III: Small: RUI: A Fairness Auditing Framework for Predictive Mobility Models

University Of Washington, Seattle WA

Investigators

Abstract

Recent years have seen an increase in the use of location data collected from devices with GPS as well as location-based social media. This type of location data can be used for various decision-making purposes in the context of urban computing and city planning. For instance, in the context of traffic management and crowd flow, location data has been shown to provide immense opportunities for understanding and predicting visitation and congestion patterns, thus helping managers to plan resources accordingly. For a long time, privacy concerns about location data have received attention from the research community. Decades of research have been examining how to strip sensitive information from location data to block the re-identification of individuals. The success of these efforts has led to new opportunities for integrating predictive and generative models that are based on historical location data into decision-making tasks. However, a critical concern that arises is the extent to which such models and analyses are representative of everyone, fair, and equitable. Ultimately, the goal of this research is to ensure that such models and underlying data are both private and fair. This project aims to define a set of methods and approaches for auditing location predictive and generative models in terms of fairness from the perspectives of individual and collective level (i.e., crowd flow) location data. At its core, this project will advocate a novel framework for auditing the fairness of mobility traces and models in both centralized and distributed systems by offering a set of domain-specific fairness metrics related to mobility. The technical aim of this project is in two research thrusts. The first thrust focuses on building infrastructure, knowledge, and methods for the creation of spatial-temporal data through fair-aware generative AI models that are inclusive and can lead to fair and equitable policy planning. The second thrust focuses on building infrastructure, knowledge, and methods for increasing and evaluating the fairness of Location Privacy Preserving Methods (LPPMs) by offering a set of novel fair-aware algorithms that will satisfy the objectives of prediction accuracy, privacy, and fairness. This award reflects NSF's statutory mission and has been deemed worthy of support through evaluation using the Foundation's intellectual merit and broader impacts review criteria.