Reclaiming the Writing Process: A CEFR-Aligned, Scaffold-First Model for AI in Foreign Language Writing

The writing between the task and the finished text

A finished text does not tell us which ideas the learner chose, which sentences they struggled with or how much of the wording a tool supplied. Those distinctions matter when writing is the practice itself.

The chapter proposes organizing AI assistance around a writing task designed in advance. Each stage asks for a particular contribution. The learner works on it, receives support when needed and builds toward a complete text.

What “scaffold-first” means here

In PrepTalk, the task comes with a purpose, an audience and a sequence of things to understand, practice and write. These are decisions for the activity designer. The AI works within them.

The model connects CEFR-informed task design with limits on assistance. Hints should help a learner make their next decision. The notepad keeps their accepted writing visible as the activity develops. The final editor gives them the whole text to work on.

A framework to investigate

The chapter also describes recording parts of the writing process, such as attempts and progression through stages. These records could help researchers examine where learners need support and how they respond to it.

This is a design argument. Whether the approach improves later, independent writing remains a question for empirical research. The explainers below examine the choices in the design and the evidence needed to evaluate them.

Related background on scaffolding

Wood, Bruner, and Ross described how a tutor can simplify parts of a difficult task while leaving other parts to the learner. Their study concerned young children building a structure. It provides historical context for scaffolding, rather than evidence about AI writing instruction (Wood et al., 1976, pp. 90, 98).