Amid an Epistemic Iteration: How AI Methodologies May Transform the Nature of Science Education Research

Abstract Advancements in artificial intelligence (AI) may transform the nature of science (NOS). These transformations may extend to science education research, where AI methodologies increasingly enable scalable, reproducible, and fine‑grained analyses of student learning. To reflect on potential transformations that AI methodologies may bring to science education research, we draw on Hasok Chang’s notion of epistemic iteration , in which successive stages of knowledge build iteratively on one another to achieve specific epistemic aims. Anchored in this notion, and further illustrated by the historical development of thermometers, this article proposes and applies a framework for reflecting on how AI methodologies may transform the nature of science education research. The framework comprises seven phases: (1) problem framing, (2) instrumentation and measurement, (3) experimentation and evidence‑based inference, (4) comparisons and replication, (5) building norms and consensus, (6) implementation and its consequences, and (7) continuous refinement. For each phase, we analyze how AI methodologies may transform the nature of science education research and draw parallels to historical explorations of temperature. Our analysis suggests that the field may currently be amid an epistemic iteration, in which AI methodologies may advance scientific inquiry within the field. Simultaneously, AI methodologies may contribute to prioritizing different epistemic aims, epistemic criteria, and epistemic practices compared with traditional methods. By recognizing these potential transformations and drawing on lessons from the history and philosophy of science, we evaluate how AI methodologies may impact how science education research is done to ensure that promising avenues for future research are epistemically well grounded.