Abstract
Many governments and companies offer incentive programs (e.g., allowance) to encourage employees to share rides for their daily work commute because ridesharing helps reduce travel costs, road congestion, and greenhouse gas emissions. For the program success, it is critical to understand employees’ motivation to use this service and effectively incentivize new users’ adoption and retain existing users by providing the best ridesharing experience. In this study, our goal is to learn user incentives and preferences as well as risks in employee ridesharing. We use a choice model to estimate the personalized utility of different choices among possible transportation modes and learn from longitudinal behaviors using reinforcement learning. The integrative model is derived theoretically and evaluated empirically using a real-world employee ridesharing data. We find that ridesharing is risky and company employees, especially drivers, adopt the service mainly because of social relationship with colleagues rather than the financial incentive.
Recommended Citation
Yan, Wangcheng; Zhou, Wenjun; Tan, Chang; and Fan, Lei, "Employee Ridesharing: Reinforcement Learning and Choice Modeling" (2019). AMCIS 2019 Proceedings. 21.
https://aisel.aisnet.org/amcis2019/data_science_analytics_for_decision_support/data_science_analytics_for_decision_support/21
Employee Ridesharing: Reinforcement Learning and Choice Modeling
Many governments and companies offer incentive programs (e.g., allowance) to encourage employees to share rides for their daily work commute because ridesharing helps reduce travel costs, road congestion, and greenhouse gas emissions. For the program success, it is critical to understand employees’ motivation to use this service and effectively incentivize new users’ adoption and retain existing users by providing the best ridesharing experience. In this study, our goal is to learn user incentives and preferences as well as risks in employee ridesharing. We use a choice model to estimate the personalized utility of different choices among possible transportation modes and learn from longitudinal behaviors using reinforcement learning. The integrative model is derived theoretically and evaluated empirically using a real-world employee ridesharing data. We find that ridesharing is risky and company employees, especially drivers, adopt the service mainly because of social relationship with colleagues rather than the financial incentive.