Abstract
In recent years, research systems have faced increasingly complex sustainability challenges, and improving system resilience has become the key to ensuring their continuous and efficient operation. Based on the data of Web of Science co-authored papers, we constructed a research cooperation network from 2005 to 2024. Subsequently, complex network analysis and principal component analysis (PCA) were used to quantitatively assess the overall resilience of the research system in terms of structural attributes and functional resilience, and to explore the adaptive mechanisms and evolutionary trends of the system in the face of external shocks by combining with the identification of policy events. The results of the study include: ① Policy events have a phased impact on the resilience of the system, and the study captures the dynamic response to resilience under policy shocks with the help of the difference-in-differences approach and finds that the scientific research system shows a certain degree of resilience in the face of the Sino-US scientific research friction around 2018; ② The study defines the resilience of the system as structural and functional resilience. The structural toughness of the research system reflects the structural robustness of the network in different dimensions, such as hierarchy, matching, access, aggregation, and heterogeneity; ③ Functional toughness is quantified based on the area under the curve of GC (Giant Component), which intuitively reflects the ability of the scientific research system to maintain functional output when suffering from local node failures; ④ The comprehensive resilience assessment framework uses PCA to effectively integrate the structural resilience indicators of the system and combines structural and functional analyses. It is found that the overall resilience of the research system during the research period shows a ”fluctuating but improving” trend: The early stage of the network structure is loose, the functional stability is weak, and the impact resistance of the system is limited. However, with the expansion of the collaborative network and the optimization of the structure, the system gradually demonstrates stronger recovery and adaptive capacity.