PROJECT TITLE :

Optimizing mobile crowdsensing platforms for boundedly rational users

ABSTRACT:

Participatory mobile crowdsensing (also known as MCS) requires users to make decisions repeatedly between a limited number of available options, such as whether or not they will contribute to a task and which task they will contribute to. The platform that is in charge of coordinating the MCS campaigns will frequently engineer these choices by selecting MCS tasks to recommend to users and offering monetary or in-kind rewards to motivate their contributions to the various MCS tasks. In this paper, we revisit the question of how to optimize the contributions of mobile end users to MCS tasks, which has received a significant amount of previous research. To the contrary, we take a different approach than the vast majority of the other research in this field by taking into account the bounded rationality that is inherent in human decision making. Bounded rationality is a phenomenon that arises as a result of cognitive and other kinds of constraints, such as time pressure, and it has received a significant amount of attention from researchers in the field of behavioral science. When modeling how boundedly rational users react to MCS task offers modeled as Fast-and-Frugal-Trees (FFTs), our first step is to draw on previous research conducted in the field of cognitive psychology. The decision-making process in FFTs proceeds through sequentially parsing lexicographically ordered features, which results in choices that are satisfying but are not necessarily optimal. This is because each MCS task is modeled as a vector of feature values. Following this, we formulate, investigate, and ultimately resolve the novel optimization problems that arise for both nonprofit and for-profit MCS platforms as a result of this context. When compared to heuristic rules that do not take into account the lexicographic structure that is involved in human decision making, the evaluation of our optimization approach reveals significant gains in both the revenue generated by the platform and the quality of the contributions made to the tasks. We demonstrate how easily this modeling framework can be extended to platforms that provide users with a variety of task options simultaneously. Finally, we discuss how these models can be trained, iterate on their assumptions, and point out their implications for applications beyond MCS. These applications include situations in which end-users make decisions through the mediation of mobile or online platforms.