Generative AI-Enhanced Science Pedagogy System for Digital Literacy Enrichment

Singh D.; Premlata; Ponce L.A.; Amalanathan S.M.; Das I.; Parida P.K. · Proceedings - 2025 International Conference on Recent Innovation in Science Engineering and Technology, ICRISET 2025 (2025) · DOI: 10.1109/icriset64803.2025.11252120

Resumen

Smart systems are required in enhancing scientific and digital literacy with regard to the fast-paced development of digital learning. The novel aspect of this study is given with a Generative AI-Enhanced Science Pedagogy System that incorporates complex ML and NL literacy methods and offers flexible and transparent learning. Methodology includes a combination of text normalization and bias elimination to achieve fairness and data quality and SHAP-based Top-K feature selection to have interpretable and efficient feature extraction. To provide high forecasting accuracy on learner progress predictions, it is using a LightGBM classifier and, to achieve personalized feedback and context-based content generation, Hugging Face Transformers are used. The system proves to be much more accurate, engaging and interpretable in comparison to traditional pedagogical models. Its potential to mitigate bias, stay scalable, and deepen digital literacy are supported by experimental evaluations of the positive effect of the interactive and accessible science learning. The framework, in general, provides a strong basis on the use of generative AI with regard to the next-generation digital pedagogy and the inclusive learning surroundings.

Instituciones: Universidad Nacional de San Antonio Abad del Cusco