Learning from Harm: Incident-Based Ethics Education for AI in the Classroom

This paper makes three contributions to AI ethics in the field of education. First, it provides a structured examination of the ethical challenges associated with the use of artificial intelligence in educational settings, including student growth, AI-generated feedback, and the limitations of AI in supporting student mentoring, emotional development, and holistic learning. Second, it demonstrates how real-world AI failures can be used as educational tools through a systematic analysis of education-related incidents from the Artificial Intelligence Incident Database (AIID). Five representative incidents, including algorithmic bias in university admissions, generative AI-assisted academic dishonesty, facial recognition failures during online examinations, and AI-powered remote proctoring, are analysed using a common framework to identify recurring ethical issues and illustrate how abstract ethical values manifest in real educational contexts. Third, the paper proposes the integration of incident-based AI ethics education into AI curricula. Drawing upon evidence from educational literature and AIID case studies, it argues that exposing students to documented AI failures can strengthen ethical reasoning, improve awareness of unintended consequences, and encourage responsible AI development. The paper concludes that human–AI collaborative models, in which AI supports teaching and assessment while educators retain meaningful oversight and professional judgement, represent the most responsible approach to educational AI. Consequently, AI ethics, grounded in real-world incidents and critical reflection, should become a foundational rather than optional component of AI education for future practitioners designing technologies that will shape learning and society.