Title: DATAFICATIONS OF EVICTION IN THE U.S. SOUTH

Date: Monday, April 7, 2025

Time: 8.00-11.00 ET

In-person Location: TSRB 223 (Spark Studio)

Zoom link: https://gatech.zoom.us/j/99748846789

Anh-Ton Tran

Ph.D. Candidate, Human-Centered Computing

School of Interactive Computing

College of Computing

Georgia Institute of Technology

Committee: 

Dr. Carl DiSalvo (advisor), College of Computing, Georgia Institute of Technology

Dr. Christopher A. Le Dantec, Khoury College of Computing & College of Arts, Media, and Design, Northeastern University

Dr. Amy Bruckman, College of Computing, Georgia Institute of Technology

Dr. Seumalu Elora Raymond, College of Design, Georgia Institute of Technology

Dr. Katta Spiel, HCI Group, TU Wien

Summary:

This dissertation comprises a five-year ethnography that explores the intersection of data, public services, and socio-technical harm in the context of eviction in one of the highest evicting cities in the U.S., Atlanta. I argue that public services, like the magistrate court’s provision of “access to justice,” are unique configurations of a data pipeline towards the development of data driven tools and products. This unique configuration of a data pipeline can inform how computing scholars consider data work and algorithmic fairness.

Despite being public, administrative eviction data is often not accessible or reliable, leading to issues of harm, especially when used by commercial entities like tenant screening services. Studies have shown the inaccuracy and challenges in aggregating administrative eviction data, finding that on average 22% of records are either inaccurate or ambiguous. These datasets cannot be concatenated cleanly either, because of the differences in state laws over eviction and the differences county to county in terms of how eviction is processed. However, there is little research that understands the data work to produce this data. Understanding this which has cascading effects when the data is employed into algorithms and interfaces like tenant screening services. How eviction is turned into data and how that data is ultimately used matters, since those affected by eviction data are those who have historically been marginalized in housing.

I explore how eviction is turned into data across two contexts: grassroots and institutional. I describe and detail the data practices of a housing activist non-profit in counting eviction to support tenants through this process, hold bad actors accountable, and advocate for more tenant rights. I also report on the ways in which the magistrate court system produces and generates administrative eviction data from public eviction records which are inevitably fed into algorithmic decision-making systems: tenant screening interfaces. Gaining a deep understanding of these contexts allows me to design tools and practices that can intervene into the datafication of eviction.

Emphasizing the complexities of eviction data shows how computing and HCI scholars should aim to gain a holistic understanding of an AI or algorithmic development pipeline. Gaining this understanding allows one to offer insights into addressing these issues by developing tools and practices that bridge various stakeholders involved in that data, which I demonstrate in this work. Ultimately, the dissertation seeks to identify solutions and insights that can help mitigate the harms caused by large data systems, while offering new ways for computing to engage with pressing social issues like eviction.