Adaptation and validation of the generative AI dependency scale: evidence from construct validity, reliability and measurement invariance.

This study aimed to adapt the Generative AI Dependency Scale into Turkish and to evaluate its psychometric properties in a Turkish sample. The study was conducted with 640 participants. The adaptation process was carried out through translation, back-translation, expert review and cultural evaluation procedures. Construct validity was examined using confirmatory factor analysis (CFA) and both first-order and second-order models were tested. The findings supported the original three-factor structure of the scale, consisting of cognitive preoccupation, negative consequences, and withdrawal. The fit indices indicated that both models demonstrated good fit. Reliability analyses showed that the Turkish form of the scale had acceptable to high internal consistency coefficients for both the subdimensions and the total score. Additional evidence based on the measurement model was obtained through average variance extracted (AVE) and composite reliability (CR) values. Measurement invariance analyses further showed that the scale achieved strict invariance across all examined demographic and AI-related groups. Overall, the findings provide evidence supporting the validity and reliability of the Turkish version of the Generative AI Dependency Scale for assessing dependency on generative artificial intelligence systems.