Surfacing educational traditions in European K–12 computing curricula: A large-scale NLP-based comparison

Educational traditions underpinning computing education reflect broader pedagogical philosophies and national priorities. Building on the framework developed by Schulte et al. (2025) as part of an ITiCSE working group, which identifies four educational traditions – algorithmic problem solving, scientific, design and making, and societal – this paper presents a large-scale comparative analysis of K–12 computing curricula across 18 European countries. Using natural language processing (NLP) and large language models (LLMs), we extracted and classified over 15,000 learning objectives to examine two research questions: (RQ1) To what extent can an LLM reliably assign pedagogical traditions to learning objectives compared to human coders? (RQ2) How are these traditions emphasized across (a) countries and (b) educational stages? Results show that LLMs can effectively support curriculum coding, particularly when categories are clearly defined. However, human oversight remains essential, especially for ambiguous categories such as the scientific tradition. The study reveals a shift in emphasis from the societal tradition to the algorithmic problem-solving tradition as students progress through educational stages. Distinct national patterns emerge, reflecting differences in the rationale for teaching computing, its disciplinary framing, and curricular philosophy. This work explores the use of NLP techniques for large-scale curriculum analysis and offers insights into how educational traditions are emphasized across national computing curricula in Europe.