DAI emphasizes pluralism, transparency, and contestability in meaning construction. Real-world applications include academic search, semantic interoperability, and policy analysis—pointing toward a more dialogical paradigm in knowledge-driven AI systems.