Project ยท Dynamic assessment of L2 writing
WriteMed: AI-assisted dynamic assessment of L2 academic writing
The idea
WriteMed brings Vygotskian dynamic assessment into second language academic writing. An AI mediator guides each revision through graduated hints, and the learner remains the author of every change.
Why dynamic assessment
Conventional feedback on L2 writing tends to arrive as a list of corrections, which shows what a learner can already do alone. Dynamic assessment asks a different question: how much support does the learner need to improve the text, and how does that need change over time? In WriteMed, instruction and assessment take place within the same revision cycle, so that each round of revision is also evidence of the learner's zone of proximal development.
Design principles
- Each revision cycle focuses on a single aspect of the draft, namely content, organization, or language, which reduces feedback overload and directs learner attention.
- Mediation is graduated, moving from questions about the learner's own wording toward targeted examples. The level of support a learner needs is treated as evidence of developing ability.
- The AI prompts reflection and revision rather than supplying corrections or model sentences, so the learner keeps ownership of the text.
- Each revision is traced alongside the mediation that preceded it, which supports research on how learners respond to different forms of help.
Yuwei Xia designed WriteMed in collaboration with a technical partner.
Related research
Yuwei Xia's published work on AI-assisted dynamic assessment examines how mediation is distributed across teacher-designed prompts, AI-generated feedback, and teacher mediation. See Distributed mediation in AI-assisted dynamic assessment (Language Learning & Technology, 2026) and a systematic review of AI in second language writing education (Digital Studies in Language and Literature, 2025). For the full body of work, see dynamic assessment research.