The network structure of research on artificial intelligence in school leadership and management: insights from a multi-method approach

Abstract

Even though Artificial Intelligence has become a rapidly growing research focus in the field of school leadership and management, it has not been statistically examined how the knowledge production structure and citation networks of this exponentially developing research field are constituted. Previous research employing techniques such as thematic analysis, co-word, or descriptive mapping has yielded themes and trends; however, limited attention has been paid to thematic and regional clustering of citation ties and to the cumulative advantage of citation counts for the further visibility of scientific works. Using a systematic-search-based scientometric and citation-network design, this study examines the citation network structure of a constructed corpus of 105 peer-reviewed studies on AI in school leadership and management through Exponential Random Graph Models (ERGMs), power-law analysis, and negative binomial regression. The analysis revealed three key findings: First, the observed citation network is sparse and centralized, with citations concentrated around a limited number of high-impact core studies. Second, intermediary studies that connect central and peripheral parts of the network are limited, suggesting weak brokerage and only emerging thematic sub-clustering within the corpus. Third, citation accumulation is highly uneven, with citations concentrated around a small number of highly visible studies, indicating a heavy-tailed and visibility-linked citation structure rather than a confirmed temporal preferential-attachment mechanism. Negative binomial regression showed that broader Google Scholar citation recognition was positively associated with within-corpus citation counts in the primary model, whereas the remaining coefficients were sensitive to the exclusion of this variable. This study contributes to discussions on the epistemic structure of AI research in school leadership and management by providing corpus-based evidence that the observed citation network is sparse, centralized, and uneven.