agent-based-model

This paper demonstrates that content moderation systems can cause disproportionate harm to bridge users connecting separate communities, even when aggregate accuracy metrics appear satisfactory, with governance loss increasing under false-positive-heavy conditions.