Do You Care Who Flagged This Post? Effects of Moderator Visibility on Bystander Behavior

Abstract

This study evaluates whether increasing information visibility around the identity of a moderator influences bystanders’ likelihood to flag subsequent unmoderated harassing comments. In a 2-day preregistered experiment conducted in a realistic social media simulation, participants encountered ambiguous or unambiguous harassment comments, which were ostensibly flagged by either other users, an automated system (AI), or an unidentified moderation source. The results reveal that visibility of a content moderation source inhibited participants’ flagging of a subsequent unmoderated harassment comment, presumably because their efforts were seen as dispensable, compared to when the moderation source was unknown. On the contrary, there was an indirect effect of other users versus AI as moderation source on subsequent flagging through changes in perceived social norms. Overall, this research shows that the effects of moderation transparency are complex, as increasing visibility of a content moderator may inadvertently inhibit bystander intervention.

Exposure and Reactions to Cancer Treatment Misinformation and Advice: Survey Study

“…Thus, reducing cancer misinformation through unique, indirect platform affordances, such as flagging and reporting, appears to be more promising. These prosocial interventions have been part of effective digital bystander interventions, with increasing evidence of their ability to encourage supportive community action in the face of misinformation that perpetuates injustice, harassment, and harm [ 39 - 43 ]. However, individuals must be able to discern what is misinformation and know how to act; we need prompts and messages to help people question suspicious information and direct community action, as knowing how to intervene is a critical step in the human-computer interaction applications of the bystander intervention model [ 41 ].…”

Shall AI moderators be made visible? Perception of accountability and trust in moderation systems on social media platforms

“…This study was part of a larger experiment, which resulted in two separate studies, each one involving different research questions, hypotheses, and dependent variables (first study: Bhandari et al, 2021). The experiment was preregistered on the Open Science Framework, with the research questions, measures, and analysis plan available at the following https://osf.io/bwm8e.…”

“…The context and intentionality, previous interactions between the harasser and the person being harassed, offline power dynamics and many other factors can all be important in determining if a post or a comment constitutes harassment (Langos, 2012). In fact, flagging can sometimes move from a mechanism of reporting and upstanding into something that can be “gamed” and abused (Crawford and Gillespie, 2016). Finally, because of ambiguity in harassment comments, building automated systems to detect harassment can be difficult and these systems cannot necessarily understand and interpret ambiguity (Nadali et al, 2013).…”

“…has its own set of community guidelines and regulations. Yet, most moderation systems and decisions remain opaque (Banchik, 2020; Crawford and Gillespie, 2016; Suzor et al, 2019), which can leave users confused and unsure about how and why content moderation occurs (Crawford and Gillespie, 2016; Jhaver et al, 2019a). While there is abundant critical scholarship focused on the issue of content moderation (e.g.…”

Cyberbullying in the Metaverse

“…Human moderators analyze reported material and then act on cyberbullying. In community moderation, users report cyberbullying and a human moderator analyzes the material (Bhandari et al, 2021). Data on public postings, such as flags and cyberbullying complaints, is needed to identify patterns and advise policy (Howell & Burruss, 2020).…”