Effect of Knowledge Sharing on Participant Performance in Artificial Intelligence Contests: A Quasi-experiment

The effective use of open data and shared knowledge to enhance digital innovation in enterprises has become a trending research area in the cross-disciplinary field of AI and management. And AI contests have emerged as pivotal forms for the industrialization of digital technologies and open innovation. Yet, in a competitive and cooperative setting, the role of knowledge sharing in solution generation among contestants remains uncertain. This study conducted a quasi-experiment in conjunction with the launch of the AI contest platform's knowledge-sharing function, and the Propensity Score Matching and Difference-in-Differences (PSM-DID) methodology was used to empirically investigate the knowledge-sharing effect on participant performance, with further analysis on participant's heterogeneity. The findings indicate that the introduction of the knowledge-sharing function positively affects participant performance, though with a latency period. Additionally, the knowledge-sharing function shows a heterogeneous impact on participant performance; while the performance of individual participants significantly improved, team participants experienced a decrease, yet not significantly. Participants with more entries and longer tenure significantly increased, whereas those with more submissions showed a significant performance decline. These results offer valuable management implications for the platform and contestants to effectively utilize the knowledge-sharing function.