Natural Language as the Bridge Between Pedagogy and AI Paper 1 of 4: Encoding Pedagogy – Elements and Parameters for Constrained LLM Generation

Pedagogical expertise is often expressed in rule-like, natural language terms. This has direct application to LLM content generation: the same natural language rules that describe a pedagogy can serve as structured constraints that guide LLM output. This paper presents a lightweight framework for encoding this expertise as elements and pedagogies. The framework decomposes educational content into discrete elements, each guided by parameters that specify both per-element constraints and inter-element relationships. These two constructs (elements and parameters) provide a grammar for expressing a wide range of approaches as a set of natural language rules that LLMs can interpret and follow. The framework has been developed through direct professional experience delivering AI professional development and sustained work groups with practising teachers and content creators across primary, secondary, further, and adult education, and reflects the practical constraints and opportunities that educators encounter when working with these tools. We believe our approach to be agnostic, subject-independent, and accessible to any educator who can articulate what they want their content to do. The framework is illustrated through worked examples that teachers can use in any LLM to begin applying the framework. This is the first of four papers; subsequent papers extend this foundational architecture to audience-specific tailoring (Paper 2), visual and spatial design (Paper 3), and temporal sequencing for video production (Paper 4). Taken together, the series describes how this framework can support structured content generation at considerable scale, with the ability to tailor to individual student needs, while maintaining the pedagogical consistency that comes from encoding expertise as explicit, reusable rules.Keywords: content generation, constrained LLM generation, prompt engineering, educational technology, natural language processing, element-based content design.