AI-Powered Generation of Teaching Materials to Support Autonomous Learning in Higher Education

Abstract

This study presents and evaluates an automated system for generating personalized lecture notes through artificial intelligence, designed to foster autonomous learning in higher education. A modular Python-based workflow integrating local large language models (LLMs) processed a bilingual corpus (Spanish– English) to extract, synthesize, and validate content. Five progressively broader queries—ranging from the definition and composition of olive pomace to its production, industrial uses, and environmental challenges—were tested under two conditions: closed mode (local sources) and open mode (including online references). The system produced coherent notes covering around 90% of key concepts, consistently citing references, with external sources enriching content without reducing coherence. Quantitative evaluation showed keyword coverage of 85–95%, strong agreement between automatic validation and experts (κ=0.85), excellent inter-rater reliability (ICC=0.92), and processing times of about 15 seconds. Results highlight the pedagogical potential, efficiency gains, and ethical challenges of integrating generative AI responsibly in higher education.