An audit of 8,946 AI-generated feedback reports on IELTS essays revealed that only 37.7% of quoted student text was an exact character match. While 92.3% of quotes retained the student's words with only minor punctuation or spacing differences, 7.7% altered at least one word. The study found that feedback on weaker essays was more likely to be exact, and the model's internal reasoning text was more faithful than its final output. A follow-up study is planned to precisely measure how often quotes silently change the student's original meaning. AI
IMPACT Highlights potential inaccuracies in AI feedback tools, impacting language learners' ability to identify and correct their own mistakes.
RANK_REASON Analysis of a large dataset of AI-generated feedback reports on IELTS essays. [lever_c_demoted from research: ic=1 ai=1.0]
- Apache Software License 2.0
- Creative Commons Attribution 4.0 International
- GLM-4.7
- Hugging Face
- IELTS
- Jackrong/IELTS-writing-feedback-reasoning
- Munim, Inc.
AI-generated summary · Google Gemini · from 1 sources. How we write summaries →
