From curriculum to strategy: tracing shifts in AI Education policy in South Korea through keyword and topic analysis

This study examines the evolving discourse of artificial intelligence (AI) education in South Korea, tracing shifts from pedagogical foundations to strategic national positioning. Using computational text analysis – combining keyword frequency analysis, N-gram modelling and Latent Dirichlet Allocation (LDA) – national curriculum documents (2015, 2022) and policy texts from preparatory (2019–2022) and internalization (2023–2024) phases were analysed. The analysis reveals three major transformations: (1) curriculum discourse exhibits surface adoption of competency-based vocabulary while maintaining deep structural continuity in assessment practices; (2) policy discourse undergoes a rapid transition from educational to strategic-economic framing, with enterprise and governance terms displacing traditional pedagogical vocabulary; and (3) asymmetric convergence emerges between curriculum and policy domains, where shared terminology masks divergent institutional logics. Theoretically, the study contributes the concept of “compressed policy diffusion” where multiple adoption mechanisms operate simultaneously in developmental contexts. Methodologically, it demonstrates how computational analysis of Korean policy texts reveals discursive patterns invisible to traditional approaches. The findings challenge assumptions about paradigmatic educational transformation, revealing instead how global innovation rhetoric overlays persistent local structures. The Korean case illustrates potential risks when AI education reforms prioritize strategic-economic alignment over pedagogical coherence, suggesting the importance of balancing innovation with educational purposes in similar developmental contexts.