Building Bridges Across Disciplines with Network Analysis

Addressing complex scientific and technological challenges often necessitates integrative solutions, which has led to an increased recognition of the benefits of interdisciplinary research (IDR). This thesis explores how network science can help us to understand and promote IDR. We examine specific case studies, including the research esponse to the COVID-19 pandemic and advancements in explainable artificial intelligence, to develop new network analysis-based techniques for visualising and characterising IDR interactions at a macro level. These methods aim to shed light on knowledge transfer and collaboration across diverse research fields. Despite efforts within the scientometrics community to identify and quantify IDR’s prevalence and impact, classifying research papers within a static discipline taxonomy remains challenging in a rapidly evolving scientific landscape. In our work we make use of citation networks to provide evolving representations of research corpora, identifying emerging topics according to dynamic communities. We also map interdisciplinary research knowledge transfer according to structural motifs in these networks. By leveraging methods from explainable artificial intelligence, we highlight important citation network structures that are specific to IDR. Furthermore, we propose novel approaches for representing research papers based on graph neural networks (GNNs). The resulting embeddings can capture local citation network structures, thereby preserving the interdisciplinary context of articles. To promote IDR, we integrate these embeddings into a research paper recommender system (RP-Rec-Sys). By adopting a novel perspective on recommendation evaluation, we demonstrate that our methods can be used to broaden researchers’ horizons by providing more far-reaching and unexpected recommendations while maintaining precision.