This academic article contrasts automated writing-evaluation feedback with generative-AI feedback such as ChatGPT and proposes integrating AIF into process-based writing alongside teacher, peer and learner verification. It organises accuracy, dependence, bias, privacy and ethics across writing stages, but includes no learner trial and does not establish improved skill or safety.
Key findings
- The framework positions AI as supplementary feedback during planning, drafting, revision and reflection, with learner evaluation and teacher safeguards against hallucination, overreliance and opaque use. Claimed learning potential remains theoretical.
Why this matters globally
The framework helps institutions build AI literacy and assessment policy without blanket prohibition or uncritical adoption, but needs testing across languages, proficiency levels, platforms and data policies.
Thai researcher contribution
Watcharee Kulprasit of Thaksin University brings a Thai EFL-writing perspective to international generative-AI pedagogy. Its current contribution is design logic, not trial evidence from Thai students.
Limitations to consider
There is no systematic review or comparative evidence on learning, equity or adverse effects. Models and policies change quickly, ChatGPT-specific assumptions may not transfer, and the framework may add workload.