Enhancing customer satisfaction is a critical focus for public transit providers, as it directly
influences ridership. Consequently, public transport operators prioritize the collection and analysis of customer satisfaction data to refine their services and support sustainable urban mobility initiatives. Traditional data collection and analysis approaches, such as surveys and focus group interviews, offer direct perspectives into passenger experiences but are often expensive, time-consuming, labor-intensive, and may not meet the immediate need of capturing customer complaints. These issues underscore the need for innovative, non-traditional data collection strategies to complement the traditional ones. Crowdsourced data emerges as a viable solution, offering riders’ real-time insights into aspects like bus cleanliness, safety, and service reliability. However, despite its benefits, this method faces challenges like providing prompt feedback, managing vast amounts of public input, and efficiently organizing diverse comments. To address these challenges, this IDEA project proposes the development of an artificial intelligence (AI)-enhanced crowdsourcing platform specifically designed to gather public feedback and complaints regarding transit services. This platform will be powered by advanced AI technologies such as Large Language Models (LLM) and Natural Language Processing (NLP). A key feature of this platform will be its ability to automatically categorize customer feedback into specific topics to help with swift identification and resolution of issues. The platform will also address riders’ complaints with pre-defined instructions from operators, suggest actions to follow for operators, and flag items for manual review when necessary. Furthermore, the platform will be able to integrate public contributions with specific transit lines or stations, overlaying this information on transit service maps to enable spatial analysis and visualization for operators.

To build the proposed AI-enhanced crowdsourcing platform, its requirements and features will be defined, based on the challenges identified by transit agencies, e.g., intelligent user interface and automatically processing rider complaints. The building process will consist of user experience (UX)/user interface (UI) design; technology stack design; AI algorithm design; and deployment and launch. Presently available public opinion solicitation platforms will be evaluated and analyzed. Visual components will be designed, and  user flows will be developed to define how users will interact with the platform. The UX/UI design will be user friendly and cater to both transit riders and operators. Technologies for the client side (such as HTML, CSS, JavaScript, React, Angular) and for server-side operations (such as Node.js, Python, Ruby) will be evaluated and the most suitable one will be selected. Given that this is a web-based platform for use across different mobile operation systems, selection for the database system will be made from systems such as MySQL, PostgreSQL, MongoDB, etc. Based on the needed features, an AI algorithm will be designed to automatically process rider complaints. The algorithm will be able to handle tasks such as data cleaning, topic classification, suggesting potential actions for operators, and updating riders about the status of their complaints. A pipeline will be established to facilitate the flow of public input to a dashboard that operators can use to monitor and manage the platform. Transit operators will be able to perform basic spatial and temporal analysis of rider inputs and derive insights to inform possible actions. 

For the deployment of the web-based platform, a hosting environment will be selected and the servers configured accordingly. Subsequently, the platform will be deployed on those servers to test individual components for functionality and to ensure seamless integration across all parts of the application. The platform will initially be released internally to collect feedback and make necessary adjustments. It will then be tested with transit riders in collaboration with the Detroit and Houston transit agencies. Based on riders’ feedback, the platform’s functionality and any inconveniences encountered by riders when providing input will be assessed. The platform can be updated, as needed, based on the riders’ feedback. Training session will be held with relevant transit agency staff. Demonstrations will be conducted on how to navigate the platform, access features like data filtering and spatial-temporal analyses tabs, and following up with a specific comment.