Adaptive Budget Allocation for Cooperative Task Solving in Crowdsourcing
Yuya Itoh, Shigeo Matsubara · 2021 IEEE International Conference on Big Data (Big Data) · 2021
This paper proposes a new budget allocation method for crowdsourced sequential tasks. Complex tasks are decomposed into interdependent subtasks that can be executed cooperatively by individual workers. In the case of sequential tasks, the output of a task becomes the input to another task, and the quality of the final artifact thus depends on the qualities of the preceding tasks. In crowdsourcing, the abilities of workers are often difficult to learn in advance. Thus, a fixed budget allocation for the component subtasks cannot deal with a dynamic situation. Also, it is often difficult for a requester to accurately evaluate the quality of intermediate artifacts, which can result in budget misallocation and waste. To overcome these difficulties, we formalize the budget allocation problem as a partially observable Markov decision process (POMDP) by introducing quality evaluation actions and developing a contingent budget allocation method, which generates a conditional plan given uncertainty about the intermediate states and action effects. Experimental simulation results show that the proposed method can find a solution in a reasonable time and improve the quality of the final artifact.