Scholarlink

<p>Generative Artificial Intelligence (GenAI) tools have transformed higher education, yet empirical comparisons with traditional digital resources remain limited. This study proposes GenAI-EduNet, a hybrid deep learning architecture integrating Long Short-Term Memory networks and multi-head self-attention transformers to predict and analyze learning outcomes when students use GenAI tools, specifically Google NotebookLM, versus conventional digital materials. The research adopts a rigorous quasi-experimental quantitative design involving 1847 undergraduate students across two academic semesters, with model training and validation using the Open University Learning Analytics Dataset (OULAD) and EdNet. GenAI-EduNet introduces several innovations: a multimodal engagement encoding mechanism with gated attention capturing behavioral, cognitive, and affective dimensions; an adaptive knowledge tracing module based on transformer attention for modeling temporal learning patterns; a dual-branch comparative performance predictor employing inverse propensity weighting for causal inference; and a binary outcome module for pass–fail prediction. Experimental results demonstrate that the proposed framework achieves 94.7% classification accuracy and an AUC of 0.967, outperforming ten state-of-the-art baselines by 8.3%. Quasi-experimental analyses reveal that students using NotebookLM achieved significantly higher learning gains, including post-test scores (d = 0.73, <i>p</i> &lt; .001), higher-order thinking skills (d = 0.74, <i>p</i> &lt; .001), cognitive engagement (d = 0.59, <i>p</i> &lt; .001), and self-efficacy (d = 0.52, <i>p</i> &lt; .001). Overall, the findings highlight the pedagogical value of integrating source-grounded generative AI tools, specifically NotebookLM in higher education and contribute a validated computational framework to learning analytics research in the era of artificial intelligence, supporting evidence-based policy development, curriculum redesign, faculty training, and sustainable digital transformation worldwide.</p>

错误：搜索内容不能为空，请输入英文关键词
错误：关键词超出字数限制，请精简
高级检索

Pedagogical transformation through generative AI: a hybrid deep learning comparison with traditional digital learning materials based on GenAI-EduNet framework

  • 期刊论文

  • 发表日期：

    2026年3月24日

查看全文
  • Faisal Alshammari

摘要

Generative Artificial Intelligence (GenAI) tools have transformed higher education, yet empirical comparisons with traditional digital resources remain limited. This study proposes GenAI-EduNet, a hybrid deep learning architecture integrating Long Short-Term Memory networks and multi-head self-attention transformers to predict and analyze learning outcomes when students use GenAI tools, specifically Google NotebookLM, versus conventional digital materials. The research adopts a rigorous quasi-experimental quantitative design involving 1847 undergraduate students across two academic semesters, with model training and validation using the Open University Learning Analytics Dataset (OULAD) and EdNet. GenAI-EduNet introduces several innovations: a multimodal engagement encoding mechanism with gated attention capturing behavioral, cognitive, and affective dimensions; an adaptive knowledge tracing module based on transformer attention for modeling temporal learning patterns; a dual-branch comparative performance predictor employing inverse propensity weighting for causal inference; and a binary outcome module for pass–fail prediction. Experimental results demonstrate that the proposed framework achieves 94.7% classification accuracy and an AUC of 0.967, outperforming ten state-of-the-art baselines by 8.3%. Quasi-experimental analyses reveal that students using NotebookLM achieved significantly higher learning gains, including post-test scores (d = 0.73, p < .001), higher-order thinking skills (d = 0.74, p < .001), cognitive engagement (d = 0.59, p < .001), and self-efficacy (d = 0.52, p < .001). Overall, the findings highlight the pedagogical value of integrating source-grounded generative AI tools, specifically NotebookLM in higher education and contribute a validated computational framework to learning analytics research in the era of artificial intelligence, supporting evidence-based policy development, curriculum redesign, faculty training, and sustainable digital transformation worldwide.