Teacher Professional Identity and Artificial Intelligence: A Theoretical Account of Deskilling, Reskilling and the Relational Core of Teaching

Abstract

Artificial intelligence is entering teachers' work from two directions at once. Lesson planning, marking and reporting are being automated, while students arrive with generative tools that change what a teacher is for. Research on teachers and AI is dominated by surveys of attitudes and readiness and says little about what AI does to teaching as a profession: to its jurisdiction, skill base, status and identities. This theoretical paper addresses that gap. Its objective is to explain how AI reshapes teaching as an occupation and under what conditions it deskills or reskills teachers. The design is a conceptual synthesis. Its population is the peer-reviewed literature on teachers and AI, on AI in other professions, and on the sociology of professions, labour-process theory and teacher professional identity. The sample comprises 39 sources: 35 journal articles published 2023–2026 and four foundational works. Sources were chosen by criterion-based (purposive) selection, with every DOI verified against its Crossref record and screened for retractions. Analysis combined thematic coding, by component of teaching and by mechanism, with proposition derivation. The paper decomposes teaching into instructional, administrative and relational components and argues that AI can relieve the second, transforms the first and cannot replace the third. The relief is potential rather than demonstrated, and some administrative tasks carry relational knowledge that automation can strip away. A deskilling–reskilling model holds that the trajectory depends on who governs adoption, whether teachers use AI as generator or collaborator, where released time goes, and whether policy seeks augmentation or substitution, with AI competence and the profession's collective power as moderators. Eight testable propositions link these conditions to skill, relationships, recruitment, retention and status. The paper recommends that teacher education and professional standards place AI competence inside professional identity, tied to judgement and the relational core, and that AI rollouts carry workload agreements stating how released time is used. African systems facing shortages and large classes, including Kenya, Ghana, Nigeria and South Africa, should use AI to protect contact time and qualification standards, not to justify fewer, less qualified teachers managing more students with software.