Algorithmic Amplification of Stigma: A Theory of Norm Externalities in Digital Platforms

Digital platforms increasingly mediate public discourse, shaping perceptions of health, identity, and vulnerability. This paper develops a new theoretical framework to explain how engagementmaximizing recommendation algorithms can unintentionally sustain social stigma, even when most users privately prefer supportive discourse. We introduce the concept of a stigma trap, a self-reinforcing equilibrium in which profitable algorithmic choices amplify stigmatizing content, altering perceived norms and reducing social welfare. The model formalizes interactions among strategic content creators, viewers, and the platform, highlighting a novel norm externality through which visibility reshapes social expectations. We show that modest algorithmic interventions, such as credibility thresholds or externality-weighted ranking, can reduce stigma without reducing platform revenue, offering profit-neutral mechanisms for aligning incentives with welfare. Extensions demonstrate that heterogeneous creators, viewer sorting, and platform competition can amplify the stigma trap, and the framework generalizes to other domains where engagement creates social costs. The paper provides a foundation for understanding how algorithmic design can generate or mitigate harmful norms, offering new insights for platform governance and policy.