Central events in self-generated thoughts: Inferring short-term mental health status through narrative networks

Self-generated thoughts are closely associated with individuals' mental health status. Recent advances in AI technology enable the analysis of natural language materials with greater precision, offering new opportunities to examine this relationship. However, it remains unclear how the contents of self-generated thoughts reflect mental health status, especially in distinguishing individuals with higher levels of psychological distress. In the present study, natural language processing techniques and a computational network model were used to transform the self-generated thoughts of 71 healthy participants into narrative networks whose nodes were assigned emotional attributes. Central events within each participant's narrative network were defined using three approaches: 1) hub of the overall network, 2) provincial hubs within modules, and 3) cluster centers of nodes with emotional attributes. Our findings indicate that the emotional experience of an individual's central events (excluding provincial hubs) showed a significant negative correlation with their psychological distress. Furthermore, the emotional experience of a central event was associated with its subsequent thought, suggesting that the emotional experience of central events may spread and influence subsequent thinking. Overall, this study demonstrates how AI-based narrative network analysis can capture meaningful structure in self-generated thoughts and provides insight into how the emotional properties of central events may reflect short-term mental health status.