Development and validation of an instrument to measure undergraduate students' attitudes toward the ethics of artificial intelligence (AT-EAI) and analysis of its difference by gender and experience of AI education

Abstract

As artificial intelligence (AI) becomes more prevalent, so does the interest in AI ethics. To address issues related to AI ethics, many government agencies, non-governmental organizations (NGOs), and corporations have published AI ethics guidelines. However, only a few test instruments have been developed to assess students' attitudes toward AI ethics. A related instrument is required to effectively prepare lecture curricula and materials on AI ethics, as well as to quantitatively evaluate the learning effect of students. In this study, we developed and validated the instrument (AT-EAI) to assess undergraduate students' attitudes toward AI ethics. The instrument's reliability, content validity, and construct validity were evaluated following its development and application in a sample of 1,076 undergraduate students. Initially, the instrument comprised five dimensions that totaled 42 items, while the final version had 17 items. When it came to content validity, experts (n = 8) were involved in the process. Exploratory factor analysis identified five dimensions, and confirmatory factor analysis found that the model was good-fitting. The reliability analysis using Cronbach's alpha and the corrected item-total correlation were both satisfactory. Considering all the results, the developed instrument possesses the psychometric properties necessary to be considered a valid and reliable instrument for measuring undergraduate students' attitudes toward AI ethics. This study also found that there were gender differences in fairness, privacy, and non-maleficence dimensions. Furthermore, it revealed the difference in students' attitudes toward fairness based on their prior experience with AI education.