Human-Centred AI in Education and Governance: Ethical Challenges of Digital Transformation

Abstract

This conceptual article examines the conditions under which AI-assisted decisions in education and public governance can strengthen institutional capacity without displacing human judgement, agency, or accountability. It employs a purposive conceptual synthesis of interdisciplinary scholarship and legal and policy materials and compares governance approaches in the European Union, the Republic of Korea, and the United States with regard to legal force, risk classification, human oversight, transparency, contestability, and institutional capacity. Rather than treating the technical system alone as the unit of ethical analysis, the article focuses on the AI-assisted decision episode: the sequence through which data, model outputs, human judgement, and institutional authority combine to affect a learner, citizen, or community. On this basis, it develops a six-element governance framework comprising legitimate purpose and proportionality; explicit allocation of roles and responsibility; traceable data, evidence, and uncertainty; competent human oversight and calibrated reliance; stakeholder participation, contestability, and redress; and continuous monitoring, audit, and institutional learning. Applied to educational assessment and public-service decisions, the framework demonstrates that a nominal human-in-the-loop is insufficient unless reviewers possess the competence, time, authority, alternative evidence, and records necessary to challenge model outputs. The article’s contribution lies in connecting legal safeguards, organisational capacity, and cognitive risks within a process-based model of human-centred AI. The framework is conceptual and requires empirical validation. Human-centred AI ultimately depends not only on technical accuracy but also on a decision architecture that preserves agency, provides effective remedies, and keeps responsibility visible.