Revisiting the small-world property of co-enrollment networks: A network analysis of hybrid course delivery strategies

Abstract

Infectious disease outbreaks bring new challenges to campus operations at institutions of higher education (IHEs). At the beginning of the COVID-19 pandemic, as a precaution against COVID- 19 transmission on their campuses, many IHEs shifted to entirely remote courses in the Spring 2020 semester. As these institutions prepared for the Fall 2020 semester, their administrations evaluated the extent to which remote courses should continue. While remote courses may reduce the number of students who interact on campus and thus mitigate the risk of disease spread, there may be academic and financial downsides. MethodologyWith de-identified enrollment, course schedule, and student profile data from the Fall 2019 semester at Georgia Institute of Technology (Georgia Tech), we construct a co-enrollment network, which is a representation of student connections through courses. In this co-enrollment network, we observe the small-world property in which the average path length among reachable student pairs is short and nearly all students are connected to each other in the main component of the network through multiple independent paths. We also see high connectivity between different majors. Then, we apply various hybrid instructional mode strategies to this co-enrollment network and use network connectivity and in-person instruction metrics to compare these strategies from health and academic perspectives, by their impacts on various groups of students, and based on the trade-off between reducing network connectivity and delivering in-person instruction. ResultsAlthough these strategies result in reduced network connectivity, they do not remove the small-world property. When comparing the strategies in terms of reducing network connectivity, we find three different strategy types to be the most favorable. Even under these strategies, the percentage of students with in-person courses that could be exposed rises quickly with initial infection rates. When comparing the strategies in terms of delivering in-person instruction, the most favorable strategies select a fixed percentage of courses for each subject to be delivered remotely. We then consider the impacts of these strategies on various groups of students to evaluate if there may be disparate impacts across these groups. In general, we find that strategies with more remote courses result in less disparate impacts on network connectivity and in-person instruction across these groups. In addition, we generally see less in-person instruction for lower-level students (i.e., first-year students and sophomores) compared to upper-level students. Regarding the trade-off between reducing network connectivity and delivering in-person instruction, the only strategy that is dominated under all scenarios in this study is the 1000s and 4000s only strategy, which keeps only the 1000- and 4000-level (i.e., first-year and senior level, respectively) courses in-person. The reason for this is that even though the university offers less in-person instruction due to this strategy's large quantity of remote courses, this strategy results in more connected networks because the 1000- and 4000-level courses tend to be large in size, the 1000-level courses are highly central, and the first-year students and seniors have high connectivity. ImportanceBy quantitatively comparing network connectivity and in-person instruction for these various strategies, campus administrators can weigh the physical health, academic performance, and mental health of students, along with the financial well-being of their academic institutions when responding to infectious disease outbreaks.