Longitudinal Effects of Human-in-the-Loop(HITL)-Based University Writing Education : Focusing on AI Utilization Patterns, Tutoring Achievement, and Learning Transfer

This study empirically examined the longitudinal effects of Human-in-the-Loop (HITL)-based university writing instruction across three semesters, drawing on 634 reflective journals, tutoring recordings, and achievement scores from students at University A. Using a mixed methods approach, the findings showed that learners' AI utilization progressively shifted from the planning stage toward content organization and revision across all semesters. Tutor feedback was categorized into corrective-directive and facilitative-questioning types, both of which promoted critical engagement with AI-generated content. Writing achievement improved significantly across midterm and final assignments, and learners reported transfer of acquired competencies to other academic and professional contexts. These results demonstrate that HITL-based writing education can reshape AI utilization strategies while fostering writing competency and learning transfer, offering practical implications for writing curriculum design in the era of generative AI.