Investigating Neural Substrates of Individual Independence and Interdependence Orientations via Efficiency-based Dynamic Functional Connectivity: A Machine Learning Approach

The self-construal is one of the most significant cultural markers in humans. Accordingly, mapping the relationship between brain activity and self-construal contributes to understanding the nature of such psychological traits. Existing studies have mainly focused on static functional brain activities in specific brain regions. However, evidence has suggested that the functional connectivity of the brain network is dynamic over time and the high-level psychological processes might require collaboration among multiple regions. In the present study, we explored the dynamic connection patterns of the two most representative types of self-construal traits, namely independence and interdependence, using machine learning-based models. We performed resting-state functional MRI (rs-fMRI) on a sample of young adults (n=359) who completed Singelis’ Self-Construal Scale, and constructed the efficiency-based dynamic functional connectivity (FC) networks. XGBoost Regression was subsequently applied to learn the relationship between the dynamic FC and the two self-construals without any priori bias or hypothesis. The performance of the regression model was validated by the nested 10-fold cross-validation. The results showed that the efficiency-based dynamic FC could identify the orientations of independence and interdependence. The comparison analyses revealed that prediction accuracy using this dynamic FC method was significantly improved compared to the conventional static FC method. By exploring key connectivities selected by the regression model, we observed that the independence orientation was mainly characterized by the right-hemisphere FC, while the interdependence orientation by the left-hemisphere FC. The results suggest that the self-construals are associated with distributed neural networks the entire brain. These findings provide the pivotal ingredients toward the biological essence of culturally related variables in the brain by taking advances in cultural psychology, neuroscience, together with machine-learning analytic technologies.