Pluralistic Discussions, Graceful Losers? The Role of Throughput Legitimacy for Decision Acceptance in Deliberative Online Citizen Participation

The winner-loser gap is a well-documented phenomenon in electoral studies, according to which electoral losers report lower levels of political support. However, the experience of being part of the majority versus the minority is also relevant to other decision-making processes. Research on direct political participation suggests that the procedure itself can reconcile those who lose. According to procedural fairness theory, decisions resulting from processes with direct citizen involvement may be perceived as fairer and more legitimate than purely representative procedures. From this point of view, expanding citizen participation is proposed as a response to crises of representative democracy, such as decreasing trust in democratic institutions. Yet, several studies call for a more nuanced perspective, as the outcomes of decisions still matter. Instruments such as referendums even seem to have an inherent polarizing component. Less research has been done on deliberative participation formats, which promise to generate legitimacy through an ideally open, equal, and reason-centered exchange of arguments. This paper addresses this gap within the framework of online participation processes. Specifically, it examines how the perceived plurality within discussions varies depending on the AI-based features of an online discussion platform, and how this perception interacts with participants’ outcome evaluations. From a deliberative democratic perspective, plurality is central: perceiving that various viewpoints ― including one’s own ― were voiced and considered can reinforce fairness beliefs and reduce disappointment among losers. Empirically, the study draws on an experiment that combines a three-wave panel survey and a simulated citizens’ assembly, involving participants from Germany (N = 1,552) who discussed one policy issue for ten days and voted on it. Thus, the institutional setup combines a talk-centric deliberative phase followed by a vote-centric aggregation procedure using majority rule. The experiment varied both the discussion topic (euthanasia, sale of alcohol) and the platform design. Participants were randomly assigned to one of five platform versions, including AI-based features that either suggested opposing viewpoints or identified comments with high deliberative quality, corresponding randomized feature versions, and a control condition without additional features. Results show that winners express higher decision acceptance than losers, but this gap narrows when discussions are perceived as exchanges of diverse views. An AI-based function suggesting responses to opposing views tends to increase perceived discussion plurality compared to most other platform versions. While the overall effects are modest, the findings indicate both the potential and the limits of carefully crafted decision procedures to foster acceptance of collective decisions. Co-Authors not ECPR-registered: M. Behrendt, S. Wagner, C. Weinmann, M. Ziegele