The Effects of Teacher-Mediated GenAI Feedback on Elementary EFL Learners’ Writing = 교사 중재 생성형 AI 피드백이 초등 영어 학습자의 쓰기 결과에 미치는 영향

다국어 초록 (Multilingual Abstract)

While feedback is essential for second language (L2) writing development, providing individualized support presents a significant challenge for teachers in Korean elementary EFL contexts. Generative AI (GenAI) offers a promising solution to these constraints, yet little is known about how its interaction modalities align with the developmental needs of young learners. This study investigated the effects of teacher-mediated GenAI feedback—specifically comparing interactive and non-interactive modalities—on elementary students’ writing performance and emotions (anxiety and boredom). The study involved 106 fifth-grade students (CEFR A1 level) at a Korean elementary school. Over an eight-week quasi-experimental intervention involving three writing-revision cycles, participants were assigned to three groups: Interactive AI Feedback (IAF; n = 42), where students engaged in guided dialogue with AI; Non-interactive AI Feedback (NAF; n = 29), where students applied teacher-curated feedback independently; and a Control Group (CG; n = 35) receiving grammar-focused instruction without systematic writing feedback. Writing performance was assessed using narrative tasks adapted from the Cambridge A2 Flyers Writing Test. Linear mixed-effects models were employed to analyze changes in outcome variables. Results indicated distinct patterns across cognitive and affective domains. For writing performance, a significant Group × Time interaction (p = .046) was observed, with only the IAF group showing statistically significant improvement (d = 0.81). This pattern suggests that dialogic interaction may be an important driver of performance gains. Regarding affect, writing anxiety decreased significantly in both experimental groups (d > 1.5), suggesting that teacher-mediated AI feedback can reduce affective barriers to writing regardless of modality. However, across both time points, the NAF group reported significantly lower writing boredom (p = .005), which may indicate that direct AI interaction can impose excessive cognitive load on young novice writers. These findings highlight a trade-off: while interactive approaches appear to enhance performance, non-interactive and teacher-filtered approaches may be more effective for managing cognitive load and sustaining engagement. Consequently, this study suggests that, in elementary EFL contexts, the pedagogical design of teacher mediation is more critical than technological interactivity itself.

목차 (Table of Contents)

• 1. INTRODUCTION 1
• 2. LITERATURE REVIEW 7
• 2.1 Theoretical Foundations: Sociocultural Perspectives on Feedback 8
• 2.2 From Automated Writing Evaluation to Generative AI 11
• 2.3 Cognitive Load Considerations for Young Language Learners 14
• 2.4 Affective Dimensions of AI-Mediated Feedback 16
• 2.5 The Current Study 20
• 3. METHODS 22
• 3.1 Participants 22
• 3.2 Instruments 23
• 3.4 Procedures 27
• 3.5 Data Analysis 29
• 4. RESULTS 32
• 4.1 RQ1: Effects of teacher-mediated GenAI feedback on Writing Scores 32
• 4.2 RQ2: Effects of teacher-mediated GenAI feedback on Writing Anxiety 36
• 4.3 RQ3: Effects of teacher-mediated GenAI feedback on Writing Boredom 40
• 5. DISCUSSION 44
• 5.1 Effects on Writing Performance 44
• 5.2 Effects on Writing Anxiety 46
• 5.3 Effects on Writing Boredom 48
• 5.4 Theoretical and Pedagogical Implications 49
• 6. CONCLUSION 51
• REFERENCES 55
• 국문요약 67
• APPENDICES 69