Generative AI interactive textbook in electrotechnics: A four-year comparative study on student performance and inclusion

Texto completo

This study presents the results of the implementation of Generative-AI Interactive Textbook, an intelligent textbook built on a large language model based on GPT-4, which is directly integrated into the subject of Electrical Engineering course and provides interactive, adaptive teaching functions. The textbook was deployed in the 2024/2025 academic year, and the data was compared with the summary results from 2021 to 2024. The sample consisted of 736 students. The results suggest that the effects associated with the introduction of GenAI-T vary depending on the type of assessment and the analytical context. In multi-year analyses, a statistically significant improvement in mid-term assessment results was consistently observed, both across the overall student population and within individual subgroups. In contrast, final assessment results showed a more variable pattern, with a statistically significant difference identified in a one-year comparison between groups with and without access to GenAI-T, while no such association was observed in multi-year analyses. In this study we present a practical method for implementing the Intelligent Textbook, which includes analyzing course objectives, creating targeted prompts for configuring the AI assistant, and linking it to existing forms of education, thereby ensuring smooth integration into teaching and a reliable framework for further testing of the system. At the same time, it sets out principles to ensure that generative AI truly develops cognitive abilities, independence, and critical thinking, including the didactic anchoring of tasks to higher levels of Bloom's taxonomy, active verification and reflection on AI outputs, preservation of student authorship, and transparency in the use of technology. The results indicate that a systematically designed generative AI Interactive Textbook is associated with improved outcomes during the semester, with the most consistent increase observed in mid-term assessment, while effects on final assessment and between-group differences remain indicative rather than conclusive.