Independent Writing in the Machine Age: The Draft–Interrogate–Reflect (DIR) Model of Human–AI Interaction in Second Language Writing

The use of generative AI in composing and editing text poses a challenge for language educators. On the one hand, writing in a foreign language (FL) is cognitively demanding. In the process of composing texts, errors in FL grammar, punctuation, and syntax (FLG/FLP/FLS) become noticeable, and learners revise them. On the other hand, writing with the help of AI may override the cognitive load necessary for learners to develop their ability to identify and revise errors in FLG/FLP/FLS. Some educators have banned the use of AI for editing and composing text; others have integrated it into their teaching without giving a rational account of when and how the technology should be used. This article aims to bridge this gap. It proposes a synthesis of sociocultural theory, cognitive load theory, and the noticing and output hypothesis. Based on the integration, it identifies the offloading-noticing paradox. Offloading the cognitive load for text revision to AI frees the writer’s working memory for structuring and other aspects of composing. However, relying on AI may hinder the writer from noticing aspects of FL text that need revision. The article provides a resolution to this paradox in the form of the Draft-Interrogate-Reflect (DIR) cycle and a curricular plan for implementing the DIR cycle for a period of six weeks. A validation agenda and six working hypotheses are provided. The article gives a reconceptualization of the role of the teacher in relation to the mediating function of AI for language learning.