Boundaries of integrity-compliant use of artificial intelligence technologies in the educational and upbringing process of an educational institution

Abstract

The article identifies the boundaries of integrity-compliant use of artificial intelligence technologies (generative language models, adaptive learning systems, automated assessment tools) in educational and upbringing processes across general secondary, vocational, and higher education institutions. The ICAI value framework of academic integrity serves as the coordinate system for determining AI permissibility depending on learning objectives, activity stage, and education level. Objective: to substantiate the teleological principle of boundary determination (the learning activity goal defines the AI permissibility threshold), analyse AI's impact on cognitive learning mechanisms and the upbringing dimension of education, and provide a comparative review of institutional mechanisms for enforcing these boundaries. Methods: systems analysis, comparative analysis of normative documents and institutional policies (UNESCO, Council of Europe, EU AI Act, Russell Group, Australian framework, Ministry of Education of Ukraine), synthesis of cognitive psychology empirical data. Results: AI-generated text detection tools are shown to be systemically unreliable (none of 14 tested achieved 80% accuracy), and the detect-and-punish approach is pedagogically unproductive; the cognitive outsourcing effect is empirically confirmed (11 p.p. knowledge retention decline); AI's disproportionate impact on subjectification is identified; a typology of boundaries by learning activity stages and education levels is proposed. Conclusions: an optimal model for Ukrainian educational institutions should be grounded in the principles of adaptiveness, transparency, pedagogical justification, and values-based foundation.