Adapting and validating an artificial intelligence literacy scale for Ethiopian higher education using structural equation modeling

Abstract

This study adapts and validates the Artificial Intelligence Literacy Scale (AILS) for higher education faculty in Ethiopia. After expert review and a pilot test with 40 participants identified four items with weak factor loadings, the research team revised those items to improve clarity and contextual relevance. The final 12-item English version was administered to 376 faculty members from four public universities. Exploratory factor analysis using principal axis factoring with oblique rotation confirmed a four-factor structure—Awareness, Usage, Evaluation, and Ethics—that explained 72.89% of the total variance. Confirmatory factor analysis showed good model fit (CFI = 0.977; TLI = 0.969; RMSEA = 0.061; SRMR = 0.049). Reliability and validity tests produced strong results: Cronbach’s α ranged from 0.856 to 0.930, composite reliability (CR) from 0.856 to 0.930, and average variance extracted (AVE) from 0.665 to 0.816. Multi-group confirmatory factor analysis supported configural, metric, and scalar invariance across both data collection methods (online and paper-based) and gender, confirming stable measurement properties. The adapted AILS provides a reliable, concise, and contextually sensitive tool for measuring AI literacy among faculty in low-resource higher education environments. The study highlights implications for curriculum design, faculty professional development, and cross-cultural adaptation of AI literacy assessments.