ChatGPT 5.0을 활용한 대학수학능력시험 영어 읽기 영역 문항 풀이 피드백의 효과성 탐색 = Exploring the Effectiveness of ChatGPT 5.0-Based Feedback on Solving Korean College Scholastic Ability Test(CSAT) English Reading Items

다국어 초록 (Multilingual Abstract)

This study explored how ChatGPT 5.0–generated feedback was interpreted and used by Korean high school students during CSAT-style English reading problem-solving, with a focus on feedback literacy (Carless & Boud, 2018). Participants were 10 high school students(Grade 11) and 6 in-service English teachers from a general high school in Jeju. Students completed 15 CSAT reading items under a time limit, received correctness-only information, and then interacted with ChatGPT to revise incorrect answers using a researcher-designed prompt that prohibited direct answer provision. During the revision process, students verbalized their cognitive processes using the think-aloud technique, followed by semi-structured interviews on their learning experiences. Teachers were also interviewed regarding their experiences with generative AI and their perceptions of its use in English instruction, based on samples of ChatGPT-generated feedback and their own teaching practices.
Findings indicated that the effectiveness of ChatGPT feedback varied by learner achievement level and item type. Mid- and high-achieving students generally understood the feedback and used it selectively to monitor reasoning, reorganize discourse-level understanding, and transfer strategies across tasks. In contrast, lower-achieving students often struggled with lengthy text-based feedback, relied heavily on translation- and vocabulary-focused requests, and demonstrated limited strategic uptake. Across item types, answer revision gains were higher for form-focused (grammar and vocabulary) and discourse-focused (sentence insertion and ordering) items than for meaning-focused inference items, which more frequently induced cognitive load and partial understanding. Teachers valued the speed and structural support provided by AI-generated feedback but cautioned against uncritical acceptance and overreliance, emphasizing the continued need for teacher mediation and explicit feedback literacy instruction.
The study suggests that generative AI feedback can function as an effective supplementary scaffold for CSAT reading; however, its pedagogical value is conditional on learners’ feedback literacy and task demands. Implications include designing adaptive feedback routines according to proficiency level and item type, strengthening AI and feedback literacy instruction, and positioning teachers as orchestrators who regulate feedback quality, timing, and instructional goals.

목차 (Table of Contents)

• Ⅰ. 서론 1
• Ⅱ. 이론적 배경 3
• 2.1 대학수학능력시험과 영어 교육 3
• 2.2 교육 환경 변화와 피드백 중요성의 강화 5
• 2.3 영어 읽기의 언어 능력 구성 요소와 인지 처리 과정 7
• 2.4 AI 기반 영어 교육 선행연구 고찰 및 본 연구의 필요성 9
• Ⅲ. 연구 방법 15
• 3.1 연구 대상 및 자료 수집 15
• 3.2 분석 절차 및 신뢰도 확보 21
• Ⅳ. 결과 및 논의 23
• 4.1 학습자의 성취 수준에 ChatGPT 피드백의 효과성 차이 24
• 4.1.1 피드백 이해(appreciating feedback) 차원 25
• 4.1.2 피드백 판단(making judgment) 차원 28
• 4.1.3 정서 조절(managing affect) 차원 31
• 4.1.4 행동 변화(taking action) 차원 34
• 4.2 문항 유형에 따른 ChatGPT 피드백의 효과성 차이 38
• 4.2.1 의미 중심 문항에 관한 ChatGPT 피드백의 효과성 41
• 4.2.2 담화 중심 문항에 관한 ChatGPT 피드백의 효과성 44
• 4.2.3 언어 형태 중심 문항에 관한 ChatGPT 피드백의 효과성 46
• 4.3 ChatGPT 피드백의 효과성에 대한 현직 교사의 인식 47
• 4.3.1 피드백 이해 차원에서의 교사 인식 48
• 4.3.2 피드백 판단 차원에서의 교사 인식 49
• 4.3.3 정서 조절 차원에서의 교사 인식 51
• 4.3.4 행동 변화 차원에서의 교사 인식 52
• Ⅴ. 결론 55
• 5.1 주요 결과 요약 55
• 5.2 연구의 시사점 58
• 5.3 연구의 한계 60
• 5.4 향후 연구 과제 61
• 참고문헌 64
• 부록 69
• ABSTRACT 129