Providing a comprehensive thematic review of the Science-Policy Interface (SPI): A probabilistic topic modeling approach

Interactions between the scientific community and policymakers have substantially increased over the past two decades, reflecting a collective effort to craft evidence-based policies and foster policy-relevant scientific research. This evolving relationship has given rise to a growing body of academic research aimed at analyzing the characteristics of, or the challenges related to, such interactions often grouped under the concept of ‘ Science-Policy Interface ’. Recent studies have initiated endeavours to consolidate this still fragmented field through global reviews, but such attempts remain scarce, which limits our understanding of this rich and complex concept. This article aims to contribute to this integrating effort by applying a text mining approach based on a latent Dirichlet allocation model on 1048 Scopus-indexed abstracts published from 1964 to 2023. This analysis uncovers 60 distinct latent topics, which are then synthesised into 15 meta-themes using hierarchical clustering. Five of these dominant themes relate to specific challenges inherent to the nature and the dynamics of the knowledge interaction process and its organization within the Science-Policy Interfaces, while the remaining themes pertain to specific issues that are addressed by these interfaces. Persistent gaps in this literature regarding the concepts of trust, power, and the issue of energy are identified. By discussing the interconnectedness of the clusters in relationship with previous reviews, this study not only provides a comprehensive thematic review of the Science-Policy Interfaces field but also identifies pivotal research directions and offers a structured synthesis that can guide future evidence-based policy research and practice. • 1048 Scopus-indexed articles are analyzed to provide a comprehensive thematic review of the SPI field. • 60 dominant topics and 15 themes are identified using an LDA-based topic modeling coupled with hierarchical clustering. • Studies from the natural sciences dominate research on SPI, with a low representation of the arts and humanities. • Persistent gaps in the SPI literature regarding the concepts of trust, power, and the issue of energy are identified. • Co-occurrence analysis reveals several underrepresented intersections between clusters.