The Short General Attitudes Towards Artificial Intelligence Scale in Japan: Psychometric Validation and Exploratory Measurement Invariance with the Original UK Sample

Abstract

As artificial intelligence (AI) becomes increasingly widespread, reliable measures of general attitudes toward AI are needed to understand public responses, and cross-national comparisons require evidence of measurement equivalence. This study addressed two objectives: validating the Short General Attitudes Towards Artificial Intelligence Scale (Short GAAIS-10; Schepman & Rodway, 2026) in Japan and exploring its comparability with the original UK sample. We translated the scale into Japanese (GAAIS-10-J) and examined its psychometric properties among 468 Japanese adults, including a one-month retest (N = 429). The original two-factor structure (Positive and Negative) received partial support, although some confirmatory factor analysis fit criteria were not met. Most items showed acceptable functioning, and both subscales demonstrated good internal consistency and adequate temporal stability. The two subscales captured related but distinguishable aspects of AI attitudes. Furthermore, the GAAIS-10-J explained additional variance in AI usage intention beyond perceived usefulness and ease of use. An exploratory comparison with the original UK sample (N = 500) approximately supported threshold and metric invariance but not full scalar invariance. Under a partial scalar invariance model, the Japanese sample had a higher latent Positive mean and a lower latent Negative mean than the UK sample, although these differences cannot be attributed directly to national or cultural factors. Overall, the GAAIS-10-J offers a brief measure of general AI attitudes in Japan, while the findings highlight the importance of accounting for measurement differences in cross-national comparisons.