From diagnostic criteria to change processes: validation of the Korean process-based assessment tool using a machine learning approach

Background: This study provides initial validity evidence for the Korean-adapted Process-Based Assessment Tool (PBAT) using a Boruta machine learning approach. Methods: A community sample of Korean adults was employed (N = 1000; M_age = 40.19, SD = 10.45). Item-level criterion validity was evaluated using the Boruta algorithm, which identified important PBAT items by comparing each item to randomized shadow variables across multiple iterations for outcomes (e.g., psychological distress, life satisfaction, and frustration). Temporal consistency was examined through test–retest reliability and within-person (ipsative) stability using Pearson correlations, D² indices, and Q-correlations using a subsample (n = 300) that completed a second PBAT 30 days later. Results: Negative selection items, such as difficulties expressing emotions (PB18), and negative variation items, such as feeling stuck (PB12), were among the strongest predictors of outcomes. Test–retest correlations demonstrated moderate to strong temporal stability; ipsative analyses revealed that most participants maintained stable profiles. Conclusions: Overall, our findings support the PBAT as a culturally adaptable tool for tracking core psychological processes in interventions. The study’s limitations include a reliance on self-reporting, which can inflate shared method variance, and the Boruta algorithm’s inability to clarify causal pathways. Future studies should incorporate experimental or longitudinal designs.