Evaluating Predictive AI for Taxi Fleet Positioning Management

Sammanfattning: Predictive AI has huge potential to increase efficiency in taxi networks, improving incomes of drivers and passengers’ time, by calculating passenger demands and making the informationaccessible to drivers. However, to gain these benefits, drivers must accept the technology, and designers must understand how their perceptions govern acceptance. This study evaluates predictive AI applied to taxi fleet management from a user acceptance perspective, in the context of a proposed extended technology acceptance model, with added user experience constructs.Process modelling of a taxi organisation was performed to see how the algorithm fit into that context, and an app was designed to give taxi drivers access to its output. After a period, semistructured interviews were conducted, and thematic analysis was performed to identify what factors were important for their acceptance. The study found that visual design, interactivity, complexity, compatibility, utility, inclusivity, norms, presentation, and trustworthiness were factors that needed to be addressed by designers for successful implementation of predictive AI in the context of taxi fleet management.