Efficiency of Community-Based Content Moderation Mechanisms: A Discussion Focused on Birdwatch

As user-generated online content has been flourishing with both useful information and misinformation. One of the complexities surrounding such phenomena is its huge amounts of data and its associated difficulties to effectively moderate content, particularly as most initiatives are centralised and fraught with its intrinsic trust issues. One of the few examples using mainly a decentralised (i.e., community-driven) mechanism is Twitter’s Community Notes (once named as Birdwatch) experimental project. This paper thus is about testing the efficiency of such community-based content moderation mechanism and scenarios of interest aiming to better understanding how the users themselves better moderate online content. This is done through an agent-based approach and three conclusions are discussed in detail: (1) to some extent the community is able to fight against misinformation, (2) a Birdwatch-like mechanism can indeed boost the community’s content moderation ability, but there is a nontrivial trade-off between social influence and content timeliness and (3) a simple proposition, in the form of a reminder mechanism to users, cannot fulfil the task of improving the content moderation efficiency, which means a different approach to design is needed.