AI-Based Learning Analytics for Lecture-Free Instruction

Abstract

Recent educational approaches emphasize the effectiveness of using lecture-free environments that use active learning to boost student outcomes. This approach has been used at Khalifa University since 2018, using its Moodle-based Learn-Smartly platform. This paper explores how AI-based learning analytics can enhance this approach by analyzing educational data to understand the factors influencing student performance and engagement. Artificial Intelligence models, such as linear regression, decision trees, random forests, and XGBoost, alongside methods like frequency analysis, sequence analysis, and statistical tests, were used to identify impactful factors. Our findings demonstrate that certain question types and sequences influence student engagement and performance. For example, question sequences with repetitive semantic question types like conceptual questions and code analysis questions lead to lower performance and engagement. Statistical analysis shows that gender does not significantly affect learning outcomes in this setting and that this approach is gender-neutral. Future research will explore the applicability of these findings across different fields and educational contexts. We also plan to explore personalized learning paths using AI and large language models such as ChatGPT to tailor educational experiences to individual needs.

Date of Award	20 Jul 2024
Original language	American English
Supervisor	Abdulhadi Shoufan (Supervisor)