Decentralized Governance on Two-Sided Platforms: Crowdsourcing, Learning, and Debiasing

Abstract

Disputes over transactions on two-sided platforms are common and are usually arbitrated through platforms’ customer service departments or third-party service providers. In this paper, we study crowd-judging, a novel crowd-sourcing mechanism whereby users (buyers and sellers) volunteer as jurors to decide disputes arising from the platform. To understand this phenomenon, we use a rich dataset from the dispute resolution center at Taobao, a leading Chinese e-commerce platform. While this mechanism enhances resolution speed, there are concerns that crowd-jurors may exhibit a form of in-group bias (where buyers favor the buyer and sellers favor the seller in a dispute), and that such in-group bias may systematically sway case outcomes given the majority of users on such platforms are buyers. We find evidence consistent with this concern: on average, a seller juror is approximately 10% likelier to vote for a seller. Such bias is 70% higher among cases that are less clear-cut and decided by a thin margin. Conversely, the bias reduces dramatically as users gain crowd-judging experience: in-group bias when jurors have the sample-median level of experience is 95% lower than when jurors are completely inexperienced. This suggests learning-by-doing may mitigate biases associated with socioeconomic identification. Partly due to this learning effect, our simulation shows that in-group bias influences the outcomes of no more than 2% of cases under the current randomized case allocation process, and can be further reduced under dynamic policies that better allocate experienced jurors. Such findings offer promising evidence that crowd-sourcing can be an effective dispute resolution mechanism to govern online platforms, and that properly designed operating policies can further improve its efficacy.