-
Title
-
Glass-Box Feedback: Turning AI Chatbots into Metacognitive Writing Partners for Year 5 Girls
-
Author
-
Makiko Ryland
-
Year Published
-
2026
-
Description
-
Many students approach writing as a task to complete rather than a process of drafting, reflecting and refining ideas. This action research project investigated how AI-generated feedback influenced Year 5 girls’ metacognitive engagement and revision practices in writing within a primary school context in Sydney, Australia. With generative AI increasingly present in classrooms, there is a need to understand whether AI feedback can support revision without replacing students’ thinking. The 10-week intervention involved designing and implementing a custom AI chatbot, Blue Bot. Blue Bot was co-constructed with students through shared success criteria, task-specific rubrics, and clear guardrails to shift AI from a “black box” to a “glass-box” tool. Students engaged in repeated drafting cycles: writing an initial draft, receiving rubric-aligned AI feedback, revising independently, and submitting a second draft for teacher assessment. A mixed-methods approach was used to collect data through student journals, pre- and post-intervention surveys, focus group interviews, chatbot interaction logs, teacher field notes, and rubric-scored writing samples with calculated revision gains. Findings indicate that, when explicitly scaffolded, AI feedback can strengthen evaluative judgement and support deeper revision beyond surface editing. However, the impact of AI feedback varied depending on students’ perceptions of the chatbot, and some learners (including EALD and lower-achieving writers) required additional scaffolding to interpret and apply feedback. The study highlights the importance of transparent design, explicit teaching and “human in the loop” principles to ensure AI supports metacognitive growth and equitable access to revision improvement.
-
Tags
-
Technology, Teaching and Learning
