Toward AI-Scaffolded Pedagogical Stylistics: A Conceptual Framework for Bridging the Language–Literature Divide in L2 English

Abstract

Literature has a long-standing place in second-language (L2) English teaching because it offers learners language that is rich, patterned, and affectively charged, and pedagogical stylistics provides a principled way of turning that language into learning by having students notice the choices through which literary effects are achieved. Recent classroom evidence is sobering. Large-scale observation and competence profiling show that literary texts still appear in L2 classrooms yet work with them seldom moves beyond literal comprehension and personal reaction; furthermore, learners’ aesthetic-stylistic competence lags well behind their interpretive and empathic competence. The reason is structural rather than attitudinal: contingent stylistic guidance, responsive to what each reader notices, cannot be given to thirty readers at once. This article proposes AI-Scaffolded Pedagogical Stylistics (ASPS), a conceptual framework in which a generative language model is positioned not as an interpreter but as a semiotic tool within a teacher-designed activity, one that issues questions directing attention to stylistic features while the learner’s aesthetic response and interpretive authority remain primary. Its contribution is threefold: it defines the construct of AI-scaffolded stylistic noticing by conditions that separate it from neighboring ideas; it specifies the mechanism by which repeated, progressively faded questioning is internalized as a transferable reading routine; and it advances the distinction between scaffolding and substitution as a falsifiable claim, namely the prediction that the same tool used to question rather than to answer will yield more learner-initiated stylistic noticing, and better delayed transfer to unaided reading, a difference traceable in a transcript. A preliminary coding framework and a set of testable propositions set out how the claim can be examined empirically.