An Incentive Mechanism Based on a Bayesian Game for Spatial Crowdsourcing

Abstract

Crowds are playing an increasingly important role in the research and application of geoscience by providing spatial data via crowdsourcing. However, low-public participation and poor quality of data submissions have greatly restricted the application of spatial crowdsourcing (SC) and other similar models, thus garnering the attention of scientists in this field. In this paper, we design a precise incentive mechanism based on a Bayesian game for SC that successfully avoids the conditions limited by the Gibbard-Satterthwaite impossibility theorem. Under this mechanism, the outsourcer carries out a Bayesian game with the participants under the circumstance of incomplete information by setting a certain amount of reference information that is not visible to the participants. Participants gain far more utility by telling the truth than that of they gain by lying and thus have a stronger motivation to submit higher-quality data. In implementing this mechanism to automatically compute the actual utility of participants and integrate data results, we propose a geometric primitive matching algorithm based on the Jaccard coefficient. Through both rigorous theoretical analyses and real experiments, the incentive mechanism that we propose is incentive-compatible and can greatly improve data quality.