Abstract

A sequential claim about collaborative learning is only as defensible as the chain of decisions that produced the sequence. In classroom collaborative game-based learning (GBL), data is typically captured via shared tablet screen recordings with group audio. However, because speakers are unidentifiable, event boundaries are absent, and the interface changes independently, applying different analytical rules can yield entirely different sequences from a single recording. This article specifies this decision-making chain as an ordered framework, asking how evidence admission, event construction, sequence building, analysis, and interpretation should be ordered and reported so that claims remain within the evidence. It was developed through theory synthesis across evidence-centred design, observational methodology, trace-based research, and sequential analysis. The proposed framework consists of six layers: conceptualise, bound, operationalise, build, analyse, and interpret. Socially shared regulation of learning serves as an illustrative lens, with a five-category coding layer. Four principles support auditability: classify screen changes by source before functional coding; retain adjacency only where observed; check event and transition totals; and assess boundary agreement separately from category agreement. A synthetic fragment shows three defensible rule sets producing three different sequences. Future work should test reproducibility, validity, and transfer across recording configurations.