Decentralization won't save us


A great deal of drama has occurred around Twitter's CEO in the last week. In addition to politics, the platform has recently pulled even more APIs from public use, severely degrading the user experience for third party clients. Combined with Twitter's outright refusal to provide long-requested features, many have called for leaving the platform altogether. 

The alternative many are looking to as Mastodon, a microblogging service built around the Activitypub standard. Mastodon touts itself as a non-corporate, fully open source, and decentralized social network with a completely open API. On the surface, this looks like a huge win, but the more I think about it -- and social networking in general -- the more I'm convinced it's a flawed alternative. Flawed, not "doomed". I, too hope for Mastodon to displace ad-driven social networks. It's core feature, decentralization, isn't the savior it claims to be.

When I first heard about Mastodon, I was intrigued. It sounded like everything I wanted it to be: Fully open source. Decentralized. No adds or monetization. And no app, just the site.

The last part was what really sold me on the platform. Smartphone applications are no longer humorous distractions, but are not constantly slurping whatever personal data they can, funneling it to external repositories where it can be sold cheaply. What I wanted was an "app-less" social network, preferably one that relied on a Progressive Web App (PWA) instead of something distributed through the Google Play or Apple App Stores. As a web application, access to contact lists, sensors, storage, and so on is restricted via the browser's sandbox as well as the underlying OS's permission model. 

I signed up with the largest Mastodon instance, Eventually, the initial hype wore off, and activity on the platform declined. I started paying less and less attention to Mastodon, and went back to Twitter. Eventually, I used a bridge utility to facilitate cross-posting between my accounts. Mastodon a while. 

I encouraged some friends and colleagues to join, still starry-eyed myself. That's lasted about a week.

Someone I knew was harassed off of the platform shortly after joining. Despite the instance claiming to have robust anti-harassment tools, all of those had failed.

The instance operator's response? Shit happens, we can't do anything.

This happened within one week of opening their account. 

The natural first thought is, "Well, you should try another instance." Moving instances isn't an insurmountable task. Today, import/export tools are even provided to migrate your follows to a new instance. So, the thinking goes, there were horrible people on that instance, maybe the next one will be better?

The problem with this argument is that Mastodon is decentralized. Even if you move to another instance, the same malicious users and harassers can simply follow you to the new instance. They need not join the instance you're on. They need not agree to it's Code of Conduct. They can continue lobbing bile at you by simply finding you once more.

Unlike email, the graph of social networks is often an easily accessible public record. Find one user who isn't your target that knows you target user, and recurse up the chain until you find the new account. It requires no technical acumen to accomplish. And that's what makes it so common and so dangerous.

"So, make your account private!" While that sounds like sensible advice, it also shifts the blame of the harassment onto the person receiving the harassment. You just made the problem worse for them, not better. If the social network is providing the technology to provide reach and broadcasting, it should also provide the tooling necessary to halt that same operation. 

Yet, Mastodon and other Activitypub projects suggest somehow decentralization solves the harassment problem. Even if you move to another instance, harassers can follow you without them having to change accounts or sign up for any new CoCs. This leaves the target of a harassment champaign two real choices:

  1. Move to another instance that is not federated with other instances.
  2. Leave the network entirely.

Again, these "solutions" place the burden on the person experiencing the harassment. They have to remove themselves from an avenue of contact with friends and loved ones. Instead, they move to purely private modes of communication via invite-only chat systems. This causes the additional social injury that you are now denied the reach and broadcast abilities so freely given to those not targeted. 

One topic that comes up often in software design is that of so-called golden hammers. The aphorism goes, "if you have a hammer, everything looks like a nail." If a piece of technology is versatile enough to solve many problems, it becomes the first thing one reaches for to solve nearly every problem. The danger is that having a golden hammer also artificially constrains the developer's options when designing a solution. This can happen -- and often does -- subconsciously. 

The way that decentralization is used by those in the open source community, one gets the sense it's yet another golden hammer. While the git version control system is decentralized, the majority of open source software is still built around Github. Few took alarmists seriously about this until Microsoft bought Github earlier this year.

The thing that worries me greatly about Activitypub is that we're now using decentralization as more than just a solution to technical problems. It's being advertised as a solution for social problems. Harassment and malicious users are a social problem. Claiming that providing basic blocking tools while leaving the rest to the nebulous hope that decentralization will solve everything is a weak prayer at best.

In some ways, the open source nature of Mastodon can work against the federation-as-cure approach. For malicious users with sufficient technical expertise, a nefarious instance could be stood up. This instance could then carry out harassment activities. Furthermore, the instance code could be modified to slurp up personally identifiable information from the federated network. This can be used to facilitate harassment, doxxing, and all manner of nastiness. The federated network would have the option to ban the instance, but this only happens after the fact. It doesn't prevent damaging activity from occurring. Real people are already injured as the cost of that design decision.

All of this prompted me to think about if social networks evolve in a predictable cycle not unlike governments, systems of commerce, and civilizations. After thinking about my own personal experiences with LiveJournal, MySpace, Facebook, Pounce (remember them?), Twitter, Ello, and Tumblr, it really did seem like there's a predictable pattern.

Early stage social networks

When a network is first founded, the network operators genuinely seem to care about helping their users. The product often doesn't work well or completely, so bug reports and feature requests are taken seriously. In so doing the operators hope to build a brand as well as a base of loyal users.

It's also the point in any social network's history where they are the most receptive to backpedaling changes the userbase finds objectionable. If a new feature or change is received poorly, the operators are more likely to rollback deployment. Thus, users have the most power at this stage of a social network's life.

Middle stage social networks

When a social network reaches the middle stage, the operators are more concerned about building a business around their network than fulfilling the needs of their users. This is not to say that the operators aren't receptive to changes, just that user satisfaction is no longer a primary concern. 

A critical mass of users and media attention will prompt network operators to think to the practicalities of operating a technical service in a capitalist society. Hosting bills and developer payroll becomes a concern. In order to fulfill these practicalities, operators start to construct metrics that are convincing to angel investors and other venture capitalists. The name of the game becomes engagement. If your network has a higher engagement than your competitors, you can acquire more attention and thus more money. 

Here is also where most networks turn to advertising to sustain themselves. Having already given away their product for free to end-users, it is nearly impossible to make the service for-pay. Many services create account tiers where paying a subscription gives you additional features or notoriety on the network.

Late stage social networks

By the time you get to the late stage of a social network, there's no more financial issues to solve. Furthermore, the majority of the engineering issues are also solved. The majority of user needs are met, and anything more is either too technically challenging internally (editing tweets, for example), or run counter to business desires (chronological timelines). 

At this point, the network now has so many users and money that many operators start finding social shaping a more interesting problem to solve. Operators buy into their own hype, and start seeing people through the lens of the technology they created. Their own internal biases and prejudices become enforced company doctrine. Given that most network operators are white, male, cis, and straight, most of them have never been the target of the constant harassment and objectification that everyone else experiences.

Often we see a transposition from harassment being a problem to the loudness of harassment targets being an existential threat. Inevitably, the solution is to remove the "squeaky wheels", rather than confront the problem of harassment directly. Particularly in America, responsibility is foisted upon individuals rather than societies, and the most prominent social networks in the West are American creations.


Is there any way users can regain power in this process? In my experience, the answer is no. The only time a network can be rewound to an earlier stage is when it begins the process of collapsing. Users start to leave for other, newer networks. In a panic, the network operators will finally give in to long-demanded features. Not only is this too little, too late, it also infuriates their VCs and business partners. The additional financial and media pressure do the rest. 

Meanwhile, users flocking to new networks see a lot of promise. Often, new networks will take advantage of features denied by incumbents and use them as advertising to gain more users. We silently thing, "This time it will be different", but the process simply starts over again.

The above narrative is heavily focused on American-run, for-profit networks. If there was no need to make money off of a social network, this cycle would break down, wouldn't it?

I really, really wish that were the case, but I think we have already tried that. Forums were and are small-scale social networks. Often they are open source, and do not operate at any sort of profit. About the only difference compared to Mastodon is forums typically were non-federated.

In the few web forums I've joined, the cycle appears to continue, but centered around different motivations. In the early stage, operators are receptive to users and try to build sufficient infrastructure to build their communities. Unless that community reaches a mass scale, there typically no need to seek out more than occasional donations to keep the server running. Social shaping, however, does persist. Too often I've seen web forums grow dogmatic and fossilized around their founding members or figures. Dissent no longer becomes tolerated, and flattery becomes the coin by which protection or vindictiveness is meted out. 

The collapse cycle for forums often happened at a smaller scale as well. Since the forums were self-contained and non-federated, it was easy to sign out, delete your account, and never return again. Sometimes you could drag your friends to a new platform or start a forum of your own. 

Removing the money doesn't really change the overall progress of the cycle, it only changes the motivations and the depth to which the network stays in each stage. 

When I took Intro to Computer Science, a great deal of hype was made of the accomplishments of our field. The moon landing. The Internet. It was feel-good chest thumping meant to encourage young programmers to continue with their chosen field of study. 

When I took Intro to Anthropology, the experience was very different. The second unit was about the atrocities and death the field had caused and abetted.

In all my years at University, I never once heard of the deaths computer science had it's hand in. We didn't talk about how databases were used to mark the progress of genocide. We weren't taught how algorithms can be used to deny people life-saving healthcare. We never thought about how pervasive connectedness could be so destructive.

We as developers often think we can solve everything with just some more code. It's not only naive, but it's also dangerous in the age of mass-scale technology. When unconscious bias can creep into design decisions without examination or second-guessing, the result is a bludgeon against real people with few options other to accept the blows. 

This post was created with the support of my wonderful supporters on Patreon.

If you like this post, consider becoming a supporter at:

Thank you!!!