Sociomeeting: Unveiling emerging moderation dynamics in Mastodon’s federated instance network.

Recently, users have moved from major platforms like Twitter/X and Facebook to Decentralized Online Social Networks (DOSNs) like Mastodon. Mastodon, part of the Fediverse, consists of independent, decentralized servers (instances) that communicate with each other. This shift stems from concerns about content moderation and control on centralized platforms. The rise in Fediverse instances affects network structure and information flow, creating challenges in managing harmful servers. Mastodon offers tools for moderating instance interactions. This study uses Mastodon's APIs to explore moderation dynamics and their impact on information propagation.

The data form a temporal, directed, and signed network, with positive links indicating user follows between instances and negative edges representing blocks by instance owners. Motif analysis shows a consistent trend in network evolution. Evaluating the network's global balance reveals steady dynamics. The banning network splits into two sets: a larger Majority (M) of banned instances and a smaller minority (m) responsible for banning. We propose a model to quantify information flow within and between these groups. Findings show the minority group shares more within itself, while the majority group depends on minority connections for knowledge propagation.

Understanding information transmission in decentralized networks is key to preventing malicious information sources.