クロセ リョウ   Ryo KUROSE   黒瀬　亮    所属   京都外国語大学  ランゲージセンター    職種   講師
発表年月日	2025/05/17
発表テーマ	Comparing JLPT and ChatGPT-generated reading comprehension items
会議名	PanSIG 2025
主催者	Japan Association for Language Teaching (JALT)
学会区分	全国学会
発表形式	口頭（一般）
単独共同区分	共同
開催地名	Kanda University of International Studies
発表者・共同発表者	Gilbert Dizon・Ryo Kurose
概要	The study examines the use of generative AI in language assessment by comparing JLPT reading comprehension items created by human experts with those generated by ChatGPT. Using sample JLPT materials as a basis, AI-generated passages and test items were developed and evaluated by Japanese speakers through a blind survey. The results showed no significant differences in the quality of the reading comprehension items. One possible explanation is that participants were not aware of the study’s exact purpose and may have assumed that all passages were created using generative AI. The findings suggest that ChatGPT can be used to create mock JLPT materials to save time and cost; however, it should not be relied upon as the sole resource.