JALT2025 - Tokyo

Comparing AI and Human Grading of EFL Student Essays

Oct 31, 2025, 2:50 PM

30m

National Olympics Memorial Youth Center / 国立オリンピック記念青少年総合センター

Research-oriented Presentation (30-minutes) Technology Room 514

Speaker

Steven MacWhinnie (Hirosaki Gakuin University)

Description

This study compares ChatGPT’s performance to human instructors in grading Japanese university students' EFL essays. Results highlight strong alignment on objective elements like formatting, but significant discrepancies in areas such as body paragraph development and language use. Findings support the cautious integration of AI as a grading aid, with human oversight remaining essential for subjective writing criteria.

Summary

This study compares ChatGPT’s performance to human instructors in grading Japanese university students' EFL essays. Results highlight strong alignment on objective elements like formatting, but significant discrepancies in areas such as body paragraph development and language use. Findings support the cautious integration of AI as a grading aid, with human oversight remaining essential for subjective writing criteria.