Shaping Pro and Anti-Social Behavior on Twitch Through Moderation and Example-Setting

Abstract

Online communities have the potential to be supportive, cruel, or anywhere in between. The development of positive norms for interaction can help users build bonds, grow, and learn. Using millions of messages sent in Twitch chatrooms, we explore the effectiveness of methods for encouraging and discouraging specific behaviors, including taking advantage of imitation effects through setting positive examples and using moderation tools to discourage antisocial behaviors. Consistent with aspects of imitation theory and deterrence theory, users imitated examples of behavior that they saw, and more so for behaviors from high status users. Proactive moderation tools, such as chat modes which restricted the ability to post certain content, proved effective at discouraging spam behaviors, while reactive bans were able to discourage a wider variety of behaviors. This work considers the intersection of tools, authority, and types of behaviors, offering a new frame through which to consider the development of moderation strategies.

Content Removal as a Moderation Strategy

“…The fact that the treatment group and the control group have similar noncompliance rates before removal supports the validity of our matching scheme. Overall, these observations confirm the causal role of comment removal in reducing noncompliance rates, echoing Seering et al [37] in spirit.…”

“…By discounting this underlying change, we can isolate the effect of comment removal from that of temporal effects. 16 Importantly, this methodology assumes the duration of the delay is not related to the behavior of interest; while in other domains this was shown to hold true [37], this assumption could be challenged in our domain. To mitigate this risk, we select u ′ such that their delay time t ′ D is as close as possible to t D .…”

Social motivations of live-streaming viewer engagement on Twitch

“…Interaction takes place in the forms of viewer-toviewer, viewer-to-streamer, and streamer-to-viewer (Anderson, 2017), and particularly the interaction between the viewer and the streamer has been shown to be influential on the use of the service (Gros et al, 2018). The communities of Twitch are not without their own problems, and studies have highlighted the potential problems that might arise in these mostly anonymous communities, and how proactive approaches such as moderation can help mitigate some of these problems (Seering, Kraut, et al, 2017).…”

Everyone Can Be a Star: Quantifying Grassroots Online Sellers’ Live Streaming Effects on Product Sales

“…The effect of live streaming has also been studied. Live streaming acts as virtual third places for users to socially interact with other people [13], creates a very intimate experience with a remote partner [24], improve learning performance for novice instructors [22], and influences pro and anti-social behaviors [25].…”