Environmental Education in AI-Mediated E-STEM Settings: Advances Toward a Framework for Pedagogical and Technological Integration

This repository contains the processed datasets, notebooks, scripts, coding rubrics, prompts, and supplementary materials used in the study “Environmental Education in AI-Mediated E-STEM Settings: Advances Toward a Framework for Pedagogical and Technological Integration.” The materials document the methodological workflow developed to transform scientific evidence into instructional design inputs for the generation of Digital Educational Resources within an E-STEM approach using Generative Artificial Intelligence and Machine Learning. The repository includes resources corresponding to the five phases of the study: construction and normalization of the scientific corpus; semantic extraction through a Multi-LLM strategy using ChatGPT, Gemini, and Qwen; assessment of inter-model agreement using Fleiss’ κ and Cohen’s κ coefficients; identification and interpretation of patterns through decision trees; and integration of these patterns with teacher-defined pedagogical parameters for the dynamic generation of a master prompt. The repository also includes derived analytical outputs such as model-specific evaluation datasets, agreement results, an integrated pattern library, figures, and variable documentation. These materials are shared to support methodological transparency, traceability, and reproducibility. The identified patterns should be interpreted as exploratory associations derived from the analyzed corpus and not as causal relationships or direct evidence of pedagogical effectiveness.