Using artificial intelligence (AI) to design better streets for pedestrians? Preference in streetscape transformation from an AI-generated video-based stated preference survey

Abstract

Street experimental interventions are gaining popularity as a means of evaluating, prototyping, and catalyzing positive changes in the built environment. However, a prevalent reliance on a trial-and-error methodology, often coupled with a qualitative design approach, characterizes many of these interventions. The scarcity of quantitative studies investigating the design of street experiments for landscape transformation may be attributed to the inherent challenges in generating and evaluating viable options. This study aims to bridge this gap by leveraging artificial intelligence (AI) techniques to craft video visualizations of street experiment setups, facilitating the creation of choice sets for a stated preference discrete choice experiment. The logit choice model scrutinizes not only the physical attributes of experiments but also encompasses the perception of functional, environmental, social, safety, and commercial aspects of streets. Our choice analysis unveils determinants at both the physical and individual levels, shedding light on perceptions related to these physical factors. The findings offer insights that can inform recommendations for enhancing the walking experience and strategically planning socially accepted street experiments. The outcomes of this research possess the potential to guide the development of street interventions, effectively transforming car-dominated streets into vibrant public spaces for people.