Weeks after Elon Musk took possession of Twitter and proceeded to tweet fast and break things, this Scientific American article by Joe Bak-Coleman offers a general theory of why it’s going so badly: social networks are complex systems rather than complicated ones.

On a social network, interactions between individuals create dizzying feedback loops and chaotic interactions that render simple mathematical models next-to-useless for predicting the future, let alone controlling it. Musk’s gamble is that applying his tinkering philosophy to Twitter will take him where no one has gone before: ownership of a large, healthy and profitable social network. The problem is that, unlike the moon or Mars, we have no idea how to get there—and that’s a challenge that engineering fixes cannot solve.

I saw this linked from Mastodon, where I (and now a pretty large fraction of my former Twitter network) have taken refuge. People there are generally passing it around in the spirit of schadenfreude that imbues most discussion of Twitter in the “Fediverse” distributed social network. However, I don’t see much awareness that what Bak-Coleman identifies as a problem for Musk running Twitter is also a problem for a network of mostly volunteer admins running Fediverse-connected Mastodon instances: we don’t actually know how design and moderation decisions will ultimately add up to create the decentralized social network we’re all using now.

Mastodon’s design is widely described as “anti-viral”, and it certainly makes sense that many of its elements should prevent the rapid global spread of individual memes that we saw on Twitter at its peak. The breakup of the network across independently-managed servers means there’s extra effort needed to transmit a post between “local timelines” on different servers. Mastodon provides a “content warning” (or “subject line”) item for each post, which hides the post behind the text of the content warning by default, requiring other users to click through to see the body of the post. (Users can also set their view to automatically open CW’d posts.) That builds a little friction into the broadcast of potentially infuriating or distressing material. There’s also no option to “quote post”, on Mastodon the way you could on Twitter, forwarding another user’s post with your own commentary appended above it — Mastodon’s creator has said that this absence is very deliberate, because quote tweeting on Twitter facilitated bullying and grandstanding and viral pile-ons. All of these factors have been cited as reasons why Mastodon (at least, before the influx of users from Twitter) was a calmer, gentler social network.

However! All three of these design factors — decentralization, content warnings, and the absence of quote-posting — have emerged as issues in the Fediverse’s growing pains as it absorbs just a fraction of Twitter’s user base.

The semi-segregation of heated discussions created by decentralization means that Fediverse-wide controversies can be harder to understand when they do arise. Over this past weekend, we started seeing reports that instances were de-federating (either silencing or fully blocking) the journa.host instance over its treatment of trans journalist — the administrator of my home instance, ecoevo.social took in what he could see of this and blocked journa.host, cutting ecoevo.social members off from any journa.host accounts they followed. Of a handful of ecoevo.social members acting as a sort of advisory committee on admin and moderation issues, I think I was the only one who’d actually seen the interactions that precipitated the mass defederating and could provide fuller context than a bunch of scattered posts denouncing journa.host — because I happened to follow the three journa.host accounts that were most directly involved. (Parker Molloy, the journalist suspended, has somewhat reconciled with the journa.host admins; ecoevo.social is now muting journa.host rather than fully blocking it. [Correction, 1 Dec 2022: ecoevo.social actually fully re-federated with journa-host.])

Content warnings and quote-posting, meanwhile, have emerged as a different kind of issue: where they were intended to calm the waters of the Fediverse, a lot of new prospective users have found these design choices — and the sometimes instance-specific conventions around their use — to be active barriers to participation. People of color, especially Black folks, arrived on on public Mastodon instances only to be met with instructions from predominantly white veteran Mastodon users that discussions of racism should be covered up with content warnings. The Fediverse is now (maybe) re-learning norms that much of Twitter assimilated years ago, that members of systematically disadvantaged minorities should not have to moderate their account of their experiences for the majority’s comfort. Quote-tweeting may have facilitated bullying pile-ons, but it also allowed for call-and-response iteration of jokes and memes, and helped elevate issues into the broader public eye. It’s a fair question (as the thread I just linked asks) whether Mastodon or another Fediverse application could support the mass awareness-raising and mobilization against injustice that Twitter did, in its heyday.

All of this should tell us that the link between the structure of the Fediverse and its reputation as a kinder, gentler Twitter alternative may not be as straightforward as we’ve assumed. Yes, Mastodon was designed to be a calmer social network experience than Twitter provided — but its early adopters were also people who wanted that experience, and built norms around the structure of the Fediverse and its most prominent application that, they thought, supported their overall goal. We’re now seeing what happens when the structure and existing norms of the Fediverse meet the expectations and needs of a much bigger, more diverse, and more active user base — how we build a new culture of Mastodon will depend as much on the people in the network as it depends on the software we’re using to connect.