Norm repair as an extension to the deliberative dialogue model

Abstract

According to research on deliberation, high-quality online dialogue is characterized by participants’ adherence to a set of ideal norms (e.g., civility, rationality). Deliberative norms are prescriptive in that they specify conditions under which dialogue ought to proceed. Empirical research consistently finds that real dialogue departs from these ideals, with norm violations routine even in dialogues that produce positive outcomes. To investigate this disparity between deliberative models and empirical reality, this article asks how online dialogue quality is conceptualized by those charged with maintaining it: content moderators. We report on three analyses of 32 interviews with Reddit moderators, collectively overseeing 123 subreddits. Analysis 1 used topic modeling to describe the moderators’ responses and found that they discuss dialogue quality through the lens of motivated practice. Analysis 2 evaluated whether moderators’ discussions of quality align with a conventional deliberative model. We found that, although moderators’ discussions generally aligned with civility and rationality norms, they deviate from the equality norm. Analysis 3 explains this discrepancy using an abductive analysis, finding that moderators prioritize excluding participants acting in “bad faith” (those who repeatedly refuse to repair norm violations) at the expense of equality. We bring these results together for a new model of online dialogue quality as systematic norm adherence (e.g., being civil) and ongoing norm repair (e.g., warnings, temporary bans). We use “repair” as a conceptual analogy drawn from conversation analysis, highlighting the ways moderators and users address norm violations through a structured sequence of interaction. Norm repair explains how online dialogue quality is maintained despite recurring norm violations, bringing the prescriptive deliberative model closer to empirical reality.