Infrastructural-epistemometrics: mapping fragmentation, selective coupling, and bias in library and information science (2000–2024)

Purpose This study investigates how bibliometric infrastructures shape epistemic visibility in Library and Information Science (LIS), examining the field’s epistemic structure, the relationship between topical attention and conceptual centrality, and the geographical, linguistic, and thematic biases embedded in database representations. Design/methodology/approach An infrastructural-epistemometric approach is employed, integrating quantitative network analysis with critical infrastructure studies. A combined dataset of 16,083 unique LIS publications (2000–2024) from Web of Science and Scopus is constructed. Co-word network analysis focuses on the 2017–2024 subset (n = 6,633), supplemented by longitudinal comparison and cross-database metrics to address three interrelated research questions. Findings The analysis reveals a moderately fragmented epistemic architecture comprising 7 distinct thematic communities with sparse connectivity (modularity Q = 0.36). Results provide evidence consistent with selective, rather than deterministic, coupling between topical attention and conceptual centrality, with structural network properties accounting for the majority of the variance in centrality. The Pearson correlation between attention and centrality is r = 0.94, while the Spearman correlation is ρ = 0.80, indicating a strong but not perfectly linear relationship. Comparative assessment suggests databases actively construct divergent disciplinary realities, evidenced by 44% non-overlapping content and a dominant Anglophone bias (92.5% English). Originality/value This study provides empirical evidence that bibliometric infrastructures function as active epistemic gatekeepers rather than neutral mirrors, co-constructing visibility and marginalizing non-dominant knowledge. It challenges linear attention–centrality assumptions and advances the infrastructural-epistemometric approach, underscoring the imperative for reflexive, pluralistic evaluation practices in LIS.