Multimodal learning analytics: A promising avenue to measure cognitive load during online complex problem solving?
Due to the Covid-19 pandemic, online learning in higher education is gaining in popularity more than ever. Since online learning is so prevalent, it is important that online courses offer learning experiences of high quality that match the skills required for the 21st century. As such, online learning environments should offer real-life complex problems to practise career adaptive competencies (Van Merriënboer & Kirschner, 2018). However, it must be borne in mind that complex problem solving can induce a great deal of cognitive load. This phenomenon can be explained by cognitive load theory (CLT) introduced by Sweller (1994).
CLT uses current knowledge about the human cognitive architecture as a baseline to design training that reduces the demands on learners’ working memory, so that they learn more effectively. CLT indicates that cognitive load (CL) can be induced by both intrinsic and extraneous load. The level of intrinsic load is determined by the task complexity, whereas extraneous load is mainly imposed by instructional procedures that induce unnecessary working memory load (Sweller, 2010; Van Merriënboer & Kirschner, 2018). When the learners’ cognitive capacities are exceeded, this might induce negative feelings such as stress and frustration, which might seriously hamper learning and subsequently undermine the effectiveness of the online courses (Morton et al., 2019; Vanneste et al., 2020). Against this background, real-life complex problems should be sufficiently challenging but still within the cognitive capacities of the learner (Larmuseau, Vanneste, Cornelis, Desmet, & Depaepe, 2019).
‘When the learners’ cognitive capacities are exceeded, [cognitive load] might induce negative feelings such as stress and frustration, which might seriously hamper learning and subsequently undermine the effectiveness of the online courses.’
In order to align the instructional design with students’ cognitive abilities, we should be able to measure cognitive load during online complex problem solving. Using these assessments of cognitive load, interactive online learning systems can be created that use artificial intelligence technologies (AI) to deliver tailored pedagogical interventions such as instructional guidance or adjustment of task difficulty.
In search of objective, real-time measures of cognitive load, a study was conducted in collaboration with Imec research groups at KU Leuven and at the University of Lille in the context of online complex problem solving in higher education. The study incorporated electro dermal activity (EDA), skin temperature, heart rate and heart rate variability measured unobtrusively by a wrist-worn wearable and chest patch developed by Imec. In order to induce changes in cognitive load this study experimentally manipulated the intrinsic load and the extraneous load based on CLT, by respectively varying the difficulty level of statistical problems and the instructional guidance.
Findings of the study highlight the potential of skin temperature and heart rate in measuring high cognitive load during online problem solving. Nonetheless, it also indicates that physiological data might not be sensitive to small changes of cognitive load (Larmuseau et al., 2019; Morton et al., 2019). Follow-up studies are therefore desirable in order to validate these findings.
This blog is based on the article ‘Multimodal learning analytics to investigate cognitive load during online problem solving’ by Charlotte Larmuseau, Jan Cornelis, Luigi Lancieri, Piet Desmet and Fien Depaepe (2020), published in a new special issue on Multimodal Learning Analytics in the British Journal of Educational Technology. It is free-to-access for a limited period, courtesy of our publisher, Wiley.
Larmuseau, C., Cornelis, J., Lancieri, L., Desmet, P., & Depaepe, F. (2020). Multimodal learning analytics to investigate cognitive load during online problem solving. British Journal of Educational Technology. Advance online publication. https://onlinelibrary.wiley.com/doi/abs/10.1111/bjet.12958
Larmuseau, C., Vanneste, P., Cornelis, J., Desmet, P., & Depaepe, F. (2019). Combining physiological data and subjective measurements to investigate cognitive load during complex learning. Frontline Learning Research, 7(2), 57–74. https://doi.org/10.14786/flr.v7i2.403
Morton, J., Vanneste, P., Larmuseau, C., Van Acker, B., Raes, A., Bombeke, K., Cornillie, F., Saldien, J., & De Marez, L. (2019). Identifying predictive EEG features for cognitive overload detection in assembly workers in Industry 4.0. In Longo, L., & Leva, C. (Eds.). H-Workload 2019: 3rd International Symposium on Human Mental Workload: Models and Applications (Works in Progress) [Conference proceeedings]. Sapienza, University of Rome, 14–15 November 2019. https://arrow.tudublin.ie/hwork19/1/
Sweller, J. (1994). Cognitive load theory, learning difficulty, and instructional design. Learning and Instruction, 4(4), 295–312. https://doi.org/10.1016/0959-4752(94)90003-5
Sweller, J. (2010). Element interactivity and intrinsic, extraneous, and germane cognitive load. Educational Psychology Review, 22, 123–138. https://doi.org/10.1007/s10648-010-9128-5
Vanneste, P., Raes, A., Morton, J., Bombeke, K., Van Acker, B., Larmuseau, C., Depaepe, F., & Van Den Noortgate. W. (2020). Towards measuring cognitive load during assembly work through multimodal physiological data. Manuscript submitted for publication.
Van Merriënboer, J. J. G., & Kirschner, P. A. (2018). Ten steps to complex learning: A systematic approach to instruction and instructional design (3rd ed.) New York & London: Routledge.