City as Text (2026): Longitudinal Sentiment Analysis

The "City as Text" project is a creative-practice based initiative exploring cities and artificial intelligence models as sites of engagement between human experience and constructed systems. This year’s iteration introduces a sentiment analysis methodology applied to poetic stanzas generated by Claude.AI in response to longitudinal urban photography. The photographs, captured across three cities over a three-year period—including concurrent and recurring photographing of sites—form the basis of the prompts for the AI’s generative output. The primary objective of this presentation is to share this novel approach and some of its output and analysis, specifically highlighting significant variances in the sentiment data associated with Tampa, Florida, in comparison to other urban datasets. By presenting these observed differences, the study probes how algorithmic models "read" and emotionally map distinct urban environments. The session welcomes critical discussion regarding methodological limitations and refinements, while offering open access to the current data set. Finally, it extends an invitation for collaborative expansion in 2026, seeking to integrate participant engagement and new city photography. This research aims to foster dialogue on the evolving relationship between human observers, AI agents, and cities as text.