I and others have talked a lot about the #Facebook / #Meta#FediVerse issue over the past few days, analyzing their strategy, and possible responses, and why pre-emptive blocking isn't an effective measure.
This leaves the question of "what should we do?" So....
ITT: actually effective measures for building the resilience of the FediVerse and #ActivityPub, informed by the experience of the #OSS movement.
Front-note: I think there's something very close to the difference of views here, so I'll try to articulate what motivates my "side".
Corporations and capital aren't all-powerful dark gods, vampires, demons, or the One Ring. They're systems that can be analyzed, outsmarted, and outmaneuvered.
We have many more options than separatism, and slowing an inexorable advance. We can fundamentally alter the rulebook, and in fact, we have in very significant ways.
For the less technical, the protocols are essentially the "laws of physics" for this space. The key goal is to keep them under the control of #OSS and the #FediVerse.
As currently designed, they prevent the doomsday speculations I've seen floating around (algorithmic curation, surveillance, etc).
There are two aspects of this: governance, and adoption.
Establishing robust test-suites to define common protocols is a time-tested method within the #OSS world for preventing things like #EEE. This is used, among other things, for the Java language and JVM specs, the major implementations of which are controlled by corporations.
OpenJDK is the reference implementation, and is completely open-source. There are other implementations, some closed, some open. The #2 implementation is arguably IBM's.
There is a governance structure, and certification tests in place specifically to keep the implementations mutually-compatible, and prevent any one implementation from pulling an EEE and edging out the others.
An extension of this is to establish a robust governance structure for extending and revising the protocols. These have been successfully employed by the #OSS movement on important parts of the commons, and frequently resist attempts at hegemonization.
As to the question of what sort of governance structures work best, a common model has emerged in #OSS that most closely resembles labor unions:
A tiered structure, but with accountability flowing downward towards the community. Developers/committers/members are appointed based on a clearly-defined process based on experience and domain knowledge, who in turn elect a core team tasked with strategy, but who make non-binding recommendations, not decrees.
Adoption is key in preventing an #XMPP-style de-invention scenario.
Google was able to remove XMPP from the commons by absorbing most of the user base into their own platform, while most external servers fragmented, producing a hub-and-spoke situation. They then just turned off their XMPP integration.
This is also why the "block anyone who doesn't block #Facebook / #Meta" proposal actually accelerates this process.
So we need to ensure that #ActivityPub remains robust and interconnected.
@emc2 your claim here is that the protocol protects against algorithmic influence and surveillance. I'll grant you that it prevents some additional surveillance in the context of federation with meta but it absolutely does not prevent algorithmic influence.
So given that it's your claim here, I have to assume that it's your conclusion elsewhere, meaning, no you haven't
@smallpatatas These comments demonstrate a basic lack of understanding of how this works.
Facebook gains no surveillance abilities by federating that they don't possess today.
Facebook can't force my instance to use their algorithms to serve content, and the long-term moderation infrastructure that is being discussed here will make that both open and customizable.
Yeah so unfortunately that's not how it will work in practice.
If facebook prioritizes certain kinds of content, that stuff will inevitably be what gets shared and commented on more often, and given the size of the potential userbase, they'll effectively be able to set the tone and the topics for the wider network. Plus, you'll find you get more engagement when fb boosts your stuff, so in this way you'll be encouraged to 'dance to their algo' as I like to call it.
In terms of surveillance: at present, your social graph here is separate from whatever social graph fb has on you on their platform. Once federated, they can merge these two data sets, making whatever data they may have grabbed from the Fedi far more valuable.
In addition, due to the fact that you'll often need to click on a profile to view it "on the original page" means fb will quickly be able to link your fedi profile with their wider web surveillance apparatus.
Nothing stops them from seeing all ActivityPub traffic today. Cross-referencing gets them nothing, as they already have the whole data set.
They can prioritize all they want. That will have zero effect on how my instance chooses to run the show, and I control my own followers (and eventually, moderation). Nothing they can do can stop me or my instance from filtering them to our heart's content.
@smallpatatas@emc2 I feel lost, what you just described largely falls under the technical side. Even if we include the social side, you and others are not being open and honest. Based on what you just said, the issue is not a Meta issue. The real issue is a people issue. If users are encouraged by further engagement that is on them and not on Meta
@damon@emc2 The claim was that Meta's algorithms wouldn't make a difference if you're not directly using their platform.
At any rate, the entire point of fb's algorithms is manipulation. Those that think they're somehow immune to manipulation are likely the very people most susceptible.
This is magical thinking. These content algorithms aren't some kind of superintelligence that engineer Rube Goldberg chains of social interactions. They're actually pretty stupid.
(And even if they could, defederation wouldn't stop them)
and they don't need to be rube goldberg superintelligences to have the exact effect I'm describing...
Facebook pushes divisive content? You will see far more divisive content.
Facebook deprioritizes anything to do with Black Lives Matter? You will see far less discussion of it.
No magic to it at all, and it's not the type of thing you can "filter".
And yes, defederation would indeed stop that influence.
Anyway so far I've been called 'bad-faith', 'dishonest', and told that I have no understanding of any of this multiple times. And yet here we are discussing how Facebook's algorithms will most certainly influence the tenor and content of the conversations on the network.
Nope. Once again, Facebook does not determine what accounts I follow (or don't), and they don't control how my instance serves content.
By the way, this "discussion" which consists of you making factually incorrect statements and then being corrected doesn't count as evidence. That's a false controversy fallacy (which is another example of bad faith argumentation)
now consider that anyone you follow on Threads will be served content that Facebook wants them to see. What content do you think they will comment on, boost, or be far more likely to post about?
So I don't follow those people... Again, Facebook cannot control who I do and don't follow, or what my instance does w.r.t. moderation or content serving.
@emc2@damon ok so you're saying you won't follow anyone on Threads in order to keep Facebook's algorithms away from your feed - what about the other people you follow, if they follow lots of people from Threads what do you think they're gonna be talking about? (Keep in mind here that the potential user base dwarfs the fediverse)
The presence of Facebook is going to somehow control the behavior of people who aren't interested in them, independent of any tangible, identifiable mechanism.
Look, I'm not going to chase you around in circles and keep correcting bad facts and fallacious arguments.
the mechanism is that if you are served a lot of content of a particular type or of a particular topic that you will naturally start to behave in a similar way and think about similar things. We are social creatures after all. It's the basis for every political movement, all of culture... if none of it worked then facebook wouldn't use algorithms at all. None of us are incorruptible islands of pure self-assured willpower, it doesn't work like that, sorry to break it to you!
This is an almost…baffling level of seemingly willful ignorance on the part of @emc2; the ability to control your feed isn't the part that makes these algorithms work. Even on FB or Twitter, I can block any user I want; the issue isn't my ability to turn off content from a particular person – it's the people who don't recognize that content as harmful or erroneous, to begin with.
That's what makes the algorithm dangerous: it takes people who may've not known about the subject or are on the fence and presents more polarized "evidence" from others regarding it (because that drives engagement) and that pushes the Overton Window of the discussion.
One person on the Fediverse making an anti-Muslim post (likely) only affects those who follow them and maybe their instance;
if that post makes it to Threads and gets put a lot in the Now Trending timeline on a bunch of Threads instances, maybe a bunch of people on all those instances get to see it; and a larger amount of people who either already agreed or were persuaded now might reblog it; and all the accounts on the non-Threads–Fediverse who followed those Threads accounts now see it. Its reach is further.
Yes – non-Threads–Fediverse instances could, say, defederate with the Threads instances: you're right that FB cannot control how we interact with FB.
But that's assuming that all non-Threads–Fediverse admin.s are as hawkish about catching these things (or that the admin. may be familiar with the subject matter! They may not understand what is dangerous about the content) or
that the content itself is even as extreme as the example we're working with. Not all dangerous content is going to be as blunt as "Kill All Trans People".
What may've been a post dangerous to some people becomes an amplified posts if everyone on the non-Threads–Fediverse doesn't know to block that stuff or can recognize that it's dangerous;
they don't defederate and use the controls they have to manage Threads content – so Threads manipulates the social graph.
That's the core of the problem with the algorithms FB, etc. employs; this is what makes it social and makes being able to stop it not a guarantee that it /is/ stopped.
Removing the red-herring components, you're basically saying "people see content from the people they follow". Ok, sure. So don't follow Facebook people, or people who repost garbage.
This idea that their algorithms are some kind of superintelligence that can manipulate people across multiple hops is magical thinking.
But frankly, I wrote this thread to talk about effective measures for repelling hegemonizing actors, based on principled, factual analysis of how these systems work.
I'm interested in engaging in those discussions, not this. I've spent enough time on these catastrophizing, bad-faith, and factually-wrong arguments already.
"Ok, sure. So don't follow Facebook people" /Almost/ sounds like what the BlockMeta people were on about…
As @smallpatatas already mentioned, there's nothing superintelligent about it; it's literally just spreading the reach of posts; that's all it was with Myanmar – it's not like the thing was orchestrated.
I already explained how being able to block doesn't mean people will (and how we need to look at humans slas social creatures and not those that always make perfect decisions)
in an environment where instances are federated with threads and the threads user base is much larger than the existing fedi, we're hardly talking about 'multiple hops' - we're talking about 'even if you don''t actively follow people on threads you'll be seeing tons of stuff where people are boosting, reacting to, or posting their own takes on whatever their algos are prioritizing.'
It may not even be obvious that what they're talking about is related to those algos. For example, simply pushing culture war stuff detracts from discussion of actually important issues, but defending people from abuse becomes necessary in that environment. There'd be no way to know why you are seeing that content, and people might well have takes that are thoughtful and not 'garbage', but still within the divisive issue framing pushed by a company long known for doing exactly what I'm describing. (It's also something that is very often a right-wing political tactic, since accentuating societal division tends to reduce class solidarity - as we know, many credit fb with the accelerating the rise of Trumpism, if not being a major cause.)
And if you're saying ok so block all those people on the fedi too that are reacting to or boosting threads content? Well, yeah, that's what I'm going to be doing: fully defederating from meta.
Back then in Usenet days, posting quality in Usenet would lower on September when large amounts of students would get access in college. Then AOL plugged in on Internet, and the drop was permanent because it flooded Usenet with low quality discussion.
This will be the same, only they come with a proven increased level of FB-steered toxicity and zero interest on becoming good Fedi citizens.
Filter doesn't mean block, it refers to moderation, which is a thing on every instance. My issue with the preemptive blocking push is its potential to fragment the network; I don't really care what individual instances choose to block, as long as it's not transitive.
(Specifically, I was pointing out factual errors in the other person's arguments, which seemed to presume Facebook would be able to somehow control all instances.)
@emc2@smallpatatas Thank you for saying this. One of the biggest issues I have with these talks in particular #eee is that people are not being open and honest about Facebook currently abilities. Nor are they saying the hard part out loud. This requires adoption. As you’ve stated Meta can’t force any implementations on instances. For the threat level that’s widely pushed here it requires adoption and compliance. Meta can’t force anyone. That’s the discussion that is not happening & it’s doing a disservice
I think the majority of that comes from genuinely not understanding the technology, coupled with a worldview that incorporates a lot more defeatism than a lot of them even realize (see my front-note, "corps aren't all-powerful dark gods")
I've read the entire thread a couple times now, and I don't see a single argument for why we should federate with meta. I don't disagree we should do things like strengthen the protocol or make migration stronger, we should do these things regardless.
The arguments you're making, that the protocol itself is a defense against corporate monopoly don't hold up historically, for Email, Xmpp or IRC, all of these became monopolized despite following open standards.
Even in the case of governance, a billion dollar corporation will naturally be able to stack steering committees or overwhelm git commits to the point where they can influence the overall direction of development in their interest. Imagine Meta makes a billoin dollar donation to Mastodon, and so every user of vanilla mastodon is basically forced to adopt their features.
Meanwhile there are plenty of positive reasons to #BlockMeta namely that all of their platforms are full of hate speech and misinformation.
We have robust mechanisms for countering large corporations stacking committees and such, and mitigations for that kind of influence, with a proven track record of working. I've analyzed the specific examples and why they don't work here.
As for Facebook, if they want to federate, fine. If they expect to be able to make demands about how other instances run things or get special treatment, hell no. They get treated like any other entity, it's that simple.
Well I really don't know how robust those mechanisms you're referring to are, from what I can tell when it comes to code projects like mastodon are surprisingly centralized,
There's also another dimension of governance, on the instance level: who instance admins choose to limit, block, what moderation policies they follow etc.
So if you want to bring up robust governance, a large number of instances choosing to per-emptively choose to block meta is exactly that, that's the strength of the fediverse over anything is the autonomy of instances.
Well I do know, from personal experience, and from having lived through the history of this movement.
To answer your original question, though, I do think that allowing Facebook to federate will lead to them losing users, provide people an offramp from unhealthy platforms to something better, and will ultimately end up having Facebook flounder and flail and burn through a lot of money to no avail.
Is there another historic example of a multi billion dollar corporation moving in putting pressure on Mastodon / AP development before now? I don't see how anyone can say with any certainty how it's going to fair against that kind of pressure. Your faith in your friends and colleagues is admirable and all but the claim that the existing governance structure is incorruptible needs more than personal assurance.
As for the fediverse basically bleeding facebook of users, that could plausibly happen, and it would happen regardless of whether or not a large section of the fediverse blocks them, and having a large section of the fediverse, inaccessible to meta users actually further incentivizes them to migrate.
The OSS movement (which I view the FediVerse being an outgrowth of) has successfully resisted many, many attempts to destroy it. In spite of all that, it has been very successful at decommodification of software.
My concern with preemptive blocking is the call for it to be transitive, as this would fragment the network and could facilitate another XMPP-type outcome. I don't have a problem with who instances block, as long as they don't fragment the network.
So you're saying you don't personally know or have experience with the governance structures in relation to Mastodon / ActivityPub ? Yet you're confident that Facebook is no match for them?
I totally disagree with you about OSS* , Apple's Darwin kernal is 'open source' along with everything Google does, RedHat(IBM) literally just closed off access to it's source code to non paying customers. Microsoft ownes github -- Open Source has been very effectively coopted and is now inseparable from surveillance capitalism. . Don't get me started on the politics relating to the Linux kernel or all the fuckery Amazon does.
On an unrelated note, have you received a suitcase full of cash recently?
*Back in my day we called it FREE and Open Source Software #FOSS
These include post importing, smoother account migration, etc.
If these are fixed, then there's a strong case that the net migration will be out of Facebook's platforms and into the broader FediVerse, away from the ads and everything else bad.
If Facebook is the only one offering these, however, the migration will be towards them.
One significant failure mode for the current #FediVerse is the burden carried by the instance admins. This opens up some key attack vectors, which have been suggested in the rumors of Facebook's terms for providing financial support in exchange for ceding autonomy.
Whether or not these rumors are true, the danger is real and needs to be dealt with.
First thing: SUPPORT YOUR INSTANCE (and others; I support several now)
Going beyond this, there will need to be a more robust infrastructure for funding these sorts of efforts. #FediVerse can be seen as #OSS branching out into domains where we have to contend with things like maintenance costs, moderation, and such.
Some work is being done here, one example being @nivenly
On a longer timeline, I think there is a lot of room for putting things like worker coops, employee ownership, and non-profit models into practice.
The aforementioned #FediLaw efforts to develop legal templates and guidelines are also important. #Facebook may be proposing onerous, non-starter terms, but there may be other situations down the road.
Business models akin to sleepycat's (BerkeleyDB) and others may emerge down the road that are both profitable and actually beneficial to #OSS and #FediVerse. Having a legal framework in place is key.
Lastly, we need to continue to push the envelope on the technical side. We need to challenge the notion that a huge SRE org and cloud infrastructure is essential to these kinds of operations, and develop more decentralized methods.
@abinmn So I'm working (slowly) on a platform based on distributed consensus. What I'd like to do is get to where we can serve a single instance from a consensus cluster run by a number of individuals.
From there, we'd need to do iterative evolution, but I think it's at least prima-facie plausible that we could eventually eliminate much of the need for huge SRE operations.
@abinmn For moderation, you can draw a lot of inspiration from the immune system (makes sense when you think about what the immune system does). I wrote a vision paper on this back in 2018.
BlueSky's labeling servers resemble a portion of this kind of infrastructure, but what they propose would need to be overhauled to be usable within the FediVerse.
Moderation is a huge challenge, and it will also demand technical innovation. Moderation by humans simply doesn't scale. There are technical options, though.
#BlueSky actually has some ideas that are on the right track, but these aren't new (I can personally show prior art from 2018, others can as well). They also need to be re-worked from what they've suggested anyway.
The short version is that moderation infrastructure can be decentralized, helped by automation, and made available as something that can be customized by both users and instances alike.
This eases the burden by eliminating a lot of duplication of labor that goes into moderation.
To be clear, I am not suggesting to simply copy BlueSky's proposal wholesale. There's problems with them, but there's good ideas that can be reworked and redesigned.
If these challenges can be overcome, and effective methods for addressing them can be developed, it will clear much of the road to effectively realizing social change in much broader contexts.
It offers a real and effective way to build a resilient commons, and a real and practical alternative to domination by unaccountable monopolies.
I've seen a few arguments based on the idea that Facebook's algorithm will somehow exert control over the entire network, regardless of follows, through some kind of spooky action-at-a-distance. I suspect these are ultimately rooted in a combination of lack of technical knowledge, plus belief in AI/AGI hype.
Origins aside, there is ongoing work and usable results on the spread of disinfo that the FediVerse can absolutely use. It's not a magical process.
The structure of social graphs, particularly in networks like this is not random. It's well-studied. Similarly, the flow of information and influence through it has been the subject of study, as are countermeasures. I have several colleagues who work on this.
The FediVerse represents an excellent opportunity for more applied R&D to put this into practice at scale, and there is a lot of interest in funding this type of work.
TL;DR version: defeatism, magical thinking, and catastrophizing can always conjure seemingly unbeatable shadows.
But incisive analysis, careful planning, strategic thinking, and a can-do attitude can overcome them. Stick with the hacker mindset; it's gotten us very, very far.
The structure of social graphs, particularly in networks like this is not random. It's well-studied. Similarly, the flow of information and influence through it has been the subject of study, as are countermeasures. I have several colleagues who work on this.
The FediVerse represents an excellent opportunity for more applied R&D to put this into practice at scale, and there is a lot of interest in funding this type of work.
TL;DR version: defeatism, magical thinking, and catastrophizing can always conjure seemingly unbeatable shadows.
But incisive analysis, careful planning, strategic thinking, and a can-do attitude can overcome them. Stick with the hacker mindset; it's gotten us very,very far.
@emc2 not a fault of yours. But it is an indicator how somehow things don't work as they should in grassroots movements.
Many of the same devs who discuss on SocialHub, have lengthy fedi thread discussions, where it is 1-to-readercount and then the knowledge gets to be lost in timeline history.
The social dynamics that make it like that where everyone is focused on their own initiative, things fragment, and there's little effort taken to mitigate, bring cohesion and coordinate, is interesting.
@emc2 I have name-dropped these places literally 1,000's of times as a fedi advocate.
For those promoting the strengthening of the "technology substrate" - people and processes that form the foundation - this dynamic constitutes yet another 'burnout vector' in FOSS (besides the well-known one of maintainers burning out, as the popularity of their project increases and the demand on them from people gets too much).
@emc2 I'm active in Social Coding Movement where inherent unsustainability in Free Software culture is an important topic.
Unsustainable not wrt the software (FOSS is eating the world) but for people who want to make a living with it and spread the culture and values that underlie.
The movement hasn't been formally launched but is very slowly evolving. This because broad-scoped communities that don't address people's incentives directly don't work. Intrinsic motivation is driver to contribute.
So I'll be straight with you, my ideas are R&D efforts, and there's a lot of "we'll figure it out" (I'm a research scientist, it's what I do)
What I'm trying to work towards is something based on distributed consensus and related technologies, where an instance could be spread among multiple people's machines.
There's also a significant anticensorship angle there.
Would love to have ongoing dialogue with people with actual SRE experience, though.
@emc2 Yes, encourage all who join via Meta apps then create an account outside of the Meta ecosystem to use actively. Make it very hard for Meta to later choke off the network of decentralized servers.
@emc2 This has all the elements to understand the power, but using a strawman “block those who block Meta” … “so let’s stay interconnected” misses the danger you outlined quoting from XMPP history.
@emc2 speaking of governance at a recent SWICG meeting @evan mentioned the need for an additional active chair, #Meta has ignored the group in the past, but now have a stake in the game.
A very thoughtful and well-argued strategic reflection. The arrival of Meta/P92 poses a real threat to many core Fediverse tenets; your essay recognises these threats and offers plausible response strategies.
My own leanings thus far have been to ask for defederation from Meta whilst avoiding mutual defederation between existing instances.
Your essay combines a range of proposals that form a more pro-active, developmental approach. Lots to think about!
@the_roamer@emc2 I began arguing for and testing sentiment for defederating commercial instances with the arrival of Medium on fedi.
With Meta moving in, more are waking up to the issue but it is clear that defederating will only divide and undermine the fediverse, which of course helps Meta.
So I'm returning to the thing I've said all along:
fedi is the lifeboat but p2p will be the shore.
Focusing on #Meta et al wastes energy and time. So support #p2p projects such as #SafeNetwork
Add comment