Use of Multimodal Learning Analytics for Complex Learning Environment

Esther Tan

Multimodal learning analytics (MMLA) offer new insights into the study of collaborative learning. Multimodal data affords a more effective and efficient investigation on the quality of collaboration, communication and interaction. In this study, the authors illustrated how specific features of MMLA can inform on the quality of students’ interaction in project-based learning using physical computing.
They investigated the use of diverse sensors, user-generated content and data from learning artefacts to study small group interaction in project-based learning. This enabled an exploration of the different aspects of interactions that contribute to a successful collaboration in open-ended tasks. Participants were 18 engineering students at a European university. The students were put into six groups of three students each and they had three days to complete an open-ended design task in each session (e.g., create a prototype for an interactive toy). Participants were given a mobile tablet to document their planning, building and reflection phase. Various machine learning classifiers were used to predict groups’ artefact quality based on the multimodal data (i.e., face tracking, hand tracking, Arduino IDE, and audio level). Students’ project outcome was scored on five aspects: clarity, independent thinking, plan, solution working and quality.

Their findings showed that the distance between learners’ hands and faces was a stronger predictor on the quality of students’ artefact, which in turn, informed on the quality of the collaboration. Also, the audio signal level (AUD) feature was able to predict students’ performance. The results of the study carry important implications for both researchers and practitioners. For the latter, an awareness of the MMLA features of project-based learning is instrumental for providing timely and appropriate learning support. For researchers, the decision on using neural networks and/ or traditional regression approaches to categorise MMLA data is contingent on the research questions and research context.

Article: Spikol, D., Ruffaldi, E., Dabisias, G., & Cukurova, M. (2018). Supervised machine learning in multimodal learning analytics for estimating success in project‐based learning. Journal of Computer Assisted Learning34(4), 366-377.