Can data science achieve the ideal of evidence-based decision-making in environmental regulation?

Data science, including big data analytics and machine-learning techniques, is increasingly utilized in environmental governance. This study explores the perspectives of environmental experts on the limitations of data science in evidence-based environmental regulations. Based on semi-structured interviews with 20 environmental experts, this study examines three key limitations: the lack of local specificity in big data, the uncertainty of the environment as an open system, and the opacity of machine learning. The main findings of the study are as follows: First, the lack of spatial and temporal specificity in big data poses challenges in using them as regulatory evidence, requiring revalidation through traditional assessment methods. Second, while machine learning performs well in closed virtual environments, it is susceptible to the fluid and uncertain nature of the environment in an open world. Third, the lack of causal explanation owing to the inscrutability of deep learning limits its utility for regulatory evidence. The study concludes that while the applicability of data science to evidence-based decision-making in environmental regulation is limited, it can still be utilized in data-driven environmental governance without the need for region-specific information or a high burden of decision-making responsibility. This includes tasks such as strategic environmental impact assessments, policy resource allocation, or environmental policy prioritization.