BlueSky is cosplaying decentralization

Created
Fri, 28/04/2023 - 05:15
Updated
Fri, 28/04/2023 - 05:15

Almost exactly six months after Twitter got taken over by a petulant edge lord, people seem to be done with grieving the communities this disrupted and connections they lost, and are ready, eager even, to jump head-first into another toxic relationship. This time with BlueSky.

BlueSky’s faux-decentralization

BlueSky differentiates itself from Hive, Post, and other centralized social media newcommers by being ostensibly decentralized. It differentiates itself from the Fediverse by not being the Fediverse, and by being funded by *checks notes* Twitter. Oh, and by being built by Silicon Valley techbros, instead of weirdos who understand consent and how important moderation is.

I say “ostensibly decentralized”, because BlueSky’s (henceforth referred to as “BS” here) decentralization is a similar kind of decentralization as with cryptocurrencies: sure, you can run your own node (in BS case: “personal data servers”), but that does not give you basically any meaningful agency in the system. Quoting the protocol docs:

Account portability is the major reason why we chose to build a separate protocol. We consider portability to be crucial because it protects users from sudden bans, server shutdowns, and policy disagreements.

And here:

ATP’s model is that speech and reach should be two separate layers, built to work with each other. The “speech” layer should remain neutral, distributing authority and designed to ensure everyone has a voice. The “reach” layer lives on top, built for flexibility and designed to scale.

So the storage layer is “neutral”, accounts are “portable”. That to me means that node operators will have no agency in the system. Discoverability/search/recommendations are done in a separate layer, and the way the system seems to be designed (nodes have no say, they just provide the data) effectively places all the power with these “reach” algorithms.

Secondary centralization in “reach” layer

The rule of thumb with search and recommendation algorithms is: the bigger, the better. The more data you have and the more compute you get to throw at it, the better your recommendations will be. So it’s a winner-takes-all system that strongly avantages whoever starts building their dataset early and can throw as much money at it as possible.

And once you’re the biggest game in town, people will optimize for you (just look at SEO and Google Search). It won’t matter much that people using the network can freely choose a different algorithm, just as it doesn’t matter much on the Web that people can choose a different search engine. And the more I read about BS’s protocol, the more I think this is done on purpose.

Why? Because it allows BS to pay lip service to decentralization, without actually giving away the power in the system. After all, BlueSky-the-company will definitely be the first to start indexing BS-the-social-network posts, and you can bet Jack has enough money to throw at this to get the needed compute. I guess decentralization is a big thing lately and there are investors to scam if you can farm enough users and build enough hype fast enough!

Another pretty good sign that BS’s decentralization is actually b.s. is the fact that the Decentralized Identifiers (DIDs) used by BlueSky are currently “temporarily” not actually decentralized. The protocol uses something imaginatively called “DID Placeholder”. If I were a betting man I would bet that in five years it will keep on using the centralized DID Placeholder, and that that will be a root cause of a lot of shenanigans.

Externalizing the work

Finally, as a good friend of mine, tomasino, noticed:

it decentralizes the cost to the central authority by pushing data load onto volunteers

A similar observation was made by mekka okoreke, too. To which I can only add: very much this, while planning to keep control by being the biggest kid on the “reach” block.

Of course, fedi could also have some search and discovery algorithms built on top. Operators of such algorithms (there had been a few attempts already) would also benefit from being first and going big. But their potential power is balanced by the power fedi instance admins and moderators have (blocking and defederating) and by the fact that fedi is perfectly usable without such algorithms. And by strong hostility of a lot of people using fedi towards non-consensual indexing.

Jack’s BS

BS is the brainchild of Jack Dorsey, which is no surprise to anyone who’s been paying any attention to BS. Jack Dorsey is of course the former CEO of Twitter, who famously said:

Elon is the singular solution I trust. I trust his mission to extend the light of consciousness.

This aged roughly as well as fresh milk out in the midday July sun in Portugal.

Jack also heavily promoted cryptocurrencies, scammed people using NFTs, and donated a bunch of BTC to Nostr, a “censorship-resistant” social media platform, because of course.

And finally, there’s this comment of his (posted on Nostr; BlueSky not good enough for Jack, it seems). Crucial bit:

Likes are superficial and exist only to inform an algorithm. Relevance algorithms have their place, but they are best informed by a truly costly action.

No, you stockholder-value-optimizing-robot, likes exist to inform the author that you liked their post. They exist to infuse some warm emotions into the cold machine. They exist so that we can connect on a human level without trivializing it by putting into words. You know, as us humans do.

With all this considered, let’s just say I question Jack’s judgement and his motives in anything related to social networks. And since, as I said, BS is his brainchild, I would be very suspicious of it.

Modeled after Twitter

In a pretty meaningful way, “speech and reach” is the model of Twitter today. You just don’t get to choose your recommendation/discovery algorithm.

Elon Musk, the self-described “free speech absolutist” (unless it’s criticism of him) has re-platformed a lot of nasty people with the idea that anyone should have a Twitter account. But only those who pay get to play with any meaningful reach.

What actual difference would being able to choose between different recommendation/discoverability algorithms make for at-risk folks who are constantly harassed on Twitter? There is no way to opt-out from “reach” algorithms indexing one’s posts, as far as I can see in the ATproto and BS documentation. So fash/harassers would be able to choose an algorithm that basically recommends targets to them.

On the other hand, harassment victims could choose an algo that does not recommend harassers to them — but the problem for them is not that they are recommended to follow harassers’ accounts. It’s that harassers get to jump into their replies and pile-on using quote-posts and so on. Aided and abetted by recommendation algorithms that one cannot opt out of being indexed by in order to protect oneself.

The only way to effectively fight harassment in a social network is effective, contextual moderation. The Fediverse showed that having communities, which embody that context and whose admins and moderators focus on protecting their members, is pretty damn effective here. This is exactly what BS is not doing. And I do not see much mention of moderation at all in its documentation.

In other words, “neutrality” and “speech” and “voice” and “protection from bans” is mentioned right there, front and center, in BS’s overview and FAQ. At the same time moderation and anti-harassment features are, at best, an afterthought. As fedi user dr2chase put it:

I’m getting a techno-Libertarian aroma from all this, i.e., these guys won’t kick the Nazi out of the bar.

People like shiny!

Of course the sad reality is that people will buy the hype, build communities under the everloving watchful eye of Jack “Musk is the singular solution I trust, likes are superficial if not paid for” Dorsey. And then do a surprised picachu face when inevitably, sooner or later, some surveillance capitalist robber baron enshittifies it to a point of complete unusefulness.

It fascinates me how quickly people forget lessons from the whole Twitter kerfuffle, and just fall for another Silicon Valley silly con. Without even skipping a beat.