AI 기반 생리심리 변인과 자기조절학습과의 관련성 탐색 연구 = An Exploratory Study on the Associations Between AI-Based Psychophysiological Variables and Self-Regulated Learning

다국어 초록 (Multilingual Abstract)

The purpose of this study was to examine the relationships between AI-derived psychophysiological indicators—emotion, stress, and attention—extracted from an AI-based learning care platform and students’ self-regulated learning. The participants were 23 high school students. Facial video data and visual behavioral signals collected during learning sessions were analyzed using an AI model to generate Valence–Arousal–based emotional indicators (joy, anger, boredom, and relaxation), stress indicators, and attention indicators derived from gaze, facial expressions, and head pose. These indicators were aggregated at the individual level and constructed as process-oriented learning variables. Self-regulated learning was measured using total scores and sub-strategies (cognitive, motivational, and behavioral strategies). Descriptive statistics, correlation analysis, partial correlation analysis, and hierarchical regression analysis were conducted. The results showed that attention demonstrated a consistently significant positive relationship with self-regulated learning, whereas stress exhibited a negative relationship. In contrast, emotional indicators showed some associations at the bivariate level; however, their unique relationships became non-significant after controlling for other process variables. Meanwhile, the relationships of attention and stress with self-regulated learning were maintained and even strengthened after controlling for other variables, and similar patterns were observed across sub-strategies. Although the overall hierarchical regression model was significant, the independent predictive power of individual indicators was not clearly distinguished, suggesting the influence of shared variance among indicators and the limited sample size. These findings suggest that, among AI-derived psychophysiological indicators, attention and stress may serve as relatively stable process-based data for understanding self-regulated learning. Furthermore, this study highlights the potential of such indicators as process-oriented data that complement self-report measures, and provides a foundation for designing data-driven learning support systems, including teacher feedback timing and intervention strategies, learner self-monitoring tools, and AI-based personalized learning support systems.