Delegated Authority and Algorithmic Power. A Rapid Review of Ethical Issues in AI-Based Educational Assessment

The growing use of AI-driven automated assessment tools in education brings complex ethical challenges that demand careful examination. A review of forty-four peer-reviewed studies spanning 2015-2025 highlights ten core dimensions (transparency, fairness, privacy, accountability, learner agency, motivation, accessibility, integrity, regulatory compliance, equity) each characterizing a critical facet of AI assessment tools. Analysis uncovers persistent tensions between efficiency and learner autonomy, accuracy and explainability, personalization and surveillance, and compliance and innovation. Conflicts threaten trust, equity, and agency when regulations such as the EU AI Act, GDPR, and UNESCO guidelines are implemented without adaptation to educational contexts. The alignment of ethical principles with identified tensions points to the need for design principles and governance frameworks capable of reconciling competing values and ensuring educational justice in AI-mediated assessment.