Urban attractiveness according to ChatGPT: Contrasting AI and human insights

The attractiveness of urban environments significantly impacts residents' satisfaction with their living spaces and their overall mood, which in turn, affects their health and well-being. Given the resource-intensive nature of gathering evaluations on urban attractiveness through surveys or inquiries from residents, there is a constant quest for automated solutions to streamline this process and support spatial planning. In this study, we applied an off-the-shelf AI model to automate the analysis of urban attractiveness, using over 1800 Google Street View images of Helsinki, Finland. By incorporating the GPT-4 model, we assessed these images through three criteria-based prompts. Simultaneously, 24 participants, categorised into residents and non-residents, were asked to rate the images. To gain insights into the non-transparent decision-making processes of GPT-4, we employed semantic segmentation to explore how the model uses different image features. Our results demonstrated a strong alignment between GPT-4 and participant ratings, although geographic disparities were noted. Specifically, GPT-4 showed a preference for suburban areas with significant greenery, contrasting with participants who found these areas less attractive. Conversely, in the city centre and densely populated urban regions of Helsinki, GPT-4 assigned lower attractiveness scores than participant ratings. The semantic segmentation analysis revealed that GPT-4's ratings were primarily influenced by physical features like vegetation, buildings, and sidewalk. While there was general agreement between AI and human assessments across various locations, GPT-4 struggled to incorporate contextual nuances into its ratings, unlike participants, who considered both context and features of the urban environment. The study suggests that leveraging AI models like GPT-4 allows spatial planners to gather insights into the attractiveness of different areas efficiently. However, caution is necessary, while we used an off-the-shelf model, it is crucial to develop models specifically trained to understand the local context. Although AI models provide valuable insights, human perspectives are essential for a comprehensive understanding of urban attractiveness. • The study leverages GPT-4 model with 1800+ Street View images of Helsinki for comprehensive urban attractiveness analysis. • AI and human ratings of urban attractiveness show strong alignment, with notable geographic variations. • Residents' ratings show more spatial variation compared to non-residents due to personal experiences. • Results suggest AI models need local context understanding to match human evaluative nuances. • Findings emphasize hybrid AI-human approaches for effective urban planning and design decisions.