Construction and Application Evaluation of a Generative AI‐Assisted Instructional Design Model for Vocational Education

ABSTRACT The rapid integration of Artificial Intelligence (AI) in vocational education (VET) has created opportunities to improve learning outcomes, engagement, and personalization. However, existing instructional models often lack adaptive feature selection and cognitive attention mechanisms, limiting their effectiveness. This study addresses this gap by proposing an AI‐assisted framework that integrates feature optimization and attention‐based learning. The primary objective of this study is to design and evaluate a GPT‐based model that enhances learning and engagement among learners compared to traditional instructional methods. A mixed‐methods research design was employed involving 250 students and 10 instructors. The proposed framework integrates adaptive fuzzy particle‐based feature selection with a feature‐attention mechanism and support vector regression for residual learning. Quantitative data were collected through pre‐test/post‐test assessments and Likert‐scale surveys, while qualitative insights were obtained via interviews. Statistical analyses included paired and independent t ‐tests, effect size (Cohen's d ), and multiple regression analysis. The Generative AI‐assisted group demonstrated significantly higher learning gains (36.7%) than the traditional group (20.5%), with strong statistical significance ( t = 9.23, p < 0.001) and a large effect size ( d = 1.17). Survey‐based evaluations showed consistently higher scores across learning outcomes, usability, engagement, and satisfaction dimensions, with most Cohen's d values exceeding 0.8. Regression models explained over 94% of the variance in post‐test performance. The results confirm that the proposed hybrid feature‐attention framework significantly enhances cognitive learning and engagement in VET. The framework offers a scalable and intelligent solution for personalized AI‐driven instructional design.