Metacognitive Strategy Networks in GenAI-Enhanced Learning: An Epistemic Network and Lag Sequence Analysis of High and Low Self-Efficacy Students

Many studies have demonstrated that generative AI (GenAI) feedback has potential in language learning; however, few studies explore how learners' self-efficacy shapes their metacognitive and behavioural engagement with GenAI feedback. This study investigates differences in metacognitive strategy networks and behavioural sequences between high and low self-efficacy learners interacting with GenAI feedback in reading comprehension tasks. By integrating epistemic network analysis (ENA), sequential analysis, and retrospective interviews, this study analysed the interactions of 16 junior secondary students with five types of GenAI feedback. The results revealed that high self-efficacy learners develop metacognitive networks centred on evaluating and dynamically cycling between feedback types to inform future learning. In contrast, low self-efficacy learners displayed a metacognitive strategy network dominated by planning and linear behavioural sequences that prioritise immediate task completion. The findings advocate for self-efficacy-driven GenAI feedback systems that dynamically tailor metacognitive scaffolding to catalyse personalised self-regulated learning.