PRIMM in Vietnamese Specialised Informatics Classrooms: A Transfer Synthesis

PRIMM — the Predict-Run-Investigate-ModifyMake pedagogy introduced by Sentance and Waite (2017) and elaborated in Sentance, Waite and Kallia (2019) — is one of the more visible school-level computing-education pedagogies of the late 2010s. Its design intent is to reframe how learners encounter unfamiliar code: rather than writing from scratch, the learner first reads, predicts behaviour, runs to verify, investigates internal structure, modifies along controlled axes, and only then makes a novel artefact in the same family. The pedagogy targets the well-documented difficulty that novices have with reading code and with operating a useful notional machine; it borrows from the worked-example tradition and from the self-explanation literature for its mechanism specification. This paper asks whether PRIMM transfers when re-deployed for Vietnamese specialised-informatics (chuyên Tin, ages 14– 18) classrooms — a cohort whose audience profile is inverted relative to the original UK Key Stage 3–4 target: advanced (selected on a competitive entrance examination), time-pressured (Olympiad season compresses each topic into a small number of weeks), and contest-oriented (the proximal goal is HSG-Tin and VOI medal placement; the distal goal is IOI selection). The transfer-prediction question is therefore not "will PRIMM work" but "which stages of PRIMM produce transfer at this cohort, and which are operationally redundant". The transfer verdict has three parts. First, the Investigate and Modify stages predict positive transfer with mid-to-strong expected magnitude: code-reading-with-active-questioning and controlled-variation-ofgiven-code are both load-management interventions whose mechanism is preserved at the cohort. The Make stage predicts positive transfer at the contest-task surface. Second, the Predict stage predicts diminishing returns for advanced learners on familiar idioms, but remains valuable for novel idioms. The Run stage predicts diminishing returns for advanced learners on canonical inputs but remains valuable for edge-case probing. Third, the deployment template the paper specifies is a compressed PRIMM in which the five stages are weighted unevenly: Investigate and Modify carry the bulk of the lesson; Predict is retained only for novel-idiom encounters; Run is retained as edge-case probing; Make is the contest-relevant transfer task at the chapter or unit close. Six audit-loop criteria specify the implementationfidelity checks a downstream deployment must pass. Per the elixresearches no-cohort scope, the deployment template is described and theoretically validated; empirical replication is identified as a downstream cohort node responsibility and is out of scope here.