Local LLM Zero-Shot Analysis of Homelessness Discourse on Reddit

Jonathan A. Karr Jr., Ben Herbst, Matthew Hauenstein, Georgina Curto, Nitesh V Chawla

Keywords: large language models, natural language processing, people experiencing homelessness

TL;DR: Our research tackles the issue of biases in online discussions about homelessness by using LLMs to analyze Reddit data from 10 U.S. cities.

Abstract: Homelessness is a persistent issue, impacting millions worldwide, and over 770,000 people experienced homelessness in the U.S. in 2024. Social stigmatization is a significant barrier to alleviation, shifting public perception, and influencing policy. Online discourse on platforms such as Reddit shape public opinion. To address this, the project leverages natural language processing and large language models (LLMs) to mitigate bias against people experiencing homelessness (PEH) in online spaces. The goal is to promote awareness, reduce harmful biases, inform policy, and improve the fairness of generative AI. We gather Reddit data for 10 U.S. cities, then perform zero-shot classification, and finally, mitigation using Llama 3.2 Instruct and Qwen 2.5 7B Instruct models. The initial results highlighted the differing classifications between models and indicated that many mitigated outputs remained biased. This suggests the need for potential model refinement for the mitigation of text related to PEH.

Archival Status: Non‑archival

Paper Length: Long Paper (up to 8 pages of content)

Submission Number: 328