Author ORCID Identifier:
Date of Graduation
5-2026
Document Type
Dissertation
Degree Name
Doctor of Philosophy in Engineering (PhD)
Degree Level
Graduate
Department
Civil Engineering
Advisor/Mentor
Mitra, Suman
Committee Member
Sasidharan, Lekshmi
Second Committee Member
Rahman, Muhammad
Third Committee Member
Hernandez, Sarah
Keywords
Causal Analysis; Deep learning; E-shopping; Explainable Artificial Intelligence; Telecommuting; Unsupervised Machine Learning
Abstract
In recent years, technological innovation has continued to reshape mobility. This transformation has been driven by the widespread use of information and communication technologies (ICT), defined in this dissertation as digital tools and platforms that allow individuals to work, shop, and carry out daily activities without always relying on physical travel. The rapid expansion of telecommuting, online shopping, app-based delivery services, and other forms of digitally mediated activity has changed how people organize time, space, and travel in everyday life. Activities that were once closely tied to physical trip making are now increasingly shaped by virtual access and digital choice. As a result, transportation demand is no longer only a reflection of population growth, land use, and economic activity, but is also increasingly influenced by how people integrate tele-activities into their routines. This shift has created an important need for transportation research to better explain who adopts ICT-enabled activities, how those activities affect physical travel, and the extent to which stated intentions about ICT-enabled activities align with realized behavior. This dissertation examines these questions through three related studies that together develop a broader understanding of ICT-enabled travel behavior in the United States. Using recent national survey data and a set of advanced data-driven methods, the dissertation combines interpretability, and predictive learning to study tele-activities from adoption, to consequence, to realization. The first study examines how telecommuting adoption changed before, during, and after the pandemic using explainable machine learning methods. It shows that the factors associated with telecommuting were not stable across periods. Before the pandemic, gender and occupation were especially influential, while during and after the pandemic, income and education became more important, with some of these shifts persisting beyond the peak disruption. The second study investigates how online shopping affects travel behavior using a debiased causal machine learning framework. Rather than supporting a simple substitution narrative, the findings show that online shopping often has complementary travel effects, especially for food and goods purchases, while grocery and service-related online shopping show more mixed relationships depending on the travel outcome considered. Item returns also emerge as behaviorally important, increasing trip frequency and shopping-related vehicle miles traveled even when they do not substantially increase total travel time or overall travel distance. The third study examines the intention-action gap in e-shopping by linking earlier stated expectations to later realized behavior for the same respondents in a longitudinal survey setting. The results show that stated intentions are useful but incomplete predictors of later behavior. The degree of alignment varies across tele-shopping domains, with grocery delivery showing the strongest match, followed by food delivery, while goods delivery shows the weakest. The study further shows that predictive performance improves when stated intentions are supplemented with demographic, household, and latent attitudinal information learned through a variational autoencoder. The dissertation shows that tele-activities should not be treated as a single uniform phenomenon. Their adoption, travel effects, and behavioral reliability differ across contexts, activity types, and population groups. By integrating explainable machine learning, debiased causal inference, and latent representation learning within one dissertation, this research provides a more complete account of how digital activities interact with physical travel behavior. The findings have direct implications for transportation planning and policy, particularly in areas such as demand forecasting, infrastructure investment, sustainability, and the interpretation of survey-based behavioral data. More broadly, the dissertation contributes to a better understanding of mobility in a period where travel behavior is increasingly shaped by the evolving relationship between technology and everyday life.
Citation
Ogungbire, A. M. (2026). Information & Communication Technologies and Physical Travel: Modeling Their Relationship Using Machine Learning. Graduate Theses and Dissertations Retrieved from https://scholarworks.uark.edu/etd/6228
