Racial Steering by Large Language Models: A Prospective Audit of GPT-4 on Housing Recommendations
Name
3689904.3694709.pdf
Size
5.41 MB
Format
Adobe PDF
Checksum (MD5)
40d0a30e2783019139ec07f28fa9c09e
Author(s)
Liu, Eric
So, Wonyoung
Hosoi, Peko
D'Ignazio, Catherine
Date Issued
October 29, 2024
Publisher
ACM|Equity and Access in Algorithms, Mechanisms, and Optimization
Citation
Liu, Eric, So, Wonyoung, Hosoi, Peko and D'Ignazio, Catherine. 2024. "Racial Steering by Large Language Models: A Prospective Audit of GPT-4 on Housing Recommendations."
Version
Final published version
Abstract
The integration of Large Language Models (LLMs) into a wide range of rental and real estate platforms could exacerbate historical inequalities in housing, particularly given that LLMs have exhibited gender, racial, ethnic, nationality, and language-based biases in other contexts. Examples of use cases already exist, with real estate listing platforms having launched ChatGPT plugins in 2023. In response to the critical need to assess the ways that LLMs may contribute to housing discrimination, we analyze GPT-4 housing recommendations in response to N = 168,000 prompts for renting and buying in the ten largest majority-minority cities in the US with prompts varying by demographic characteristics like sexuality, race, gender, family status, and source of income, many of which are protected under federal, state, and local fair housing laws. We find evidence of racial steering, default whiteness, and steering of minority homeseekers toward neighborhoods with lower opportunity indices in GPT-4’s housing recommendations to prospective buyers or renters, all of which could have the effect of exacerbating segregation in already segregated cities. Finally, we discuss potential legal implications on how LLMs could be liable under fair housing laws and end with policy recommendations regarding the importance of auditing, understanding, and mitigating risks from AI systems before they are put to use.
Description
EAAMO ’24, October 29–31, 2024, San Luis Potosi, Mexico
MIT Department
Massachusetts Institute of Technology. Department of Electrical Engineering and Computer Science
Massachusetts Institute of Technology. Department of Urban Studies and Planning
Massachusetts Institute of Technology. Department of Mechanical Engineering
Terms of Use
Creative Commons Attribution
Persistent DSpace Link
DOI of Published Version
https://doi.org/10.1145/3689904.3694709
