A reproducible science-mapping pipeline for educational-technology research domains: Design, validation, and application to game development education

Background: Bibliometric mapping of educational-technology research domains is typically performed through GUI-driven tools that hide parameter choices behind dialog boxes, complicating reproducibility and sensitivity analysis. We address this gap with an open Python pipeline and apply it to a domain that has itself escaped systematic mapping: game development education.Objectives: We (i) release a reproducible pipeline pairing VOSViewer exports with seeded NetworkX analyses, (ii) test pre-registered Rogers/TAM hypotheses, (iii) characterise game-development education's intellectual structure, and (iv) quantify corpus completeness.Methods: The pipeline ingests 1,511 Scopus-indexed documents (1997-2026) under a platform-agnostic Boolean query. It runs seeded Louvain detection with stability sweeps over five γ values, four min-occurrence thresholds, and 500-replicate bootstrap, plus bibliometrix-standard analyses (Lotka, Bradford, Callon thematic map, trend topics, three-field plot).Results: The cumulative-publications curve fits a Gompertz S-curve at R2 = 0.994. Author productivity follows a steep Lotka law (α = 3.38). The corpus is geographically diverse (88 countries; HHI 0.065). The keyword network reveals eight thematic clusters; three are motor themes (game development; serious games / game-based learning; education) and one is the immersive-tech / Unity cluster, now peripheral. On a controlled comparison against a biased-query subset, Unity degree centrality collapses by 89% - evidence that retrieval-design choices can manufacture centrality artefacts.Conclusions: The pipeline plus its application yield a sensitivity-aware bibliometric mapping of game-development education, a query-bias case study showing how artefactual centrality results can be detected, and a literature-informed curriculum proposal grounded in GRADE-style learning-outcome evidence.