Developing a valid and reliable scale to evaluate decision making processes of artificial intelligence applications

This study was conducted to develop a valid and reliable measurement tool for multidimensional assessment of individuals awareness, usage behavior, trust perception, and general attitudes toward artificial intelligence-assisted nutrition applications. The study was conducted following systematic scale development procedures. Content validity was assessed through a panel of seven experts; pilot (n=44) and main (n=220) applications were conducted in Turkey, while test-retest reliability was evaluated in a subsample of 30 participants residing in Germany with a two-week interval. Item-total correlation analysis, exploratory and confirmatory factor analysis, Cronbach's alpha, composite reliability, convergent and discriminant validity, and Intraclass Correlation Coefficient were used in data analysis. The AI-ANB Scale reached its final structure consisting of four subscales and 22 items. The four-factor structure explained 61.35% of the total variance; confirmatory factor analysis fit indices were found to be acceptable. The overall reliability coefficient was 0.936 and the test-retest ICC value was 0.974. While AI-assisted application usage and higher education level were positively associated with scale scores, scores decreased with increasing age and BMI. Regression analysis revealed that application usage status and AI knowledge level were the strongest predictors. The AI-ANB Scale is a psychometrically robust measurement tool that multidimensionally assesses attitudes toward AI-assisted nutrition applications and is the first known instrument in the literature in this field. Its use in academic research and clinical practice is recommended.