한국어 쓰기 교육을 위한 AI 피드백과 교사 피드백 비교 연구 -피드백에 대한 혼합적 평가를 중심으로-

초록

This study compares AI-generated and teacher-provided feedback on Korean L2 writing through a mixed-methods meta-evaluation. Forty intermediate-level learner texts—20 Level 3 narrative essays and 20 Level 4 opinion essays—received feedback from three teachers and ChatGPT, powered by GPT-5.2 Pro. The resulting 80 feedback samples were evaluated by two expert raters on five criteria: accuracy, specificity/usefulness, prioritization, proficiency fit, and supportive tone. While overall scores showed no statistically significant difference, criterion-level analyses revealed differentiated strengths: AI feedback scored higher in specificity and prioritization, whereas teacher feedback excelled in accuracy, proficiency fit, and supportive tone. The teacher advantage—particularly in supportive tone—was more pronounced in Level 3 narrative writing than in Level 4 opinion writing. These findings suggest that AI feedback functions best as a complement to teacher feedback, and the study proposes a task-sensitive hybrid workflow aligning prompt design with genre demands and learner proficiency.

키워드