Answers are grounded in this paper and cite their sources.
Purpose This study examined whether Agile Product-Focused Scaffolding (APFS) with teacher-orchestrated large language model (LLM) use could support postgraduate academic writing, particularly method-related reasoning and transparent reporting, in a short research-writing course.Methodology A quasi-experimental mixed-methods design was implemented in a 9-week postgraduate academic writing course involving 89 Chinese postgraduates in two intact classes. The comparison group (n = 42) received lecture-centric instruction, whereas the APFS group (n = 47) received APFS combining staged sub-products, short-cycle checkpoints, rapid teacher feedback, and bounded LLM use. Data included pre- and post-test measures of knowledge, skills, and attitudes (KSA), analytic-rubric ratings of authentic group papers, and reflective diaries from the APFS group. Quantitative and qualitative findings were interpreted together to identify both overall and scaffold-targeted effects.Findings No significant between-group differences were found in post-test KSA or overall paper quality. However, the APFS group performed significantly better in Method Fit and Transparency, the rubric dimension most directly targeted by the intervention. Reflective diaries indicated that teacher scaffolding anchored research direction, clarified methodological choices, and externalised reasoning, while bounded LLM use supported outline refinement, instrument development, methodological clarification, and data-related tasks. The pattern therefore suggests a specific rather than broad instructional effect.Originality/value The study contributes a practical model for integrating LLMs into postgraduate academic writing through visible, checkpointed, teacher-orchestrated workflows. It highlights the importance of aligning AI use with clearly defined pedagogical purposes and shows how such orchestration may strengthen method-related reasoning and transparent reporting.
No takes yet. Share an insight, caveat, or question.
Xu et al. (2026) studied this question.
Synapse has enriched 10 related papers. Consider them for comparative context: