"Perception of Artificial Intelligence in Education"

Abstract

"Artificial intelligence (AI) is changing how candidates prepare for high-stakes examinations, yet evidence from small-island systems is scarce. This convergent mixed-methods study examined how 171 stakeholders in 15 Caribbean territories perceive AI\u2019s role in examination preparation. Twenty four-point Likert items were analyzed with polychoric exploratory factor analysis, structural equation modeling (SEM), and nonparametric tests. The 137 substantive open-ended responses were thematically coded and integrated via a joint display. 57.3% agreed AI users perform better on high-stakes examinations, versus 88.3% who agreed AI aids preparation. The SEM explained 45% of variance in perceived performance benefit, below the a priori 70% criterion, with perceived pedagogical value as the dominant predictor. Qualitative stance converged with survey scores, and critical-thinking comments converged strongly with the over-reliance item. Some respondents attributed AI-related gains to misuse rather than learning, threatening score interpretation. Implications for assessment design, integrity policy, and instrument refinement are discussed."