Why do similar homes in different Chicago neighborhoods sell for vastly different prices? Beyond property characteristics, a neighborhood's reputation, or the informal "gossip" and social narratives circulating online, shape its economic value. This work-in-progress explores whether AI systems can help us quantify these intangible "vibes" by prompting 3 LLMs: Claude, ChatGPT, and Gemini to describe how Chicago neighborhoods are perceived on platforms like Reddit (much in the way a prospective home seeker may search online for information about neighborhoods). By comparing these digital narratives with actual neighborhood premiums derived from a traditional economic hedonic model, we hope to reveal how collective perceptions and historical stigmas become embedded in both algorithmic knowledge and market valuations. This talk presents preliminary findings and invites discussion about the promise and limitations of using large language models to explore the real-world economic consequences of the stories we tell about places.
Registration is required for this event. Lunch will be provided.
Speaker: Elizabeth Delmelle is an Associate Professor in the Department of City and Regional Planning at the University of Pennsylvania’s Weitzman School of Design, where she also directs the Master of Urban Spatial Analytics Program.
Discussant: Chris Herbert is the Managing Director of the Harvard Joint Center for Housing Studies.
This event is co-sponsored by the Center for Geographic Analysis.
As our planet becomes increasingly urban, this series seeks to expand our understanding of cities and urbanization across sites and scales. The Urban Conversations aim in particular to link humanistic approaches with spatial investigations. We host public talks, and provide a venue for researchers to share works-in-progress with an interdisciplinary community, in a conversational format. Urban Conversations is chaired by Bruno Carvalho and Daniel Agbiboa.
LOCATION:Plimpton Room (Barker Center 133) STATUS:CONFIRMED DTSTART:20260226T170000Z DTEND:20260226T180000Z END:VEVENT END:VCALENDAR