Trust in IoT-enabled mobility services: predictive analytics and the impact of prediction errors on the quality of service in bike sharing

Real-time communication and information flows among vehicles, infrastructure, travelers' smartphones and service providers' backend systems constitute the everyday experience of the Internet of Things (IoT) in the transport sector. Prediction Errors (PEs), either as outcomes of imperfect model performance or as exploitation of model vulnerabilities to compromise the security of the system, can affect the Quality of Service (QoS) of transport systems. This paper contributes to the research literature in trust in IoT for transport. It does so by developing a methodological framework to quantify the users' impacts of prediction reliability in IoT-enabled mobility services. We apply such framework to London's bike sharing scheme. Two predictive algorithms are used to forecast bike availabilities at different docking stations. Prediction errors affect the reliability of information provided in real time to bike users, and hence the utility accrued by users from the information provided. The variation in consumer surplus for increased reliability and the dissatisfaction of the users are analyzed through simulation. The results demonstrate the impact of predictive algorithms on the QoS of transport services and highlight the value of data collection for empirical estimations.