Alternative Social-Media Platforms

Posted 6/12/20

This post is an informal discussion related to a research project I’ve been involved in and the broader context of alternative media platforms, deplatforming, and community building. This is not peer-reviewed academic work; see the paper for that.

Background

Mainstream social media platforms (Facebook, Twitter, YouTube) have been under public pressure to limit hatespeech and hate groups active in their communities, especially white supremacist and fascist groups. Most of their response has involved deplatforming (banning users) and demonetization (disabling advertisements for content so users can’t profit off of questionable speech). This is usually seen as a broadly good thing because it cleans up the community and removes the most dangerous and toxic members. Most of the debate circles around which users should be deplatformed or demonetized and whether platforms are doing enough, and why platforms are disincentivized from acting, and the odd role of cultural censor this puts private companies in.

However, deplatformed users don’t simply disappear. Those people still exist, and still want to produce content and regain their source of income. So where do they go when they’re exiled from mainstream social media?

Unsurprisingly, banned alt-right users have flocked to a number of alternative social media platforms. Minds stepped into Facebook’s shoes, Voat is a clear clone of Reddit, Gab stands in for Twitter, and BitChute serves as alternate YouTube. Importantly, these platforms don’t advertise as being havens for the alt-right or neo-nazis; they all self-describe as bastions of free speech that take a hands-off approach to moderation. Whether this is a cover story or the platforms are innocent and have been co-opted by exiled horrible people is up for some debate, but doesn’t change the current state of their communities.

Alternative communities face the same user-acquisition problems the mainstream platforms had when they were young: The point of social-media is to be social, and you can’t build a social community without a lot of people to talk to. The network effect is important here; most alternative social networks will fizzle out quickly because nothing interesting is happening there (and we’ve seen that most Reddit clones have stuttered to a halt within a year or so), but successful social networks will attract more and more people, creating a more vibrant community and attracting even more users, amplifying the effect in a feedback loop. Centralization is natural here. It’s a basin of attraction if you want to use system science terminology.

External influence is also important for selecting which alternative communities will thrive. When a major celebrity like InfoWars is banned from YouTube, they carry their substantial following with them, and whatever platform they land on will receive an explosion of accounts and activity.

The Radicalization Problem

On mainstream platforms the alt-right needed to take a careful stance. They want to express their ideas to recruit others, but have to “behave” and stay within acceptable language for the platform and self-censor. When a popular post contains something hateful, the comments are filled with detractors explaining why they’re wrong and offering countering-views.

On alternative platforms these limitations vanish. Content producers can advocate for race war or genocide or fascist dictatorship or whatever flavor of abhorrent views they hold, without repercussions from the platform. For the most part the only people that join the alternate platform were either banned from mainstream platforms or followed content producers that were, creating a magnificent echo-chamber. Because of the network effect, these users converge to a handful of alternative platforms and all meet one another. Under group-grid theory these platforms would be classified as enclaves - not much social hierarchy, but a clear sense of in-group and out-group and shared attitudes.

Obviously, the content producers on alternative platforms have a smaller reach than on mainstream platforms. However, the intensity of their rhetoric increases dramatically, and so, perhaps, does the threat of radicalization. Deplatforming hateful users from big platforms “cleans up” those platforms, but does it potentially fuel violence by cutting the audience off from counter-viewpoints?

The role of alternative social platforms in radicalization is difficult to measure, and is confounded by other communities like image boards and Discord and Telegram groups. What we know is that incidences of alt-right violence are increasing, and many shooters are active on alt-right media platforms, of either the more private Telegram and 8Chan variety or the more public BitChute flavor. What we can say most confidently is “there may be dangerous unintended consequences of deplatforming that should be investigated.”

Solutions?

If deplatforming is dangerous, what alternatives exist?

More Deplatforming

An obvious answer is to increase deplatforming. If we pressured Twitter and Facebook to deplatform harmful users, then we can pressure hosting and DNS providers to delist harmful communities. This has precedence; after a terrible series of shootings by 8Chan members, Cloudflare terminated service for the 8Chan image board. The board is back online, but only after finding a more niche domain registrar and hosting provider Epik, known for supporting far-right sites. Epik was in turn shut down by their own backend hosting provider, who wanted nothing to do with 8Chan, and only came back online after Epik agreed to only provide DNS services for 8Chan. The site is now hosted by a Russian service provider.

This highlights both the successes and limitations of deplatforming. Through collective agreement that a site like 8chan is abhorrent we can pressure companies to stop cooperating, and we can make it very difficult for such communities to operate. However, once a site is committed to using alternative services to stay operational, they can move to increasingly “alternative” services until they find someone willing to take their money, or someone with resources and an agreeable ideology. Deplatforming pushes them away, but they always have somewhere further to go.

De-recommending

The opposite strategies is to let the alt-right remain on mainstream platforms, but find alternative means to limit their influence and disperse their audience. A key piece of this strategy is recommendation algorithms, responsible for selecting similar YouTube videos, relevant search results, and prioritizing content on a feed. These algorithms can be amended to lower the relevance of alt-right content, making it less likely to be stumbled upon and suggested to fewer people. If the content producers still have a voice on the mainstream platforms then they will be disinclined to leave for a small alternative soapbox with a miniscule audience, and they may not even know that their content is de-prioritized rather than unpopular.

An important consideration: Changes to recommendation algorithms are fraught with challenges, and place more authority in the hands of media platforms, who would be increasingly responsible for shaping culture through mysterious and unobserved means.

Counter-Suggestions

Social Media platforms have been unusually quick to combat misinformation about COVID-19 during the ongoing pandemic. At any mention of COVID, YouTube includes links to CDC and World Health websites with reliable information about the state of the disease. This protocol could be expanded, linking to the Southern Poverty Law Center or Anti-Defamation League or other positive influences at mentions of hate trigger-phrases.

Is this strategy effective? Does it combat hateful views as well as misinformation? Could countering misinformation help prevent the formation of some hateful views to begin with? This is an ongoing research area. One benefit of this strategy is that it can be deployed widely; attaching an SPLC link just below a video title does not deplatform the uploader, and does not need to carry the same weight as making decisions about censorship.

Federating

Both de-recommending and counter-suggestions place more authority in the hands of platforms and see them as the arbiters who decide which cultural problems must be addressed. Detractors to this idea regularly suggest moving to decentralized platforms like Mastodon and Diaspora. In federated social networks, users have a “home” on a particular server, which has its own rules of conduct and permitted content. Servers can interact with one another, agreeing to bridge content between servers and further its reach (a kind of dynamic collaboration loosely reminiscent of connections between Pursuances). In theory this provides a more organic solution to censorship, where everyone is allowed to exist on the federated platform, but if their content is unsightly then it won’t propagate far.

Unfortunately, in practice these federated platforms have suffered from low membership and high centralization, both as a consequence of the network effect. Projects like Mastodon and Diaspora have good intentions and intriguing designs, but without a large community they cannot attract members from Twitter and Facebook, and so mainstream platforms remain viable recruiting grounds for alt-right spaces. Further, running your own federated server within one of these platforms suffers from the same network effect, and frequently leads to centralization on a small number of servers. I wrote about this very briefly a few years ago, and the problem has persisted, to the point that more than half of Mastodon’s users are on three federated instances.

Almost exactly one year ago we witnessed a case study in how federated platforms can handle censorship, when Gab rebuilt itself as a Mastodon instance. The response was positive: Most server operators promptly blocked Gab, isolating its instance on the network. Gab can use the Mastodon code, but not its community. This suggests that federated social networks could be an effective solution; if they can grow their populations.