Trading privacy for personalisation: Ethical trust and high school students’ intention to use AI in education

Abstract

Educational AI tools personalise learning by drawing on sensitive student data, yet the privacy calculus assumes that people can freely weigh benefits against risks and decline when risks dominate. This assumption is doubtful for adolescents, who have limited ability to scrutinise institutional data practices and little scope to opt out when AI is embedded in schooling. This study examines how perceived personalisation benefits, perceived data sensitivity and privacy concerns relate to high school students’ intention to use AI, and tests whether ethical trust, defined as confidence that an AI system will handle personal data responsibly, transparently and in the user’s interest, links privacy concerns to intention. Survey data from 323 students across IB, A Level, CBSE, American and other curriculums were analysed using partial least squares structural equation modelling. Perceived benefits strongly predicted ethical trust (β = .605, p < .001) and intention to use AI directly (β = .454, p < .001). Perceived data sensitivity raised privacy concerns (β = .563, p < .001), and privacy concerns modestly reduced ethical trust (β of negative .185, p = .002). Ethical trust in turn predicted intention to use AI (β = .453, p < .001), whereas the direct effect of privacy concerns on intention was negligible (β = .023, p = .507). The model explained 65.9% of the variance in intention. The pattern indicates that privacy concern shapes adoption through eroded trust rather than through avoidance, and that usefulness can build trust without proof of trustworthy data practices. Limitations include the cross sectional design, self reported intentions and weaker measurement of data sensitivity. Schools and providers should therefore demonstrate responsible data governance rather than simply reassure students.