Natural Language As The Bridge Between Pedagogy & AI

Paper 1 presents a framework for encoding pedagogical expertise as structured constraints for LLM content generation. Educational content is decomposed into labelled elements (such as [Definition], [Example], [Surprising Fact]), each governed by natural language parameters that specify what the element should do and how it relates to others. The paper argues that teachers already express their expertise in rule-like terms that map directly onto the constraint structures known to improve LLM output, making the framework accessible without technical knowledge. Worked examples, copy-paste prompts, and batching strategies demonstrate how a single pedagogy can generate entire content libraries and support systematic comparison of pedagogical approaches. Paper 2 extends this to audience tailoring through variant profiles, allowing one pedagogy to produce content adapted for different languages, reading levels, and neurodiversity needs. Paper 3 addresses visual and spatial design, applying deterministic parametric controls to how content is rendered as slides, worksheets, or screen layouts. Paper 4 covers temporal sequencing for narrated video production, compiling elements into timed, narrated walkthroughs. Together, the four papers describe a pipeline from pedagogical encoding through audience adaptation, visual rendering, and video compilation.