생성형 AI(Generative AI) 활용 학습 효과 메타분석

초록

This study conducted a systematic literature review and meta-analysis to comprehensively examine the effectiveness of learning using generative AI in South Korea and provide recommendations for its application in education. Following the PRISMA meta-analysis guidelines, 24 studies were selected for analysis. These were categorized into pre/post, control/treatment, and correlational studies, and coded for effect size calculation. Results showed that in pre/post studies, the random effects model yielded an effect size of 0.7646, indicating a medium to large effect. For control/treatment studies, a fixed effects model revealed a medium effect size of 0.4262. Correlational studies showed a medium effect size of 0.5417 in the random effects model. In pre/post design studies, additional effect sizes were examined for moderating variables such as generative AI type and learner type. Results indicated that complex generation (including images) showed greater learning effectiveness (0.8627) compared to text-based generation (0.5859). Regarding learner type, adults (1.0265) showed greater learning effectiveness than school-age learners (0.6400). This study systematically verified the effectiveness of learning using generative AI in South Korea, providing evidence-based practices for future educational applications of generative AI and offering foundational data and implications for setting directions in AI-enhanced education.

키워드