Incentive Mechanism for Macrotasking Crowdsourcing: A Zero-Determinant Strategy Approach

Abstract

Macrotasking crowdsourcing systems (MCSs), such as Google Helpouts and Elance have emerged as an effective paradigm for improving human intelligence and activity to solve a wide variety of tasks. Requesters often post tasks to the MCS and competitive workers solve the tasks to earn the reward. However, rational and selfish workers in the MCS aim to strategically maximize their own benefit by exhibiting malicious behaviors, thereby decreasing the efficiency of systems. Herein, we present a novel game-theoretic mechanism to incentivize the competitive and selfish workers to provide high-quality solutions in the MCS. We first formulate the crowdsourcing problem as a multiplayer iterated game with incomplete information, where each worker has certain private information (such as solution quality), but does not know what other workers do. Subsequently, we propose an incentive mechanism in terms of zero-determinant (ZD) strategies aiming to improve the social welfare of the MCS, which serves to incentivize the competitive selfish workers toward high-quality solutions. Moreover, we find the conditions for reaching the maximum social welfare of the MCS. Numerical illustrations demonstrate a high and stable social welfare of the MCS with the proposed ZD strategies mechanism.