When AI Gives the Feedback: Evaluative Outsourcing and Learner Agency in the English Writing Classroom

Generative artificial intelligence (GenAI) is increasingly used to provide feedback and support revision in English writing. Existing debate has largely examined the quality, accuracy, and effectiveness of AI-generated feedback. This conceptual paper argues that such concerns risk overlooking a more fundamental pedagogical issue: the redistribution of evaluative work between learner and system. Drawing on feedback literacy and evaluative judgement, the paper introduces evaluative outsourcing to describe the delegation of interpretive, judgemental, and decision-making work in feedback engagement to AI. It argues that GenAI can enable substantial textual improvement while requiring limited learner participation in the evaluation underlying revision, creating an AI feedback paradox in which better revision does not necessarily indicate better judgement. The paper proposes four principles for retaining learner evaluative agency: making evaluative decisions visible, designing for comparison, preserving consequential decisions for learners, and creating opportunities for evaluative transfer.