From Blocks to Code: An AI-Driven Curriculum for Elementary Python Programming

Abstract

This design-based research study explores the development of an AI-generated Python programming curriculum tailored to elementary school learners transitioning from block-based to text-based coding. The study responds to the increasing prioritization of computational literacy in early education, highlighting the global shift towards integrating programming as a foundational 21st-century skill. Drawing upon the TPACK (Technological Pedagogical and Content Knowledge) framework and cognitive load theory, the study utilized large language models (LLMs), including Veed, OpenAI’s GPT-4 and Claude, to independently generate modular lesson sequences, interactive exercises, multimedia resources, and assessments. The resulting curriculum comprises three progressive units covering Python basics, variables, and data types, with multimodal features designed to support cognitive development and sustain engagement. The study also examines the affordances and limitations of generative AI as an instructional design agent, highlighting its efficiency in content creation while underscoring the crucial role of human oversight in ensuring pedagogical coherence and developmental appropriateness. This work contributes to the emerging discourse on AI-assisted curriculum development, offering a model for scalable, self-guided programming education that bridges early block-based learning to general-purpose coding fluency.