Understanding AI Mechanisms Supports Disciplinary Reasoning and Ethical Judgment in K–12 AI Literacy Education

While the urgency of AI literacy has been broadly acknowledged, little empirical research has examined how AI education supports the internalization of AI concepts. Whereas much AI education research reports short-term learning gains, this study investigates how students use AI concepts to mediate collaborative reflection and ethical reasoning after participation in a three-year co-design based cumulative AI education program. Drawing on sociocultural and cultural-historical activity theory, we qualitatively analyzed group interviews with 40 K–12 students in grades 6 and 9. The results show that students fluently used AI concepts as mediating tools. Instead of commonsense explanations and folk theories, students relied on scientific explanations and used AI concepts to explain classifier behavior, social media mechanisms, user engagement, and filter bubbles. These concepts and theoretical orientation to everyday data-driven systems extended to ethical reasoning and critical analysis of practices, structures, and problems in everyday life both at personal and societal levels. The study highlights the potential of cumulative AI education to foster transformative learning and agency, extending AI literacy beyond functional understanding toward volition and abilities to question existing practices, envision alternatives, and take informed action.