Project leads: Afra Mashhadi, Assistant Professor, Computer Software and Systems, University of Washington, Bothell, and Ekin Ugurel, Ph.D. Candidate, College of Engineering, University of Washington.

Data scientist: Bernease Herman, eScience Institute, University of Washington.

DSSG fellows: Apoorva Sheera, Jiaqi He, Manurag Khullar, Sakshi Charvan.

Click here for participant bios.

Generative crowd-flow (CF) models are machine learning models that are capable of producing mobility flow resembling population movement in a city. Such models traditionally come from physics models that were fundamentally based on association of distance, population and propensity of travels. Recently, new crowd flow generation models based on neural networks have emerged. These models can incorporate additional information from the city (such as amenities and reviews) and have been shown to outperform older models. However, to date there is little evidence as to whether any of these new models can create synthetic data that results in equitable flow prediction for all parts of society or whether they exacerbate social biases by widening the representation gap. These concerns are pressing given that government agencies often consider crowd-flow models in planning for public safety, traffic management, and other critical areas of operation. 

Fairness is increasingly recognized as a critical component of machine learning (ML) systems, but little attention has been given to the applications of city planning and urban related research that directly relies on spatial temporal data. We argue that understanding fairness in practice relies on observing a model’s behavior in the context that is intended to be used.

In this DSSG project, we will develop definitions of group fairness of GenAI-CF models. Such definitions could help us to measure the equity of CF synthetic datasets. By incorporating a more equitable representation of under-served groups’ travel demands, our project aims to ensure that future transportation policies, infrastructure investments, and service improvements are informed by a comprehensive understanding of all community needs.

Learn more on the project website. Watch the presentation here, or the full recording of all 2024 team presentations here.