Abstract

We study the design, evaluation, and deployment of data science methods for problems disproportionately faced by marginalized communities in the United States. We focus on needs and priorities identified by local stakeholders and consider how user involvement shapes the design and efficacy of our implementations. We demonstrate how to design, implement, and deploy a platform with local support, presenting 412Connect, a platform designed to connect university students with Pittsburgh Black-owned businesses. Working with local stakeholders within and outside of the university community, we develop and evaluate badge and recommendation mechanisms that increase student engagement while distributing attention equitably across businesses. Our platform serves as a model of how researchers can use data science to engage with and design platforms for the communities around them. The subsequent chapters address specific challenges in the deployment of data science tools to inform cities' decision-making on pressing housing problems. In one, we study door-to-door tenant outreach for eviction prevention. Using property-level eviction filing data from counties across the United States, we show that optimization-based outreach reaches substantially more at-risk properties than neighborhood-based canvassing. At the same time, we reveal how segregation and eviction disparities shape the distribution of service across marginalized and non-marginalized renters. Then, in the final chapter, we test whether involving end-users in training an AI decision-support tool changes how they later use it in a civic resource-allocation task. Grounding our experiment in the prioritization of vacant properties for demolition, we find involvement in training increases both concordance with AI recommendations and decision-making accuracy, but also leads to participants making more concordant decisions when the AI is wrong. Together, these three works provide insight into using data-driven methods to serve marginalized communities. We illustrate the importance of not only accounting for existing inequalities and disparities but also considering the stakeholders in the decision-making process.

Committee Chair

Patrick Fowler

Committee Members

Alvitta Ottley; Chien-Ju Ho; Sanmay Das; William Yeoh

Degree

Doctor of Philosophy (PhD)

Author's Department

Interdisciplinary Programs

Author's School

McKelvey School of Engineering

Document Type

Dissertation

Date of Award

8-17-2026

Language

English (en)

Download

Available for download on Monday, August 14, 2028

Share

COinS