대학교육에서의 멀티모달 데이터 기반 학습분석 연구동향

초록

Purpose: This study aimed to systematically analyze domestic and international research trends in multimodal data-based learning analytics (MMLA) within higher education contexts, and to derive educational implications for university teaching, learning, and Teacher Education. Methods: A systematic literature review was conducted on 170 studies (118 international and 52 domestic) retrieved from Web of Science, Scopus, ERIC, and RISS. Studies were analyzed using a coding framework comprising five categories: research context, data collection methods, analytical techniques, research topics, and key research outcomes. Only empirical studies utilizing two or more data modalities were included, and Cohen’s κ = .87 confirmed the reliability of the coding process. Results: The findings reveal that multimodal learning analytics research evolved through three distinct stages: the first stage centered on LMS log-based behavioral data analysis, the second stage expanded into emotional and behavioral analysis utilizing sensor, biometric, video, and audio data, and the third stage advanced into real-time integrated analysis based on eye-tracking, wearable devices, and deep learning. Conclusion: Multimodal data-based learning analytics is driving a paradigm shift from outcome-centered to experience-centered analysis in higher education. The findings suggest that MMLA can be applied in the following concrete ways within university settings. First, instructors can detect real-time signals of decreased engagement, increased cognitive load, and emotional shifts to adjust lesson difficulty and activity structure accordingly. Second, personalized feedback informed by behavioral and affective data can strengthen learners’ self-regulated learning and motivation. Third, analysis of interaction patterns in collaborative learning environments enables early identification of participation imbalances and supports timely instructional intervention. Fourth, in the context of Teacher Education, integrating multimodal learning analytics literacy into pre-service and in-service teacher preparation programs can systematically strengthen data-informed teaching competencies and support the coherent integration of instructional design, implementation, and reflective practice.

키워드