Feeling a sense of belonging is fundamental to students’ academic success, well-being, and long-term retention. While sense of belonging is inherently relational, it has up to now rarely been studied from a network perspective. This study evaluates how students’ social networks and sense of belonging relate to their academic performance, in the context of a large undergraduate statistics course that offers recitation sessions ( [Formula: see text]). We use a network autocorrelation model, accounting for grade associations between peers connected in friendship and advice networks. Despite our observation of positive network autocorrelation on course final grades in both networks, the model did not discern any network effects net of individual-level characteristics. We find that students’ sense of belonging is positively associated with their grades, and that students who have no classmates to ask for course-related advice score lower than those who do. We posit that the positive observed association between recitation attendance and grades is likely driven by student motivation: students who attend more recitations than required score on average considerably higher than those who do not. This study reiterates the importance of students’ peer support and sense of belonging, and addresses implications for course instructors and beyond the classroom.
We introduce a novel framework for teaching data science ethics that places students in the "driver’s seat" of ethical decision-making with data. This approach is designed to make ethics instruction more accessible to statistics instructors without a background in ethics. To address common barriers, including limitations of traditional methods, the breadth of content, and lack of practical guidance, we propose a flexible course framework centered on a core skill we call algorithmic translation: the ability to act as an informed intermediary between algorithmic decision-making processes and the people affected by them. This framework can be implemented as a full course or as standalone modules within statistics classes. We present a course implementation taught as an upper-level elective in a statistics program at a large public research university, in which students learn to incorporate and evaluate fairness, explainability, interpretability, and privacy using emerging algorithmic techniques. These techniques are implemented in Python using AIF360 and AIX360 and applied to real-world datasets. We also provide a curated set of techniques and guidance showing how they can be introduced as extensions of topics already familiar to statistics students, along with open-source course materials.
