The IEEE ACCESS published an article authored by a team of researchers from the University of Aizu in August 2025, titled "Large Language Model-Driven Dynamic Assessment of Grammatical Accuracy in English Language Learner Writing". The team consists of one post-graduate student, Timur Jaganov, and three faculty members: John Blake, Julián Villegas, and Nicholas Carr.

Have you ever wondered how teachers can give students meaningful feedback on their writing? Have you ever wondered if all the hype around AI is going to produce something truly beneficial for learners? That is exactly the type of questions this research team has been investigating.

One scientifically proven way for learners benefit from the feedack they receive from a teacher is for it to be provided in graduated form, i.e. to first receive an implcit hint, and then for hints to gradually become more explicit as needed to give the student the right level of assistance--not too much, not too little. The problem is, though, that this takes time and cannot by done to a whole classroom of learners.

This study explored whether Large Language Models (LLMs) could be used to provide learners with this type of graduated feedback, and thus allow it to be applied in larger classroom settings. To do this, we created DynaWrite, an online tool which allows learners to write a text and have a LLM process their text and provide graduated feedback. After testing multiple LLMs, we found that OpenAI's ChatGPT 4o was capable of providing such feedback at reliable levels. This opens up the door to exciting new ways of teaching and learning foreign languages.

20250905IEEE ACCESS.png

Details of the paper:
Name of Journal: IEEE ACCESS
Authors: T. Jaganov, J. Blake, J. Villegas, and N. Carr.
Title: Large Language Model-Driven Dynamic Assessment of Grammatical Accuracy in English Language Learner Writing
Date of publication: August 2025
Link to publication: https://ieeexplore.ieee.org/document/11142695

About IEEE ACCESS

SCImago Journal & Country Rank

IEEE Access® is a multidisciplinary, open access (OA), applications-oriented, all-electronic archival journal. It has a SCImago Journal Rank indicator of 0.849 (2024), and a H-index of 290. Further details of this journal are available at https://ieeeaccess.ieee.org/about-ieee-access/learn-more-about-ieee-access/