2024 ASEE Annual Conference & Exposition

This paper discusses the developments during Year 2 for a project concerned with analyzing the curricula of engineering programs in the United States to understand the structural barriers embedded in degree requirements that could push out diverse groups of students. We are using an emerging method for quantifying the complexity of these programs called Curricular Analytics. This method involves treating the prerequisite relationships between courses as a network and applying graph theoretic measures to calculate a curriculum’s structure complexity. In Year 1, we collected 497 plans of study representing five engineering disciplines (i.e., Mechanical, Civil, Electrical, Chemical, and Industrial) across 13 institutions - spanning a decade. To ensure the dataset is as useful as possible to engineering education researchers, we have intentionally aligned our data collection with institutions available in the Multiple Institution Database for Investigating Engineering Longitudinal Development (MIDFIELD).

One of the outputs of this project is an R package that will enable researchers and practitioners to explore and leverage the dataset in their work by enabling the calculations to be completed at scale. With the efforts in Year 1, the package has the required functionality to compute the necessary metrics for Curricular Analytics. During Year 2, we have been building functions to manipulate course-taking trajectories of actual student data such that they can be compared to one another using association analysis. Association analysis will enable us to mine common course-taking patterns disaggregated by strata like institution, discipline, first-generation-status, and transfer-status and reconstruct them as networks to complement the plan of study data. Moreover, after sharing this work in preliminary forms with faculty, there was a desire for more customized functions. Thus, we are currently conducting a systematic literature review of how Curricular Analytics has been applied and extended to search for usable metrics to add to our package.

Much of Year 2 has been spent verifying the data and correcting errors that would impact the results of any analysis, whether quantitative or qualitative, by exploring the dataset using a combination of descriptive statistics and visualizations like histograms, boxplots, and longitudinal plots. As the data currently exists, the mean structural complexity of all engineering programs we considered (n = 497) is 313, and the median is 294. Chemical engineering has the highest mean structural complexity of 430, followed by mechanical engineering with a structural complexity of 369. The remaining disciplines were more tightly clustered together: electrical with 287, industrial with 248, and civil with 232. Although we are finalizing corrections to these data, it is not expected that the results will change significantly. We are currently sampling cases at the distribution's tails in the box plots of structural complexity to explore the extreme cases in our dataset and jumpstart analyses regarding curricular design patterns.

This paper will provide details on the preliminary analyses we have conducted using Curricular Analytics, an introduction to the R package, and updates from our systematic literature review.