2025 ASEE Annual Conference & Exposition

The Hidden Costs of Complexity: Using Causal Inference and Double Machine Learning to Uncover Important Relationships in Higher Education Data Sets

Presented at DSAI Technical Session 4: Workshops, Professional Development, and Training

Graduation rates are critical performance metrics for higher education institutions, reflecting student success and the effectiveness of educational programs. Among various factors, the complexity of university curricula, measured by prerequisite course sequences, total credit requirements, and course flexibility within degree programs, significantly influences outcomes such as timely graduation and retention rates. Previous studies analyzing these effects often lack a unified framework to address how factors such as gender, academic preparation, and socioeconomic background shape these relationships. Using data from 26 U.S. universities funded by the Ascendium Foundation, this study employs a multifaceted causal inference framework to evaluate the impact of curricular complexity on graduation rates. Our methodology combines Hierarchical Linear Models (HLM) to account for the nested structure of students within universities, Generalized Propensity Scores (GPS) to adjust for confounders, and Double Machine Learning (DML) within GPS-stratified quintiles to provide robust causal estimates. Furthermore, we construct and refine a causal network using the Peter-Clark (PC) Algorithm and the Bayesian Information Criterion (BIC) score, with sensitivity analysis to
ensure robustness. Our results reveal a significant negative causal relationship between curricular complexity and four-year graduation rates, with an effect of -3.88% per unit increase in complexity. Sensitivity analysis supports the robustness of this relationship, showing a consistent effect of -3.76% when accounting for unobserved confounders. Introducing a hidden node representing socioeconomic status in our causal network further strengthens these findings, showing minimal change in the estimated effect even when accounting for potential unobserved factors. These findings suggest that complex curricula create barriers to completing the degree in time, underscoring the need for educational strategies that balance academic rigor with accessibility. This study provides a foundation for evidence-based policy reforms that aim to improve student success through optimized curricular design.

Authors

Melika Akbarsharifi The University of Arizona [biography]

Melika Akbarsharifi is a PhD candidate in Software Engineering at the University of Arizona, studying under Professor Gregory L. Heileman. Her research at the Curricular Analytics Lab focuses on developing advanced computational methods and machine learning approaches to optimize educational pathways and student success.
Ahmad Slim [biography]

Dr. Slim received his B.E. and M.S degrees in Electrical and Computer Engineering from the Lebanese American University in 2008 and 2011, respectively. In 2016, he received his PhD degree in Computer Engineering from the University of New Mexico where he worked as a research assistant professor for three years after graduation. In 2020, Dr. Slim joined the Mechanical Engineering Education Research Center at Purdue University as a Postdoctoral Research Fellow. His main interests involve educational data mining, machine learning and curricular analytics. He published several papers on methodologies that support inferences in STEM education, student success and other educational practices that advance education outcomes.
Prof. Gregory L. Heileman http://orcid.org/0000-0002-5221-5682 The University of Arizona [biography]

Gregory (Greg) L. Heileman currently serves as the Associate Vice Provost for Academic Administration and Professor of Electrical and Computer Engineering at the University of Arizona, where he is responsible for facilitating collaboration across campus t
Roxana Akbarsharifi The University of Arizona [biography]

Roxana Akbarsharifi is a PhD student in Software Engineering at the University of Arizona. Her research focuses on educational analytics and developing tools to improve student outcomes and support academic success. Her research interests include software engineering, data analytics, and data visualization, with an emphasis on applying these technologies to solve educational challenges and enable data-driven decision making in higher education.
Kristina A Manasil The University of Arizona [biography]

Kristi Manasil is a second-year PhD student in the College of Information Science at the University of Arizona. She received her bachelor's degree in Computer Science from the University of Arizona. Her areas of focus are data visualization, machine learning, learning analytics and educational data mining.

Note

The full paper will be available to logged in and registered conference attendees once the conference starts on June 22, 2025, and to all visitors after the conference ends on June 25, 2025

« View session

For those interested in:

Broadening Participation in Engineering and Engineering Technology
engineering
gender
undergraduate