Mapping students' attitudes and identities without imposing a priori demographic categories: A quantitative study using topological data analysis

We describe our use of an innovative analytic technique adapted for use with student educational data for the first time: topological data analysis via the Mapper algorithm. In this approach, students are clustered based on their similarity and a filter which partitions them into subsamples for iterative clustering. This results in a two-dimensional representation of a higher space showing both how students cluster together and how those clusters are related to each other. By using a variety of attitudinal constructs to inform the process, we limit the a priori categorization of students and can instead group students independently of demographic information such as race/ethnicity, gender, sexuality, and disability status, and produce new ways of describing students based on these emergent groupings. We outline the requirements for using Mapper, describe the steps to implement it, and demonstrate an example mapping of student data. Using a clustering algorithm like Mapper opens new avenues for discussion and interpretation of student attributes in both current and future work on student identity, how students navigate STEM programs, and normative cultures in STEM education by creating new emergent categories and describing how those categories are related to each other.