UW Data Science Seminar: Will von Geldernon

Описание: Title: “Using Retrieval-Augmented Generation to Analyze Evictions”

Abstract: I will present preliminary results from a project that uses computer vision and natural language processing to document tenant responses to eviction summonses and connect tenants’ response patterns to subsequent case outcomes. As a part of the eScience Institute’s Data Science and AI Accelerator, I worked with eScience Data Scientist Curtis Atkisson to measure tenant behavior and case outcomes using text extracted from ~195,000 pdf documents from ~8,500 eviction cases in Pierce County, WA filed between 2022 and 2024. Using retrieval-augmented generation (RAG) including LLM analysis, we documented the links between tenants’ submission of written responses, attendance at show cause hearings, and access to legal assistance from Washington’s right to counsel (RTC) program. This novel data also allows us to examine the relationship between legal representation and case outcomes. Results show that legal representation is provided to less than 50% of tenant households facing an eviction despite the broad eligibility criteria in Washington’s RTC program. Our findings also highlight the importance of several critical “administrative checkpoints” during eviction cases and suggest possible reforms that could increase tenants’ access to legal representation.

Biography: Will von Geldern (he/him) is a PhD candidate at the UW Evans School of Public Policy & Governance. He uses mixed methods to examine the effects of public policies and legal systems on the health and wellbeing of marginalized communities. His dissertation uses qualitative analysis, data science, and experimental methods to study the barriers that prevent tenants from accessing legal assistance during evictions.

#publicpolicy #datascience #eviction #research #retrievalaugmentedgeneration