Brief, low-cost in-class activities can introduce undergraduates to urban planning and sustainable mobility, although few studies measure what one achieves. As generative AI (GenAI) enters planning practice, we compare two ways of delivering one: a conversational GenAI agent that generates imagery as it speaks, and a human facilitator working with sketches. Fifty-five undergraduates, none studying planning or architecture and most in their first two years, completed both in counterbalanced order, about three minutes each, rating four attitudes on 5-point scales before, between and after the sessions. The two modes moved different attitudes: among students who had met only their first method, support for cycling rose 0.63 of a point after the facilitator and 0.04 after the agent, and that advantage held once both sessions were counted. Willingness to participate in planning rose 0.39 after the agent and stayed flat after the facilitator, an exploratory result appearing at first contact only. A second, back-to-back session did not compound the first: willingness fell back whichever mode delivered it. Such an encounter triggers situational interest; it does not teach content. GenAI thus gives educators a low-cost way to stage that first encounter. The mode should therefore match the aim.

Urban vitality is a key quality of the livability and attractiveness of a city, with street-level social dynamics recognized as a central representation. However, existing approaches to capturing such dynamics, primarily through pedestrian volume, rely on static imagery, location-based proxies, or custom sensors. Such methods are typically discontinuous, limited in spatio-temporal resolution, costly, and overlook the role of sound in shaping environments. To address these gaps, this study explores the feasibility of using Street View Video (SVV), a novel, promising, yet under-utilized data source, for capturing street-level dynamics. Leveraging widely available crowdsourced city tour videos as a representative form of SVV, we develop an automated and scalable framework for high-precision geo-localization of SVV by extracting textual and depth information. Further, we propose a visual–auditory pipeline that integrates continuous pedestrian activity and sound environments as complementary dimensions that reveal street-level social dynamics. Using Melbourne as a case study, we identified 892 high-quality and geolocatable videos totaling over 542 h and achieved an accuracy of 77% for geo-localization. We generated spatio-temporal maps that not only reproduced well-studied pedestrian volume patterns, but also previously neglected aspects, including human actions (walking, standing, sitting), sound intensity, and sound types (speech, music, vehicle). Correlation and quadrant analyses further explored the relationships among these vitality dimensions, revealing distinctive street characteristics and highlighting the limitations of conventional road classifications. Overall, these findings demonstrate the potential of SVV as a multimodal and reproducible source of urban vitality indicators, providing a proof-of-concept for future urban sensing applications.

Urbanization often exacerbates psychological stress, partly due to a lack of psychological connection to urban spaces. While place attachment — the multidimensional bond between individuals and their surroundings — is known to benefit well-being, its specific influence on restorative perceptions at city or regional scale remains underexplored. This study introduces an innovative framework to investigate how place attachment levels affect restoration perception by analysing links between visual features, restorative qualities, and outcomes. Using street-level imagery, a survey (n = 547) was conducted in Hong Kong to gather restorative perceptions from residents with varied attachment levels. Machine learning models were then trained on this data and 25,194 street view images to map attachment variations at a regional scale. The research identifies key differences in restorative perceptions between populations with different place attachment levels, analysing survey-based variations, feature importance in machine learning, spatial disparities, and pathway relationships. The findings highlight the significant role of place attachment in contributing to restorative experiences, with high place attachment individuals showing stronger psychological recovery in urban environments. Moreover, spatial differences between high- and low-PA groups are greater in suburban areas. While high-attachment residents derive restoration from greenery, symbolic features, and high design qualities, low-attachment groups derive restoration from structural elements (e.g., roads and walls) that provide clear scope and spatial legibility. These insights offer perspectives on planning principles that foster a deeper connection to place. They propose tiered strategies tailored to the distinct needs of new developments versus historical districts, potentially supporting citizens’ long-term psychological well-being.

Multimodal large language models (MLLMs) are increasingly deployed for urban informatics applications, from objective built environment attribute extraction to subjective assessments. However, do MLLMs apply the same standards across regions while executing these urban-related tasks? Are MLLMs reliable and neutral judges? Do they adapt without criticizing, raising concerns about equitable deployment and further hidden bias? Here, we present a geographic red-teaming framework for diagnosing geographic biases in MLLMs applied to image object detection and urban visual perception. Using annotated imagery from more than 200 cities as a standardized benchmark, we demonstrate that MLLMs exhibit significant geographic disparities in both object detection accuracy and perceptual assessments across different regions. Quantitatively, perceptual scores for attributes such as Wealthy and Beautiful dropped by an average of 80% for African cities after geo-referencing, whereas regions such as Asia and North America showed score increases of 26%. A spatial-scale sensitivity test further showed that coordinates, city, country, and continent information produced broadly similar directional shifts, while combined geographic cues generated the strongest perceptual changes. Among the multiple state-of-the-art models, GPT-4o showed the lowest perceptual bias. Our geographic red-teaming stress-tests MLLM performance across underrepresented urban environments, revealing consistent patterns of bias that mirror training-data imbalances. We also provide quantitative metrics for measuring geographic equity in model outputs and establish recommendations for responsible MLLM deployment in urban research. These findings highlight critical limitations in current multimodal AI systems and demonstrate the urgent need for geographically inclusive model development to prevent the perpetuation of urban inequalities through automated analysis systems.

Human–environment interactions, a classic topic in geography, suggest that individuals and their environments might shape each other. Yet the specific mechanisms underlying these interactions regarding human personality traits have not been explored. This study examines the associations between human Big Five personality traits and built environment characteristics derived from street view imagery across four cities in Texas, United States, providing a descriptive foundation for understanding these complex human–environment dynamics. By integrating fine-resolution self-reported personality assessments with computer vision analysis of urban environments, we identified significant spatial clustering of personality traits at the ZIP code level. Our regression analyses reveal that built environment features and socioeconomic characteristics explain substantial variance in personality distributions, with Openness showing the strongest model fit (R2 = 0.47), followed by Agreeableness, Conscientiousness, Extraversion, and Neuroticism. Grouped built environment categories, socioeconomic factors, and demographic composition showed trait-specific patterns of association. These findings illustrate how personality traits could be associated with physical spaces at a smaller geographic scale than previously examined. Our results provide empirical evidence for understanding the link between psychological characteristics and environmental features, which can potentially enrich geography studies from a human-centered perspective.