Physiological response data in education: could this be the future for analyzing social interactions and learning? A systematic literature review

Abstract

This thesis systematically reviews the literature which integrated physiological and other biodata to study human social interactions in the fields of education, physiology, neurophysiology, affective neuroscience, and psychology over the past decade (2011-2021). The aim was to identify the benefits and drawbacks of their use of physiological data and inform future directions in education. This systematic literature review pre-defined data collection methodology, which involved keyword selection, database selection and searching, title and abstract screening and sorting, methods section appraisals, and full-text screening. Studies were categorized to address distinct research questions. Qualitative-only studies were separated into a pool to address research questions 2 and 2a from the qualitative perspective. The remaining quantitative and mixed methods studies were then segregated based on the field of research: in or out of the education field to address research questions 1-1c and 2-2a respectively. In education, studies most often measured electrodermal activity (EDA) (65%) and eye movements (27%), especially interested in the synchrony of bio-physiological signals (85%), in-person (65%) with groups of two to four students (85%). The analysis revealed the plausible utility of biodata for research in education involving social interactions, particularly in learning analytics research. In combination with other data, biodata can measure prior knowledge and individual and group level emotional, cognitive, and relational components of collaborative learning. Additionally, biodata can indicate learning gains, collaboration quality, task performance, and cognitive challenge, though in a context- and time-dependent manner. Multidimensional recurrence quantification analysis, matrix analysis, and minimum width envelope were identified as promising data analysis techniques to gain insights about interpersonal cognitive, attentional, metacognitive, social, and emotional process dynamics from time-series bio-physiological data of multiple subjects. Further, the visualization of eye-tracking data was identified as a useful tool for intervention in learning as well as for qualitative content analysis. The analysis found that current methodologies in education suffer from paradigmatic ambiguities that specifically arise from experimental design, data sources, data handling, and data analysis. A 2 x 2 confusion matrix revealed methodologically based ambiguities in the reviewed literature, namely the weak ability of several studies to address true negative, false positive, and false negative results. Multimodal biodata, particularly for triangulation, can address limitations imposed by the sources of data. Standardization of protocol for signal selection, thresholding, and data processing were recommended. As well, standardization of statistical test usage would help reduce current bias of diverse approaches. Suggestions are made to clarify the tacit features of experimental paradigms such that study replicability, comparability, and hence value increases. Practically, a ramping up of interdisciplinary efforts is recommended to tackle the challenges of multimodal biodata handling and analysis. This research concludes that the inclusion of physiological and other biodata in the education field offers greater potential with these adjustments.