Unveiling the Nexus of Science and Policy Communities: Studying the Science-Policy Interface Using Natural Language Processing and Network Science

This study employs computational methodologies to understand the dynamics underpinning the interfaces between scientific evidence and policy making, using New Zealand's COVID-19 response as a case study. My research motivated by governmental rhetoric emphasising "following the science" amid the COVID-19 pandemic response and long-standing theoretical quandaries surrounding the role and uptake of research in policy making, I apply Natural Language Processing (NLP) and network science techniques to furnish innovative quantitative insights into two-communities theory, timeliness and relevance barriers, and the knowledge dissemination process.

The research employs the exploratory approaches to examine the process of science-based policymaking through network analysis, while also investigating and providing explanatory insights into patterns related to the two-communities theory and barriers to research application.

Linguistic analyses displayed contrasts in lexical preferences between the scientific and policy spheres. Academic publications exhibit predilections for disciplinary jargon while policy documents orient towards more accessible cross-disciplinary and stakeholder-centric language, potentially hindering knowledge transfer. However, there are also common languages that exist between two communities, which provide the bridge for communication.

Contrary to self-reported findings, my research did not observe the perceived time pressures on policymakers from the linguistic perspective.

In terms of barriers for the application of research in policy making, my analysis indicates that, in fact, scientific publications ahead of government policy decisions on interventions do provide timely information—potentially actionable by policymakers. However, content similarity between research outputs and policy documents addressing common intervention topics was modest.

To elaborate the details of this process by how knowledge flows from science into policy, we develop three species of networked representations capturing the interactions between: (1) scientific publications and policy documents as knowledge products; (2) actors and stakeholders represented by scientists, policymakers, institutions, government agencies, and parliamentary members; and (3) disciplinary knowledge domains embodied in the disseminated information. These multi-faceted, dynamic network analyses identify key actors and pivotal documents in ways instrumental to delineating how scientific knowledge is transformed and re-contextualized to policy formulation.

Unexpectedly, the study found limited direct citations of scientific literature in key policy documents and an unanticipated prominence of non-COVID-19 topics in central network positions. Analysis of parliamentary debates revealed that scientific knowledge was often transformed into political and governance issues rather than directly informing policy. These findings challenge conventional assumptions about the straightforward application of scientific evidence in policy making, especially during crises.

The research also uncovered an asynchronous relationship between global scientific output and local policy development, particularly evident from December 2020 to February 2021. This observation highlights the complex, non-linear nature of science-policy interactions and the importance of considering local context in knowledge dissemination processes.

The computational approach application in the science-policy interface area as well as the quantitative evidence in my research complementing and expanding findings from prior qualitative studies. This thesis provides nuanced insights into the complexities underlying evidence-informed policymaking processes. By revealing unexpected patterns and gaps in knowledge dissemination, it contributes to a more sophisticated understanding of science-policy interactions. The methodological approach developed here offers a framework for future investigations into knowledge transfer dynamics in various policymaking contexts, potentially leading to more effective integration of scientific evidence in policy decisions.