To Share, or Not to Share? Community-Level Collaboration in Open Innovation Contests

Open innovation contests have been incredibly successful at producing creative designs and solutions. While participants compete for prizes in these contests, over half of contests conducted on online platforms allow participants to share ideas during contests, for benefits such as individual learning, community building, and cumulative innovation. Such sharing, however, is at tension with the competitive nature of crowd contests. To understand this tension, this study investigates community-level sharing of code on Kaggle, a contest platform for predictive modeling. Analyzing data on 25 contests in 2015 and 2016, we find that 10% of users shared code during contests, that participants doing medium well in the contest were the most likely to share code, and that sharing code improved individual, but not collective performance. These findings allow us to contribute insights about the participants, conditions, process, and outcomes of community-level collaboration to both research on and design of open innovation contests.