Automated Creativity Prediction Using Natural Language Processing and Resting-State Functional Connectivity: An fNIRS Study

Abstract

ABSTRACT Evidence from fMRI research indicates that individual creative thinking ability – defined as performance on divergent thinking tasks, subjectively assessed by human raters – can be predicted based on the strength of functional connectivity (FC) between the brain’s default mode network (DMN) and frontoparietal control network (FPCN). Here, we sought to replicate and extend these findings in two ways: 1) using a natural language processing method to objectively quantify creative performance (instead of subjective human ratings), and 2) employing functional near-infrared spectroscopy (fNIRS), a neuroimaging method that allows measuring brain activity in more naturalistic settings (compared to fMRI). By applying elastic-net regression to resting-state functional connectivity data, we constructed two separate prediction models to predict participants’ creative performance based on static FC and dynamic FC respectively. Results from the static network analysis indicated that fNIRS-functional connectivity between the DMN and FPCN can reliably predict creative ability (assessed objectively via natural language processing; R2 = .38). Moreover, we show that dynamic DMN-FPCN functional connectivity predicts creative ability nearly twice as strong as static connectivity (R2 = .67). Our work demonstrates that objective measures of creativity can be predicted from resting-state functional connectivity and that the procedure can be efficiently implemented within highly naturalistic settings with fNIRS.