From Dialogue to Depth: A Conceptual Model of Generative Artificial Intelligence Mediated Learning in Physics Education

The rapid uptake of generative artificial intelligence (GenAI) in science education calls for theoretical frameworks that explain how AI-mediated interaction shapes learning processes, rather than merely whether particular tools are effective. This conceptual paper proposes the Dialogue-to-Depth model as a theory-building framework for understanding how GenAI can support complementary epistemic functions in physics education. The model conceptualises learning as a pedagogical movement from dialogic exploration, in which learners externalise intuitive ideas, negotiate meaning, and engage in explanatory questioning, toward analytic consolidation, in which emerging conceptions are stabilised through formal representations, canonical explanations, and disciplinary language. Drawing on constructivist and sociocognitive perspectives, engagement-oriented theories, productive failure, and contemporary work on AI literacy, the model positions generative AI as an epistemic mediator whose pedagogical role is enacted through instructional framing and teacher orchestration rather than determined by platform-specific affordances. The paper articulates theoretical implications for conceptual change research and AI-in-education, derives design-oriented principles for AI-supported physics pedagogy, and outlines a research agenda for empirical validation of the proposed trajectory across topics and learner populations. By offering a coherent conceptual vocabulary linking dialogue to depth, the model aims to support principled integration of GenAI into physics classrooms in ways that preserve learner agency, foster critical engagement with AI-generated explanations, and promote deep conceptual understanding.