Optimizing expertise management in crowdsourcing contests: The impact of information structures on open innovation...

For companies and governments alike, crowdsourcing contests have emerged as a popular approach to harness the benefits of open innovation. Consequently, numerous game models have been developed to analyze these contests and model the interactions in equilibrium between a contest's sponsors and its participants. Some models assume incomplete information where player abilities are private information, while others model complete information scenarios where these abilities are revealed. However, to date, no uniform framework has been proposed to comparatively analyze the theoretical outcomes of these two distinct information structures. This paper presents a formulation to derive a complete information model from a previously published incomplete information model using the variational method. This formulation allowed us to derive the effort levels of participants given a mixed-strategy equilibrium in a single-prize, complete information crowdsourcing contest. We then conducted a comparative analysis between the complete and incomplete information models, considering contests with and without a registration phase (i.e., where the sponsor knows/does not know the participants' abilities beforehand). The key findings are as follows. With a registration phase, incomplete information tends to yield higher quality solutions when the abilities of top participants exceed a certain skill threshold. Without registration, complete information may lead to better solutions, especially when the sponsors have a tendency toward risk-taking. Finally, the complete information mixed-strategy equilibrium is used to provide a game-theoretic interpretation of the recent 'involution' and 'lying flat' phenomenon observed in China, where players withdraw from a contest when competition is excessive. This offers the first explanation for this behavior based on an equilibrium strategy, in contrast to the previous explanations, which are all psychological.