Online Task Allocation by Price Discrimination

Quality control in crowdsourcing is an important issue. A major approach is to select a best worker for a task. Many of the existing studies on a task allocation in crowdsourcing assume a ´push´ crowdsourcing model where the system can decide which tasks are sent to a worker. On the other hand, a task allocation problem for a ´pull´ crowdsourcing model such as Amazon Mechanical Turk has not been sufficiently studied. To tackle this issue we try to employ the price discrimination theory in microeconomics. However, the theory assumes the unlimited supply of resources, which does not fit the crowdsourcing situation and is difficult to apply directly. To overcome this difficulty, we provide a model of price discrimination for task allocation in crowdsourcing and give an efficient method for calculating the optimal payment setting. Furthermore, we evaluate the proposed method by using the data obtained from MTurk experiments and show it performs well.