Abstract

This study aimed to validate the Korean version of the Attitudes Toward Artificial Intelligence Scale (ATTARI-12), which measures attitudes toward AI across cognitive, emotional, and behavioral dimensions. Although ATTARI-12 was developed to reflect three psychological components, it was originally designed with a unidimensional structure. The scale was translated and administered to 223 Korean adults aged 19 to 69. Exploratory factor analysis showed a two-factor structure separating positively and negatively worded items. To verify this, confirmatory factor analysis was conducted with another independent sample of 223 adults. Competing model analysis was performed, comparing the two-factor model according to the exploratory factor analysis, a bifactor S-1 model reflecting method effects of negative wording, and a modified bifactor S-1 model based on modification indices. The bifactor S-1 model, structurally consistent with the original model, showed the best fit, and the modified version of the bifactor S-1 model was confirmed as the final model. Both convergent and discriminant validity were found to be acceptable. Implications, limitations, and future directions are discussed.