Programming Language Policy as an AI Literacy Equity Problem: A 15-Nation Comparative Analysis

Synthesis: Across 15 nations, the paper examines how secondary computer-science education embeds AI literacy into general-track subjects (Digital Literacy, ICT, TIC, SNT) rather than specialized tracks, creating structural inequities in who develops AI capability. The comparative analysis shows that policy choices about which programming language and subject bears 'universal' AI literacy determine differential access to computational futures.

Frames AI literacy 'for all' as an Equity problem rooted in K-12 curriculum policy, with direct relevance to Educational AI Policy and Teaching decisions. It ties to AI Literacy as a civil competency and to Educational Development for teacher preparedness, arguing that without equitable language/policy grounding, AI literacy widens rather than closes gaps in Higher Education readiness.

What this means for practice

• Administrators. Audit which track carries the AI literacy mandate in your system and who is excluded by that placement: where AI literacy is lodged in general-track subjects (Digital Literacy, ICT, TIC, SNT) rather than specialist tracks, it is the language and depth attached to that subject that decides who develops computational capability.
• Administrators. Choose a reform mechanism you can actually operate. France's structural break required strong central decree capacity, Poland's shift to Python arrived through exam task redesign, and Switzerland coordinated a decentralized system only with a strong institutional anchor and extended timelines.
• Instructors. Expect the specialist language to arrive in your general classroom through you: in resource-constrained systems the general-track curriculum cannot be delivered independently of the specialist track, and the implementation gap widens where specialist CS teacher pipelines are thin.
• Instructors. Check your programming choice against the competency tier you claim to teach — the evaluate-and-create tier presupposes reading, tracing, and modifying a system's computational logic, not operating its interface.
• Administrators. Budget for enactment, not just adoption: statutory frameworks and examination specifications state intended curricula, and the paper's own warning is that classroom reality is mediated by teacher preparation, resources, and institutional inertia.

Limitations

• The method is documentary: the analysis covers intended curricula and statutory frameworks, not enacted classroom practice, and the authors state that the implementation fidelity gap between policy and delivery is likely significant where specialist CS teacher pipelines are thin.
• The 15 systems were selected by purposive sampling biased toward systems with strong documentation and reform activity; lower-income nations across Africa, South and Southeast Asia, and small island states are absent as an artifact of method — the authors are explicit that this is not evidence the challenges are absent there.
• Scope is upper secondary (ISCED 3): earlier exposure embedded in lower-secondary mathematics or science (for example Norway's LK20) and informal pathways fall outside the study.
• The classifications are a snapshot of a fast-moving policy field and may already be dated — the paper flags Kazakhstan's 2025/2026 and Romania's planned 2030 reforms — and federal or devolved systems are classified by dominant or nationally representative pattern where no single authoritative standard exists.

Citation

Adrian-Marius Dumitran, Iulia-Maria Popescu (2026). Programming Language Policy as an AI Literacy Equity Problem: A 15-Nation Comparative Analysis. arXiv preprint.