The Washington PostDemocracy Dies in Darkness

Musk is wrecking speech moderation on Twitter. There’s an alternative.

Twitter alternative Mastodon moderates speech not through banning, but shunning across a broader ‘Fediverse’ of services.

Sites like Mastodon and Tumblr are emerging as new (or renewed) alternatives to Twitter. (Barbara Ortutay/AP)

Elon Musk doesn’t seem to have any obvious theory of how to moderate speech on Twitter, now that he owns it. His rhetoric about free speech has run into the difficulties of managing a service where many users would prefer not to have to deal with certain kinds of speech, and where big companies and famous people would certainly prefer not to have “verified” Twitter accounts that claim to be them, saying things that mess up their carefully cultivated public images.

The chaos of the past week has led many Twitter users to explore alternative services such as Mastodon, a decentralized platform based on ‘open source’ software (which is freely available and can be modified under reasonable conditions). Alan Rozenshtein, an associate professor of law at the University of Minnesota is one of the few people who have done research on how Mastodon moderates speech (his research paper can be found here). I asked him to explain the differences between how Twitter and Mastodon work.

Q: You argue that “the dominance of closed platforms [like Twitter] is an aberration” in the history of the internet. Why so?

A: The internet was designed to have little centralized, top-down control. This is true both in terms of its fundamental architecture, by which the network self-organizes, and the most popular applications that run on it, from email to web browsing. While closed platforms like AOL were popular in the 1990s, they fell out of favor in the early 2000s and it wasn’t until the rise of Facebook and other social media sites that most internet users came to spend the bulk of their online time in “walled gardens.”

Q: Many people are disappointed with how closed platforms like Twitter and Facebook moderate speech. Why do you believe that this disappointment is nearly inevitable?

Large platforms are faced with what we might call the moderator’s trilemma. They are trying to moderate a, one, large and diverse community with, two, a single moderation policy without, three, having a sizable fraction of users be upset. The problem is that it’s not clear that you can satisfy all three parts of this at the same time. A small community with shared values can have a single moderation policy that is generally accepted. But Facebook and Twitter don’t want to be small; they want to connect the entire world. Nor do they want to have an anything-goes platform, because (as Elon Musk is currently learning) neither users nor, more importantly, the advertisers that pay the platform’s bills will tolerate that. So what they’ve de facto accepted is that they’re going to be the punching bag for lots of people, different groups of which think they’re moderating too much, too little, or just the wrong content.

Q: Mastodon, a Twitter alternative that many people are moving toward, is based on an open platform and the ‘Fediverse.’ What’s the difference?

The main difference between Mastodon and Twitter is architectural: Twitter is a closed platform, whereas Mastodon, like all applications built on the ActivityPub protocol that runs the Fediverse, is an open platform. A good analogy to Mastodon is email. There is no single entity that runs email. Rather, email is a system by which different email servers communicate with each other according to a shared protocol. Some of these servers, like Gmail or Outlook, are large and have millions of users, and lots of us choose to use them because they’re convenient. But any organization — indeed any one individual — can set up and run their own email server and participate as part of the email network. And while any given email server can block and refuse to communicate with any other email server, there’s no central authority that can force anyone off the email system entirely.

This is, in essence, how Mastodon works. Mastodon servers are called instances, and, just as one can send emails to users on other email servers, Mastodon users can, using the shared ActivityPub protocol, interact with users on other Mastodon instances. And while a Mastodon instance can choose not to communicate with some other instance (see below for what happened when Gab joined the Mastodon network), there is no central authority that can block a Mastodon user or instance from the network entirely. At the same time, no instance can be forced to host users or content that it doesn’t want to.

An important point is that Mastodon’s decentralization is a matter of technical architecture, rather than policy. We can see this by comparing Mastodon to Reddit, a platform that is often held up as an example of decentralized social media. Reddit operates by setting up a bunch of subreddits, discussion boards that choose their own moderator and moderation policy. But this decentralization is a matter of Reddit choosing to allow it. At any moment, Reddit HQ can reach into a subreddit and moderate content, kick a user off, or even disable an entire subreddit, as occurred when Reddit banned “The_Donald,” a subreddit for Donald Trump fans. Whether you think that Reddit should or should have done that, the point is that Reddit was able to. This is the difference between a platform that is decentralized as a matter of (changeable) policy and one, like Mastodon, that’s decentralized as a matter of (unchangeable) architecture.

Q: How does the balance between ‘voice’ and ‘exit’ in the Fediverse change arguments over the moderation of speech?

With a closed platform like Twitter or Facebook, if you don’t like what the platform is doing, you don’t have a choice other than to threaten to leave the platform. But because being on Twitter and Facebook is valuable for its users, they tend not to leave, and so Twitter and Facebook don’t really have to listen to their users. Mastodon and other Fediverse applications are different. Although you sign up with a particular Mastodon instance, you can move instances and preserve your followers. This puts a check on instances moderating content in a way that their users don’t like, whether that’s moderating too much or too little. It also means that you don’t have to have a single content moderation policy for all instances. Users can sort into instances that better reflect their values and preferences.

Q: What do fights over the relationship between the far-right social media platform Gab and Donald Trump’s Truth Social tell us about how content moderation might work in a more decentralized internet?

Gab, in particular, is an interesting case study, because it shows how the Fediverse can organically self-organize to deal with high-level content moderation disputes. When Gab joined the Mastodon network, this upset a lot of existing Mastodon users, who objected to Gab’s far-right content. But it was immediately clear that Gab could not be removed entirely from Mastodon. As Eugene Rochko, Mastodon’s creator, explained at the time, “You have to understand it’s not actually possible to do anything platform-wide because it’s decentralized. … I don’t have the control.”

But though Rochko didn’t have control, every instance could decide what relationship it wanted to have with Gab. Within days, the major instances had severed ties with Gab, which in turn severed ties with the rest of the network. The outcome was that each community — Gab on the one hand, the rest of the Mastodon on the other — was able to come to equilibrium as to whom it wanted to be associated with. No one was denied their ability to say what they wanted and to whom. And all occurred organically and bottom-up, without centralized control.