Abstract:Researchers are constantly leveraging new forms of data with the goal of understanding how people perceive the built environment and build the collective place identity of cities. Latest advancements in generative artificial intelligence (AI) models have enabled the production of realistic representations learned from vast amounts of data. In this study, we aim to test the potential of generative AI as the source of textual and visual information in capturing the place identity of cities assessed by filtered descriptions and images. We asked questions on the place identity of a set of 31 global cities to two generative AI models, ChatGPT and DALL-E2. Since generative AI has raised ethical concerns regarding its trustworthiness, we performed cross-validation to examine whether the results show similar patterns to real urban settings. In particular, we compared the outputs with Wikipedia data for text and images searched from Google for image. Our results indicate that generative AI models have the potential to capture the collective image of cities that can make them distinguishable. This study is among the first attempts to explore the capabilities of generative AI in understanding human perceptions of the built environment. It contributes to urban design literature by discussing future research opportunities and potential limitations.

What problem does this paper attempt to address?

The core problem that this paper attempts to solve is to evaluate the potential and reliability of generative artificial intelligence (Generative AI) in capturing and representing urban place identity. Specifically, researchers hope to understand the collective place identities of different cities through the text and image information generated by generative AI models (such as ChatGPT and DALL·E2), and verify whether these generated contents can truly reflect the urban characteristics in reality. ### Main problems: 1. **Can generative AI recognize the place identity of a city?** - Researchers give a series of names of global cities, let ChatGPT generate place identity descriptions of these cities, and let DALL·E2 generate corresponding streetscape images. 2. **To what extent can the generated content be trusted?** - In order to evaluate the reliability of the generated content, researchers compared the generated text with the data on Wikipedia and compared the generated images with the real images obtained from Google search. ### Solutions: - **Text data set**: Use ChatGPT to generate place identity descriptions for each city according to prompts, and limit it to ten points to ensure the consistency and comparability of the output. - **Image data set**: Use DALL·E2 to generate streetscape images for each city, generating 10 256x256 - pixel pictures for each city. - **Verification methods**: - For text similarity, use the BERT model to calculate the cosine similarity between the text generated by ChatGPT and the Wikipedia text, and make a visual comparison through word clouds. - For image similarity, use the Learned Perceptual Image Patch Similarity (LPIPS) index to evaluate the perceptual similarity between the images generated by DALL·E2 and the Google search results. ### Results: - **Text similarity**: The text generated by ChatGPT is highly similar to the Wikipedia description in some aspects, but there are also differences, especially when describing complex and subtle local features. - **Image similarity**: The images generated by DALL·E2 show high perceptual similarity in some cities, but in other cities they are more general and fail to fully reflect specific urban characteristics. ### Conclusions: Although generative AI shows certain potential in capturing and representing the place identity of cities, there are still limitations. For example, some of the generated images are too general to fully reflect the identity characteristics of specific cities. In addition, there is also uncertainty in the evaluation results of image similarity. Future research can improve the reliability and accuracy of the generated content by improving prompt design and introducing more diverse verification data sets. Through this research, the author hopes to provide new tools and methods for urban planners and designers to understand and analyze the place identity of cities in a more efficient and low - cost way, thereby promoting sustainable urban construction and brand promotion.

Understanding Place Identity with Generative AI

Urban Visual Appeal According to ChatGPT: Contrasting AI and Human Insights

Generative AI and the History of Architecture

Testing the Capability of AI Art Tools for Urban Design

Re.Dis.Cover Place with Generative AI: Exploring the Experience and Design of City Wandering with Image-to-Image AI

Urban Visual Intelligence: Studying Cities with Artificial Intelligence and Street-Level Imagery

Towards Automated Urban Planning: When Generative and ChatGPT-like AI Meets Urban Planning

Urban Visual Intelligence: Studying Cities with AI and Street-level Imagery

Generative AI and Building Design: Early Photorealistic Render Visualization of Façades using Local Identity-trained Models

Leveraging Generative AI Models in Urban Science

Leveraging Generative AI for Urban Digital Twins: A Scoping Review on the Autonomous Generation of Urban Data, Scenarios, Designs, and 3D City Models for Smart City Advancement

Generative agents in the streets: Exploring the use of Large Language Models (LLMs) in collecting urban perceptions

Generative AI May Prefer to Present National-level Characteristics of Cities Based on Stereotypical Geographic Impressions at the Continental Level

Discovering Place-Informative Scenes and Objects Using Social Media Photos

Mapping sense of place as a measurable urban identity: Using street view images and machine learning to identify building façade materials

GANs for Urban Design

The Ethics of AI-Generated Maps: A Study of DALLE 2 and Implications for Cartography

Measuring Human Perceptions of a Large-Scale Urban Region Using Machine Learning

Generative AI in the Wild: Prospects, Challenges, and Strategies

Urban-GAN: An artificial intelligence-aided computation system for plural urban design

Urban Generative Intelligence (UGI): A Foundational Platform for Agents in Embodied City Environment