Youlong Gu

Understanding people’s preferences is crucial for urban planning, yet current approaches often combine responses from multi-cultural populations, obscuring demographic differences and risking amplifying biases. We conducted a large-scale urban visual perception survey of streetscapes worldwide using street view imagery, examining how demographics—including gender, age, income, education, race and ethnicity, and personality traits—shape perceptions among 1,000 participants with balanced demographics from five countries and 45 nationalities. This dataset, Street Perception Evaluation Considering Socioeconomics, reveals demographic- and personality-based differences across six traditional indicators—safe, lively, wealthy, beautiful, boring, depressing—and four new ones: live nearby, walk, cycle, green. Location-based sentiments further shape these preferences. Machine-learning models trained on existing global datasets tend to overestimate positive indicators and underestimate negative ones compared to human responses, underscoring the need for local context. Our study aspires to rectify the myopic treatment of street perception, which rarely considers demographics or personality traits.

Urban visual perception is important for the human experience in cities, shaped by intertwined characteristics of urban landscapes. By quantifying and explaining these perceptual experiences, researchers can gain insights into human preferences and support decision-making in planning and design. However, past studies have shown inconsistencies in survey design and ambiguities in reporting, leading to concerns about the reliability and reproducibility of results. This study proposes the first comprehensive framework to guide image-based survey design for capturing perceptions of outdoor urban environments across different scenarios, addressing the lack of methodological standardization in current research. We reviewed existing surveys to identify key parameters, conducted comprehensive between-subject and within-subject surveys, and performed statistical analyses to determine best practices for survey design across different contexts. Aiming to set a potential community standard, our study doubles as a blueprint for a reporting protocol for survey designs. Based on the results, we recommend: (1) meeting a minimum of 12 and 22 ratings per image for Likert Scale and Pairwise Comparison studies to reach survey reliability, respectively, and reporting these alongside other survey design parameters to enhance transparency and reproducibility; and (2) when resource allows larger experiments, adopt a ranking method such as Pairwise Comparison to achieve firmer rating results; and (3) using perspective (non-panoramic) images more frequently, as they exhibit comparable overall scores to panoramic images (R mostly \textgreater0.7), while being more widely available via crowdsourced sources, supporting their use in large-scale visual perception research.