Mapping AI learning readiness self-efficacy worldwide: Scale validation and cross-continental patterns

Abstract

In today’s world, knowing how to use artificial intelligence (AI) technologies is becoming an essential skill. While methods for measuring the perceived efficacy of AI use are emerging, brief measures of users’ self-evaluated learning and self-efficacy regarding AI use are still lacking. This study aimed to validate the five-item AI Learning Readiness Self-Efficacy (AILRSE) scale and examine cross-national differences between 12 countries on six continents. We used large-scale, adult population samples from Australia, Brazil, Finland, France, Germany, Ireland, Italy, Japan, Poland, Portugal, South Africa, and the United States collected in 2024–2025 (N = 20,173), enabling both cross-sectional and longitudinal analysis. Scale validation involved confirmatory factor analysis and measurement invariance testing across countries and over time. The results supported a one-factor structure with high internal consistency and scalar invariance across countries as well as strict invariance in Finnish cross-sectional and longitudinal data. AI positivity emerged as the strongest predictor of AILRSE-5 scores across all models, followed by younger age and more frequent use of text-to-text AI tools (e.g., ChatGPT, Copilot). Education and gender effects were small and context dependent. The findings indicate that AILRSE-5 is a brief, reliable, and valid tool for assessing self-efficacy in AI learning readiness. Its invariance across diverse national contexts supports its applicability in cross-cultural research, while its longitudinal invariance suggests stability over time. Furthermore, our results provide rare cross-national evidence on the individual factors shaping AI learning readiness self-efficacy. The study advances understanding of how people adapt to the rapidly evolving AI landscape.

Cite this work

No public copy is stored for this record, so only the metadata is cited.

Stored only when both papers are in the corpus — 10 citations exist upstream.

Papers that cite the same references — the relational fallback while embeddings are still backfilling.