Making the Crowd Wiser: (Re)Combination Through Teaming in Crowdsourcing

Firms have widely adopted crowdsourcing because of its advantages in conducting parallel search by recruiting a large number of solvers to generate innovative solutions to challenging problems. Formerly independent solvers may join forces to collaboratively develop solutions through teaming to cocreate solutions that would not have emerged from any parallel yet independent path alone. However, when teams are formed on the fly during crowdsourcing, they inevitably alter the distribution of the quality of all solutions via reducing the number of independently developed solutions. There is scant research examining whether, to what extent, and how teaming could compensate for the loss in parallel search and eventually increase the likelihood of firms obtaining high-quality solutions, namely crowdsourcing effectiveness. In this paper, we posit that solvers are likely to make use of a variety of publicly available information that crowdsourcing platforms provide as quality signals in determining potential teammates for teaming. In turn, this reliance on observable quality signals will shape the self-selected teaming process and subsequent crowdsourcing effectiveness. Using simulation experiments, we find that the impact of teaming on crowdsourcing originates from the immediate returns from identifying other solvers (or their solutions) to integrate with and the potential returns from teamwork-based collaboration, albeit conditionally depending on problem complexity and the timing of teaming. Moreover, under certain conditions, teaming may not compensate for the loss in parallel search and will hamper global crowdsourcing effectiveness. The findings shed new light on crowdsourcing effectiveness and design and point to exciting new lines of inquiry.

History: Wonseok Oh, Senior Editor; Zhengrui Jiang, Associate Editor.

Funding: This work was supported by the Singapore Ministry of Education [Social Science Research Thematic Grant MOE2017-SSR]. The authors also acknowledge that the computational work involved in this research was partially supported by NUS IT’s Research Computing group.

Supplemental Material: The online appendices are available at https://doi.org/10.1287/isre.2023.0556.