#115: Theoretical Construction and Case Study of Temporal Learning Analytics Based on Learning Engagement Dynamics Framework

Temporal learning analytics faces challenges of lacking systematic theoretical guidance and "black-box" feature extraction. This study proposes the learning engagement dynamics (LED) framework, decomposing digital learning behavior into three complementary, cognitively grounded dimensions: cognitive temporal regulation (CTR), task engagement dynamics (TED), and information exploration and integration (IEI), enabling a transparent mapping from clickstream data to cognitive constructs. In a theory-driven case study, we derived 16 temporal features and analyzed 837,144 interactions from 117 middle school students in AI-literacy modules using cross-classified nested random-effects models. Findings: (1) CTR (behavioral regularity) shows significant associations with cognitive engagement; (2) TED signals relate positively to accuracy on structured tasks, suggesting the value of sustained deep engagement; and (3) IEI exhibits task-specific patterns, systematic exploration associates with improved accuracy, while broader exploration relates to potential interference. LED provides a theoretical foundation and preliminary empirical insights for personalizing intelligent learning platforms.