This article investigates what causes an urban environment to be perceived as safe using Stockholm, the capital of Sweden, as the study area. The study integrates AI-generated safety scores from street view images, image segmentation techniques and conventional and crowdsourced data using Geographical Information Systems (GIS) and regression models. After accounting for income, crime and other area characteristics, the models reveal that areas with lower safety scores primarily consist of areas with a relatively large percentage of roads in industrial and/or interstitial mixed residential areas. Conversely, higher safety scores are found in large but distinct combinations of buildings, vegetation and open sky, from detached single-family housing to inner city high-density built areas. To enhance safety in an area, good contextual knowledge of the area is fundamental to prioritize interventions in interstitial mixed residential zones where roads and highways may be the dominant features.

This paper analyses the alignment between urban planning targets and realized projects in Stockholm from 2012 to 2019. Employing both conformance and performance-based evaluation methods, it scrutinizes changes in street networks at both citywide and project -specific levels. The study entails a comprehensive review of planning documents, detailed configurational analyses, and an assessment of public transportation access. By examining the discrepancy between planning objectives and actual urban transformations, the research underscores the necessity for more effective monitoring techniques to achieve planning targets accurately. Notably, the analysis reveals that while municipal comprehensive plans (MCPs) articulate ambitious goals for urban development, the translation of these objectives into tangible changes in the built environment remains challenging. Despite the existence of regulatory frameworks and legal mandates, such as the MCPs and zoning plans, the paper identifies significant inconsistencies in target implementation across different projects. These disparities are attributed to various factors, including changes in political will, long timeframes and limited understanding of the complexity in the built environment. Furthermore, the research advocates for the integration of digital tools and methodologies to enhance monitoring and evaluation processes in urban planning. Overall, this study contributes to the ongoing discourse on urban development by providing valuable insights into the challenges and opportunities inherent in aligning planning targets with actual urban transformations, thereby offering recommendations for enhancing the efficacy of urban planning practices.

The safety perception of the built environment, rather than the sheer number of crimes and incivility behavior, is a fundamental driver of public policies intended to improve urban safety. Traditional surveys often capture neighborhood residents’ perceived safety, but may not fully reflect the perceptions of people who are unfamiliar with the area. In this study, focused on the city of Stockholm, Sweden, we develop a geospatial artificial intelligence (GeoAI) approach using street view images and recruiting locals to create a measure of citywide residents’ safety perceptions. We compare the measures from the survey based on neighborhood residents’ responses with those from the GeoAI approach to better understand the relationship between these safety measures. We model the two forms of safety perceptions and their disparities (i.e., perception bias) as a function of the city’s land use and its socio-demographics. Results confirm that while the GeoAI-based measures better capture people’s instant impressions of the built environment across the city, the survey-based measures reflect their overall daily experiences of specific areas. Regions that appear to be economically vibrant and have inner-city streetscapes are perceived as safe places from visual appearance but are not always perceived as such by residents. Older adults tend to overestimate their likelihood of being victimized by crime, which may enlarge perception bias. The study concludes by critically assessing the potential ethical issues (e.g., spatial bias, population bias) in the proposed methodology and making suggestions for future research.