AI 활용 컴퓨팅 교육의 학습위기에 관한 구조관계 연구 : 생존분석을 적용한 매개효과 검증 = Structural Relationships of Learning Crisis in AI-Integrated Computing Education: Mediation Effects Through Survival Analysis

다국어 초록 (Multilingual Abstract)

This study aimed to explore the structural relationships between the timing of learning crisis onset and the factors influencing its hazard in AI-integrated computing education, and to verify the mediating role of creative convergence competency. Data were collected from 473 students enrolled in AI-integrated computing liberal arts courses during the second semester of 2025. Survey data on self-determined learning motivation, perceived stress, learning flow, and creative convergence competency were collected alongside attendance records from LMS log data and academic achievement data from the academic administration system. Learning crisis was operationalized as two consecutive or four non-consecutive absences, and direct and indirect paths to the hazard of learning crisis were estimated using a structural relationship model applying survival analysis. The findings were as follows. First, learning crisis occurred in 65 of 473 participants （13.7%）, with a mean onset at Week 7.9 （SD = 4.1） and a median at Week 8. The baseline hazard concentrated in Weeks 2–5, peaking at Week 4 （h（t） = .022） before declining, and remained low from Week 8 onward （h（t） = .005–.014）. Second, comparison between students who experienced learning crisis and those who did not revealed no significant differences at the overall scale level; however, identified regulation, a subfactor of self-determined learning motivation, showed a significant difference between the two groups （t = -2.309, p = .021）, indicating that the degree to which students internalize the value of learning—even under extrinsic motivation—functions as an early differentiation point for learning crisis. Third, self-determined learning motivation had significant positive effects on both creative convergence competency （β = .463, p 〈 .001） and learning flow （β = .548, p 〈 .001）; learning flow positively predicted creative convergence competency （β = .337, p 〈 .001）, whereas perceived stress negatively predicted learning flow （β = -.114, p 〈 .01）. Perceived stress influenced creative convergence competency only through the indirect path via learning flow, and its direct path was not significant. Fourth, the simple mediation indirect effect for the path self-determined learning motivation → creative convergence competency → learning crisis was significant （0.288, BC 95% CI [.002, .606]）, confirming that creative convergence competency functions as a process variable linking learning motivation and learning crisis. Fifth, a suppression structure was identified at an exploratory level: the direct effect of learning motivation was protective, whereas its indirect effect through creative convergence competency increased risk. Learners with higher competency may set more challenging tasks and higher self-expectations, thus perceiving a greater burden from the gap between expectation and performance. The linear assumption that higher competency lowers crisis therefore does not always hold. This study identifies Week 4 as a critical threshold by operationalizing learning crisis from LMS log data and quantifying week-by-week hazard rates, thereby providing a basis for early-warning system design. The integration of survival analysis with structural equation modeling enables the temporal risk structure and the psychological pathways underlying learning crisis to be verified simultaneously within a single model, a methodological contribution to the field. The suppression effect, undetectable without this integration, demonstrates the value of the approach. These findings suggest that instructional design and intervention strategies aimed at preventing learning crisis in mandatory AI-integrated computing courses should prioritize the internalization of motivation and the development of creative convergence competency from the early weeks of the semester. Keywords: AI-integrated computing education, learning crisis, self-determined learning motivation, perceived stress, learning flow, creative convergence competency, discrete-time survival analysis, structural equation modeling, mediation effect

목차 (Table of Contents)

• Ⅰ. 서론 1
• 1. 연구의 필요성 1
• 2. 연구의 목적 6
• 3. 연구문제 8
• 4. 용어의 정의 10
• Ⅱ. 이론적 배경 15
• 1. AI 활용 컴퓨팅 교육 15
• 2. 학습위기의 개념 21
• 3. 학습위기에 관한 영향요인 24
• 4. 생존분석을 적용한 매개효과 검증 연구 39
• Ⅲ. 연구방법 49
• 1. 연구설계 49
• 2. 측정도구와 연구모형 57
• 3. 자료수집과 분석 63
• 4. 연구의 윤리적 고려 77
• Ⅳ. 연구 결과 78
• 1. 측정모형 검증과 기술통계 78
• 2. 학습위기 발생 위험의 주차별 분포 87
• 3. 구조경로 추정 96
• 4. 매개효과 검증 99
• Ⅴ. 논의 및 결론 108
• 1. 논의 108
• 2. 결론 및 제언 117
• 참고문헌 125
• 부 록 147