Learning, and potentially thought itself, is an inherently social process, whether directly from other humans, such as teachers, parents, or mentors, or indirectly from the artifacts other humans create. However, the social nature of the learning process doesn't come without its social learning, as opposed to cognitive learning, challenges. Sometimes we disagree with, offend, or otherwise harm one another in the learning process, or simply don't know one another enough to engage with and understand each other. How does social development, and specifically the development of dyadic teacher-student relationships impact individuals' learning processes? Here, I apply a multivariate time-series approach to understand how teacher-student dyads, randomly assigned to partners they know or have never met, differ in their nonverbal communication behavior, and how these differences impact student learning processes.
Through a custom-built online portal, open-source computer vision software, and a newly-derived state-of-the-art multivariate time series analysis, I show how teacher-student dyads from an undergraduate institution benefit from familiarity, nonverbal coordination, and their development, and how this development improves students' scientific reasoning performance. I also show how the degree of nonverbal coordination that enables high performance in the reasoning tasks develops over as little as 10—15 minutes of dedicated face-to-face interaction. Three implications of the work are highlighted. First, the results imply that social interaction processes are crucial to individual reasoning in face-to-face online contexts. Second, a potentially necessary route to improving STEM education at the undergraduate level may be more dedicated face-to-face time between students and their instructors.
Finally, the step-by-step guide provided by the work to apply multivariate techniques to non-stationary diachronic processes illuminates the value of combining evolutionary correspondence analysis with locally stationary vector autoregression. The combination of methods reduces the complexity of high-dimensional datasets to explanatory latent factors, and then quantifies the linear predictability of each original dimension on all of the others within each explanatory latent factor. In the current analysis, I identify familiarity and affect-attention tradeoff effects as the two most explanatory latent factors, and quantify how both familiarity, and the tradeoff between affective and attentive signalling between the dyads evolves over the course of 20-25 minute teacher-student interactions. Thus, beyond the implications for dyadic reasoning and STEM learning processes, the methodological implications could be applied to any high-dimensional diachronic processes, such as two bodies, or brains, interacting in other teacher-student contexts, as well as parent-child, therapist-client, and manager-employee environments in order to simplify the complexity of social interactions and uncover their impacts on individual change processes.
| Identifer | oai:union.ndltd.org:columbia.edu/oai:academiccommons.columbia.edu:10.7916/tqdr-3731 |
| Date | January 2024 |
| Creators | Friedman, Joshua |
| Source Sets | Columbia University |
| Language | English |
| Detected Language | English |
| Type | Theses |
Page generated in 0.0023 seconds