AI Chatbots and Video Analytics: Revolutionizing Urban Perception for Safer Cycling

Bridging the Perception Gap in Urban Planning

      Conventional urban planning and transport studies often face a significant challenge: understanding how people truly perceive their environment. Traditional surveys, while valuable, tend to rely on participants recalling past experiences, which can be subjective and prone to memory bias. This is particularly true for dynamic environments like urban cycling routes, where factors influencing safety perceptions are constantly changing. As highlighted in a recent pilot study from New York City by Ren et al. (2026), there's a critical need for methods that can capture "in-the-moment" experiences to truly grasp human perception of cycling safety. Such insights are crucial for developing infrastructure that genuinely encourages sustainable transportation and enhances public well-being.

The Innovation: Video-Based Conversational AI Surveys

      To overcome the limitations of traditional surveys, researchers are exploring innovative approaches that leverage advanced technologies. One such groundbreaking method combines video-based surveys with conversational AI chatbots. This allows for the collection of rich human perceptions regarding cycling safety and, more importantly, the underlying reasons behind these perceptions. The AI chatbot developed for this study uses a modular Large Language Model (LLM) architecture, a design approach where different components work together seamlessly. This architecture integrates prompt engineering (crafting precise instructions for the AI), state management (tracking the conversation's flow), and rule-based control (pre-defined rules for interaction) to facilitate natural and structured human-AI conversations. The goal is to create a dynamic interaction where users can respond to specific video segments, offering immediate, contextual feedback that is often missed in static questionnaires.

Pilot Study in New York City: Assessing Feasibility

      The pilot study, conducted across nine distinct street segments in New York City, involved sixteen participants whose complete responses were analyzed to assess the feasibility of this novel approach. The research evaluated two key aspects: method feasibility and data feasibility. Method feasibility focused on the user experience with the video-based conversational chatbot. Participants rated ease of use, supportiveness, and efficiency on a seven-point scale, and chatbot usability (personality, roboticness, friendliness) on a five-point scale. The results were positive, with mean scores of 5.00 out of 7 for user experience and 3.47 out of 5 for chatbot usability, indicating a generally favorable and effective interaction.

Unpacking the Findings: User Experience and Data Potential

      The positive user experience scores underscore the potential for engaging citizens in urban planning processes through interactive AI. Beyond user satisfaction, the study rigorously assessed data feasibility using multiple analytical techniques. Natural Language Processing (NLP), specifically KeyBERT, was employed to extract key built-environment attributes from participants' qualitative feedback, helping to understand overall safety and specific feature analysis. K-means clustering, a machine learning technique for grouping similar data points, was used for semantic analysis to identify common reasons and suggestions provided by users. Finally, regression analysis helped estimate how built-environment features and demographic variables influenced perceived safety outcomes. These comprehensive analytical methods demonstrate the capability of this AI-driven approach to generate nuanced, actionable insights far beyond what traditional surveys typically provide. The combination of qualitative and quantitative analysis yields a holistic understanding of how cyclists experience urban infrastructure, making the data highly valuable for future planning.

Transforming Urban Planning with AI-Powered Insights

      The findings from this pilot study signify a major step forward for urban and transport planning. AI chatbots offer a new avenue for collecting dynamic data on human perception, behavior, and even future visions for urban development. By capturing immediate, context-rich feedback, urban planners can make more informed decisions about infrastructure design, policy formulation, and safety interventions. This methodology has the potential to move urban planning from reactive problem-solving to proactive, citizen-centric design, creating more bikeable, safer, and sustainable cities globally. For enterprises involved in AI Video Analytics or smart city initiatives, this research highlights a powerful application for combining computer vision with natural language processing to enhance operational intelligence and public engagement. Companies like ARSA, with expertise in deploying robust, edge AI systems, are well-positioned to implement such solutions in real-world urban environments.

      This innovative approach is a testament to how cutting-edge AI can be harnessed to address complex societal challenges, fostering smarter, more responsive urban ecosystems. ARSA Technology, an AI & IoT solutions provider, has been experienced since 2018 in delivering practical AI deployments that provide real-time operational intelligence for governments and enterprises across various industries.

      **Source:** Ren, F., Zhang, Z., Mendel, T., & Yabe, T. (2026). Assessing the Feasibility of a Video-Based Conversational Chatbot Survey for Measuring Perceived Cycling Safety: A Pilot Study in New York City. arXiv preprint arXiv:2604.07375. Available at: https://arxiv.org/abs/2604.07375

      To explore how ARSA Technology can help your organization leverage AI and IoT for advanced data collection and operational insights, we invite you to contact ARSA for a free consultation.