소크라테스식 질문법을 결합한 생성형 AI 기반 시뮬레이션 제작 교육 프로그램의 초등 과학 수업 적용과 학습자 변화 탐색

초록

This study implemented a generative AI-based simulation authoring instructional program incorporating Socratic questioning in regular elementary science classes in the motion and energy domain. The program was delivered to approximately 320 fourth- and fifth-grade students from 16 classes and consisted of two preliminary sessions and six instructional sessions. Data from 260 students who completed both pretests and posttests were analyzed. Students first observed physical phenomena, including balance, sound vibration, and magnetic force, and then created simulations using natural-language prompts. Teachers guided students to compare AI-generated outputs with their observations and revise their prompts to incorporate relevant scientific conditions and variables. Paired-samples t-tests showed statistically significant pretest-posttest increases in the total critical-thinking disposition score and in the reconstructed subscales labeled truth-seeking and self-confidence, as well as in science-learning emotion and science-learning motivation. However, all effect sizes (Cohen's d) were small, suggesting that the findings reflect a direction of change and the feasibility of the approach rather than substantive effects from a short-term intervention. In the analyzed prompt-revision cases, students often refined AI-generated outputs by incorporating scientific conditions and variables aligned with their observations, rather than passively accepting the generated results. These findings suggest that a sequence of activities, 'observation, simulation authoring, comparison with AI-generated outputs, prompt revision, and reflection,' may support reflective model revision and active participation in scientific inquiry in elementary science classes.

키워드