一种面向众包的基于信誉值的激励机制

Reputation-based Incentive Mechanisms in Crowdsourcing

众包是互联网带来的一种分布式问题解决模式。然而,由于工作者和任务发布者具有自私特性并且致力于获得自身效益的最大化,使得在众包应用中,存在内部的激励问题。该文主要完成以下工作:首先,基于重复博弈,提出一种基于信誉值的激励模型,用于激励理性工作者高质量地完成任务;其次,该激励模型中同时设置了惩罚机制,将针对恶意工作者做出相应惩罚。仿真结果表明,即使在自私工作者比例为0.2的条件下,只要合理选择惩罚参数,均可有效激励理性工作者的尽力工作,众包平台的整体性能可以提升至90%以上。

Crowdsourcing is a new distributed problem solving pattern brought by the Internet. However, intrinsic incentive problems reside in crowdsourcing applications as workers and requester are selfish and aim to maximize their own benefit. In this paper, the following key contributions are made. A reputation-based incentive model is designed using repeated game theory, based on thorough analysis for current research on reputation and incentive mechanism; and a punishment mechanism is established to counter selfish workers. The experiment results show that the new established model can efficiently motivate the rational workers and counter the selfish ones. By setting punishment parameters appropriately, the overall performance of crowdsourcing system can be improved up to 90%, even if the fraction of selfish workers is 20%.