複雑な社会問題への対応に向けたメカニズム的エビデンスの役割

―EBPM+におけるモデル及びデータの要件―

抄録

This paper discusses the necessity of mechanistic evidence in the context of evidence-based policy making plus (EBPM+), particularly for addressing complex social issues. Traditional EBPM relies heavily on randomized controlled trials to establish statistical causal relationships; this method faces limitations in non-laboratory settings like public health. To overcome these limitations, EBPM+ incorporates mechanistic evidence, which provides a deeper understanding of the interactions and factors influencing a phenomenon. This approach is crucial for handling multiple dynamic variables and uncertain situations, such as the COVID-19 pandemic. The requirements for adequate mechanistic evidence in EBPM+ are outlined with an emphasis on the need for state spaces that describe boundary conditions of interactions, clear distinctions between empirical and actual realities, and consideration of the time constants for changes in structures and elements. The importance of adapting data environments is also highlighted, particularly regarding the granularity and timeliness of public statistics, to support the application of mechanistic evidence. The discussion includes the challenges and solutions for integrating real-time data into policy making models, while underscoring the potential of agent-based models for scenario simulation in the EBPM+ framework.