Generative AI-based Framework for Modeling Longitudinal Travel Behavior Adaptation Under Transportation Interventions

Term Start:

June 1, 2026

Term End:

May 31, 2027

Budget:

$250,000

Keywords:

Artificial Intelligence, Simulation, Travel Behavior

Thrust Area(s):

Data Collection Mechanisms, Data Modeling and Analytic Tools, Understanding User Needs

University Lead:

Georgia Institute of Technology

Researcher(s):

Srinivas Peeta

Transportation agencies deploy a wide range of interventions to influence travel behavior and manage demand. While a broad range of approaches have been used to model travel behavior dynamics, modeling how behavioral responses manifest, strengthen, or decay over repeated exposure to transportation interventions remains relatively underexplored. This project will address this gap by developing a novel framework that integrates a Generative Artificial Intelligence (AI)-powered behavioral simulation engine with a longitudinal stated preference (SP) study to model, calibrate, and validate the temporal evolution of travel behavior in response to transportation interventions. Leveraging large language models’ capability to reason through open-ended behavioral scenarios, the engine will construct heterogeneous traveler agents with persona profiles that vary across sociodemographic characteristics, mode preference, trip purpose, risk tolerance, and sensitivity to intervention type. These agents will simulate the full range of behavioral adaptation over time, capturing dynamics such as habit formation, inertia, and resistance to change. A longitudinal stated preference study, deploying scenario-based stimuli via Qualtrics to a panel recruited through Prolific, will provide the empirical basis for calibrating, fine-tuning where appropriate, and validating the engine against observed traveler responses. Safety-focused interventions will serve as the instantiation domain. The resulting framework will be intervention-agnostic and transferable, providing transportation planners an evidence-based tool for evaluating long-term behavioral implications of interventions prior to their deployment in the real world.

Scroll to Top