cross-posted from: https://lemmy.ca/post/72467063

In The Open Society and Its Enemies [53], philosopher Karl Popper introduces the “paradox of tolerance:” unlimited tolerance can undermine itself by allowing intolerance to flourish. This philosophical insight highlights an ongoing challenge for online communities: determining which behaviors to accept and which to reject to maintain inclusive and civil discourse.

Online communities operationalize this principle through content moderation, the process of shaping information exchange and user activity by deciding upon and filtering posted content according to established rules [30, 42, 68]. In centralized social media like Facebook [20] and Reddit [17], previous studies have examined content moderation practices at three levels: platform [17, 28, 55], community [33, 35, 60], and personal [20, 37]. However, these studies predominantly reflect centralized contexts, leaving significant gaps in understanding moderation within decentralized platforms, where even the very notion of a “platform” is contested.

Decentralized systems are built based on “protocols” that define standards for building interoperable digital infrastructure [49]. Consider, for example, Mastodon [4], a widely used open-source software implementation of the ActivityPub protocol [63]. Mastodon is a decentralized alternative to mainstream centralized social media platforms like X. With Mastodon, people can create or join independently operated servers commonly referred to as “instances” (e.g., mastodon.social, mstdn.party), which are part of the “Fediverse.”

The Fediverse refers to the network of thousands of such instances running Mastodon and other ActivityPub implementations. And the process of peering between instances to allow information exchange in the Fediverse is referred to as federation [9]. Instances can interact with one another through the ActivityPub protocol. However, each instance in the Fediverse is independently administered by a user or group who sets local rules and moderation policies, effectively functioning as a self-governed online community. This independence allows communities to reflect local values but also creates challenges around consistency and coordination across instances [9, 21, 32, 34, 56].

A key moderation mechanism available to Mastodon instances to moderate cross-instance interactions is community-level blocklists. This allows the moderators of an instance to block entire instances of users they believe are problematic [14, 19]. These decisions can completely shape the experience of their own users. For example, if the instance users reside in blocks or are blocked by the most popular instances, they would be cut from a broad swathe of content and interactions in the Fediverse.

Prior work has examined blocklists at the personal level — when individuals use them to block other individuals — on centralized platforms, noting their potential for collective action [26] as well as individuals’ contestation of harassment [37]. In comparison, decentralized systems enable community-level blocklists, which both constrain who makes blocking decisions (only the instance admins) and expand the scale of the decision’s impact (everyone in the instance). Prior work already highlights concerns about transparency [40] and democratic decision-making by communities in online spaces [57]. As social media systems like Mastodon become more widely adopted [46, 69], understanding blocklists in this context is crucial for improving content moderation in decentralized settings. However, despite their prominence, little is known about how Mastodon moderators currently manage these blocklists, what tools they use, and what additional support they need.

…

Summary of Results. We found that existing blocklists and tools differ substantially in their purposes, inclusion criteria, transparency, and accountability. Some prioritize broad consensus across instances, while others focus on severity-based or manual curation (RQ1). Through interviews, we identified how moderation styles shape the understanding and usage of blocklists. Different styles embody different trade-offs of how open, safe, and context-specific moderation should be (RQ2). Finally, we identify several avenues for improvement in the design of blocklist tools, including category filters, severity toggles, comment receipts, and collaborative moderation features (RQ3).

A sort of twin of the other paper I just posted about the Fediverse, but focused on blocklists rather than the process of making rules.

All of this is still so new (and kind of really old too) and that makes this kind of work super important. Definitely check this paper out!