A Citation Discipline Framework for Vietnamese Upper-Secondary and Early Tertiary Writers

Vietnamese upper-secondary and early-tertiary student writers (audience 14–19) approach citation under two compounding pressures: a Confucian-heritage academic tradition in which the explicit attribution norms of Western academic writing were never institutionally enforced at the secondary level, and a 2024–2026 AI-tooling environment in which large language models routinely fabricate plausible-looking bibliographic entries. The Document Engineering Vol 4 first edition needs a citation pedagogy that transfers academic-writing research into this audience without inheriting either the formatting-fixation failure mode of imported Western style guides or the patchwriting failure mode catalogued in the second-language-writing literature. This paper specifies a citation discipline framework of ten named practices: source-capture-at-discovery; paraphrasewith-attribution discipline; direct-quote-rationale rule; singlesource bibliography; secondary-source explicit labelling; AIsummary-as-secondary; claim-citation mapping; citation-density as signal; reflective citation revision; and verify-before-pasteAI. Each practice is presented with a definition, a Cognitive Load Theory classification (intrinsic, extraneous, or germane), a desirable-difficulty rationale, a common failure mode rooted in the Vietnamese student context, and a worked example anchored in a capstone-project setting typical of the audience. The framework is engineered against three rules: practices preserve generative friction (Bjork desirable-difficulty programme); practices reduce extraneous formatting friction so cognitive resources flow to evidence reasoning rather than punctuation rules; and practices are auditable at chapter scale by a single auditor. The framework draws on Booth, Colomb and Williams' Craft of Research, Hyland's citation-practices programme, the Pecorari and Howard patchwriting/plagiarism research, the Sweller 2019 cognitive-load update, the Bjork desirable-difficulty programme, Mayer's multimedia principles, the Bastani 2024 GPT-Tutor randomised trial, and the 2023–2024 Walters and Magesh studies on LLM citation hallucination. Three open questions — bilingual citation discipline, institutional friction with Vietnamese examination boards, and the long-term stability of verify-before-paste under agentic citation tools — are flagged for downstream Layer H research. The deliverable is the framework itself, which serves as the research backing for Vol 4 core chapters.