Analisis Psikometrik Modified AI Literacy Scale (MAILS) Menggunakan Model Rasch pada Mahasiswa di Sulawesi Tenggara

Abstract

This study aimed to analyze the psychometric quality of the Modified AI Literacy Scale (MAILS) using the Rasch Model. A quantitative approach with a psychometric analysis design was employed to evaluate 34 instrument items adapted from the MAILS, which were administered to 102 university students. Data were analyzed using Winsteps software by examining instrument reliability, separation indices, item fit, item difficulty, person fit, and the Wright Map. The results indicated that the instrument demonstrated excellent psychometric properties, with both person reliability and item reliability of 0.95, and a Cronbach's alpha ranging from 0.96 to 0.97. The person separation and item separation indices showed that the instrument effectively distinguished respondents' ability levels and item difficulty levels. A total of 25 items (73.53%) met the fit criteria, 5 items (14.71%) were classified as borderline fit, and 4 items (11.76%) were identified as misfit, indicating the need for further refinement. Item difficulty ranged from −1.02 to 1.64 logits, suggesting that the instrument is capable of measuring AI literacy across varying levels of student ability. Overall, the Modified AI Literacy Scale (MAILS) demonstrated satisfactory validity and reliability and is suitable for assessing AI literacy among university students in Indonesian higher education.