AI enhancing writing skills: Moving from feedback to reflective learning for primary school students

Abstract

Writing skills are widely acknowledged as fundamental to
academic achievement and lifelong learning, yet primary
school pupils often struggle with higher-order aspects of
composition such as planning, revision, and reflection. The
research employed a quasi-experimental design informed by
constructivist and instructional design theories. A total
of 259 fifth-grade pupils from a public school in Shenzhen
were divided into experimental and control groups. The
experimental group engaged with an AI platform that
combined automated writing evaluation, portfolio-based
progress tracking, and iterative task redesign, thereby
operationalising principles of double-loop learning. The
control group received conventional teacher-led feedback.
Writing literacy was assessed across five dimensions:
context, content, process, strategies, and feedback. In
addition, rubric-based scores were analysed using Epistemic
Network Analysis (ENA) to explore structural connections
among writing dimensions. Findings revealed that the
experimental group achieved significantly higher post-test
scores in all dimensions, with notable improvements in
content development, contextual awareness, and engagement
with feedback. ENA results confirmed that AI-supported
feedback not only enhanced overall writing performance but
also strengthened higher-order linkages between structure,
content, and norms. Although effect sizes were modest, the
convergence of results across ANCOVA, ENA, and survey
measures demonstrates the reliability and validity of the
findings. This study contributes theoretically by extending
current understandings of AI in education beyond corrective
assistance, conceptualising AI feedback as a catalyst for
reflective practice and structural integration in writing.
Practically, the research highlights the pedagogical value
of multi-round, personalised evaluation and portfolio-based
scaffolding, offering implications for writing pedagogy,
curriculum design, and equitable resource sharing in
primary education. The study suggests that AI can
complement rather than replace teacher expertise, providing
a sustainable pathway for fostering students’ reflective
engagement, metacognitive strategies, and higher-order
writing competence.