Social Incentive and Solution Reliability: Evidence from Crowdsourcing Competitions

Crowdsourcing competitions are increasingly used by firms to tackle internal R&D tasks; however, the reliability of solutions generated through these competitions remains underexplored. This study examines how social incentives influence solution reliability in such contests. Drawing on the attention-based view (ABV) and the exploration-exploitation (E- E) framework, we frame social incentives as attentional cues that redirect participants' focus toward socially visible performance metrics. This shift in attention can disrupt the balance between optimizing explicit performance and maintaining implicit reliability. To test this hypothesis, we leverage a natural experiment on a machine learning contest platform, where the introduction of a “following” feature generated exogenous variation in users' exposure to social incentives. Our empirical results indicate that this feature negatively affects solution reliability. The robustness of this finding is confirmed through multiple checks, including assessments of parallel trends, placebo tests, difference-in-difference-in-differences models, regression discontinuity in time analyses, and instrumental variables. Further analyses reveal that participants who gained followers or followed others submitted more solutions; however, these submissions were less reliable. Notably, high competition intensity mitigated this negative effect. This study contributes to the engineering management literature by addressing a crucial but understudied aspect of crowdsourced R&D projects: solution reliability. By integrating the E-E framework with ABV, it also advances research on open innovation by framing reliability as an attentional outcome shaped by social incentives rather than as a purely technical attribute. Overall, the findings reveal an unintended consequence of social incentives and provide actionable guidance for managers overseeing crowdsourced R&D initiatives.