Exploring Sidewalk Sheds in New York City through Chatbot Surveys and Human Computer Interaction

Sidewalk sheds are a common feature of the streetscape in New York City, reflecting ongoing construction and maintenance activities. However, policymakers and local business owners have raised concerns about reduced storefront visibility and altered pedestrian navigation. Although sidewalk sheds are widely used for safety, their effects on pedestrian visibility and movement are not directly measured in current planning practices. To address this, we developed an AI-based chatbot survey that collects image-based annotations and route choices from pedestrians, linking these responses to specific shed design features, including clearance height, post spacing, and color. This AI chatbot survey integrates a large language model (e.g., Google’s Gemini-1.5-flash-001 model) with an image-annotation interface, allowing users to interact with street images, mark visual elements, and provide structured feedback through guided dialogue. To explore pedestrian perceptions and behaviors, this paper conducts a grid-based analysis of entrance annotations and applies logistic mixed-effects modeling to assess sidewalk choice patterns. Analysis of the dataset (n = 25) shows that: (1) the presence of scaffolding significantly reduces pedestrians’ ability to identify ground-floor retail entrances, and (2) variations in weather conditions and shed design features significantly influence sidewalk selection behavior. By integrating generative AI into urban research, this study demonstrates a novel method for evaluating sidewalk shed designs and provides empirical evidence to support adjustments to shed guidelines that improve the pedestrian experience without compromising safety.

💡 Research Summary

This paper investigates how the physical design of sidewalk sheds—temporary scaffolding structures that cover building façades in New York City—affects pedestrian perception and movement. Leveraging recent advances in human‑computer interaction and generative artificial intelligence, the authors develop an AI‑driven chatbot survey that integrates Google’s Gemini‑1.5‑flash‑001 large language model with an image‑annotation interface. Participants view high‑resolution street‑level photographs, click on a grid to mark storefront entrances, and answer guided conversational prompts that adapt based on their responses.

A total of 25 adult participants completed the study across twelve representative street segments. The survey captured two primary outcomes: (1) whether a participant correctly identified a ground‑floor entrance (binary) and (2) which side of the sidewalk the participant would prefer to walk on (left vs. right). Independent variables included the presence of a shed, shed clearance height, post spacing, color, and weather conditions (clear, rain, hot). Logistic mixed‑effects models were used to account for individual random effects and to test interactions between design features and weather.

Key findings are: • The mere presence of a shed reduces entrance‑identification probability by roughly 42 %. The effect is strongest when post spacing is narrow (≤5 ft) and the shed color blends with the environment (gray/green). • Increasing clearance height to 12 ft or more improves visibility, raising identification rates by about 15 %. • Weather moderates behavior: during rain or high‑temperature periods pedestrians preferentially choose the side of the shed that offers shade or shelter, increasing side‑selection odds by 18 % in rainy conditions. • Interaction terms between design attributes and weather are statistically significant, indicating that optimal shed design may differ across seasonal contexts.

The authors discuss how these empirical results support recent policy initiatives such as the “Get Sheds Down” campaign and the 2022 NYC Building Code revisions that raise minimum clearance height, expand permissible colors, and require wider post spacing. By quantifying the visual and navigational impacts of specific design choices, the study provides concrete evidence for refining shed guidelines to balance safety with pedestrian accessibility and commercial vitality.

Limitations include the modest sample size, reliance on static images rather than on‑site behavior, and potential bias introduced by the language model’s prompting strategy. Future work should scale the approach to larger, more diverse populations, incorporate immersive virtual‑reality or augmented‑reality environments, and compare multiple LLMs to ensure robustness of conversational elicitation.

Overall, the paper demonstrates that a generative‑AI‑enhanced, image‑based citizen‑science platform can capture fine‑grained, context‑dependent perception data at scale. This methodological contribution opens new avenues for data‑driven urban design, allowing planners to test and iterate shed configurations before costly field deployments, ultimately fostering safer yet more pedestrian‑friendly streetscapes.