Attention Allocation in L2 Learners' Writing Feedback: Eye-Tracking Contrasts Between GenAI and Teacher Multidimensional Feedback

This study investigates the attention allocation in L2 learners' processing of generative AI (GenAI) and teacher feedback in academic writing, employing eye-tracking technology to address two research gaps: (1) the lack of a real-time comparative for analysing attention allocation across feedback sources, and (2) the underexplored attention allocation for multidimensionality of learners' feedback. A GenAI-enhanced interactive system was developed to deliver feedback across four dimensions (structure, clarity and coherence, completeness, grammar and vocabulary) from GenAI and teacher feedback to 10 Chinese undergraduates, with eye-tracking capturing attention allocation. Results revealed no significant difference in attention allocation between GenAI and teacher feedback in the same dimensions. This suggests real-time attention allocation parity with teacher feedback when GenAI feedback quality is optimised through pedagogical constraints. However, learners exhibited different attention allocation in different dimensions: completeness, grammar, and vocabulary demanded more attention than structure, clarity and coherence, reflecting learners' prioritisation based on immediate evaluative impacts. These findings highlight the critical role of feedback dimensions in cognitive load allocation. The study advocates strategic integration of AI-human feedback systems to address learners' writing revision priorities.