Pedagogical Agency of Teachers as Algorithm Users in AI-Supported Education

Abstract

This paper examines how the spread of artificial intelligence in education changes the professional position of the teacher. Its purpose is to clarify what pedagogical agency means when a growing share of classroom judgement is mediated by algorithmic systems, and to specify the conditions under which teachers remain authors of pedagogical decisions rather than executors of machine outputs. The method is conceptual analysis and framework construction. Three bodies of literature are brought into dialogue: ecological and sociocultural accounts of teacher agency, research on artificial intelligence in education, and scholarship on algorithmic accountability, human oversight and automation in human factors research. The analysis shows that the teacher's role is shifting from user of technology to user of algorithmic outputs, a shift that relocates agency from operating tools to interpreting, contesting and taking responsibility for recommendations, scores and predictions produced elsewhere. Building on this diagnosis, the paper proposes the Algorithmic Pedagogical Agency Framework, an original five-dimensional model comprising algorithmic literacy, interpretive discretion, override and contestation rights, data stewardship, and collegial and institutional conditions, together with six propositions that link the dimensions to pedagogical agency. The framework identifies where agency is gained and lost in recommendation systems, automated grading and learning analytics dashboards, and specifies the conditions that convert nominal human oversight into real agency. Its novelty lies in treating agency as a property of sociotechnical arrangements rather than of individual teachers, and in naming accountability without authority as the central failure mode of current deployment. The paper concludes with implications for teacher education and for the procurement of educational AI.