Open: Where to go after Twitter
As of this morning, Elon Musk was still planning to buy Twitter but many signs suggest he may be getting cold feet. Being against ‘censorship’ sounds well and good if the only material you see being blocked is political. The vast majority of content complaints on every social media platform are directed at men sending pictures of penises to women, at crude personal abuse and at spam.
Musk’s supporters are comparing threats to leave Twitter should their hero take over to threats to leave the country if Trump is re-elected. It is a poor analogy and not just because changing your social media platform is rather easier than changing your country. Like music acts, social media networks have a natural arc. They rise and they fall. In the course of over thirty years I have been a part of dozens of social media communities. Usenet, GeoCities, Myspace; the social media graveyard is large and it has an ample waiting room.
Also, like music acts, new social media networks are being launched every day. Most quickly fold, but some carve out a niche in some otherwise underserved community and a very small number make the big time. TikTok, the latest social network craze was only formed in 2016.
So, for me, the question of ‘where to go after Twitter’ is an easy one: It is time to build a social media platform of our own choosing, one in which we the users are its owners, not the product, one in which we have control. A social media network that is Open.
As some of you will know, I have spent the past three years designing a cryptographic infrastructure, the Mathematical Mesh which (among other things) is designed to secure data at rest. What I have not mentioned in public until now is that one of the applications the Mesh was designed to support from the very start is end-to-end encrypted social media. The current gold standard for Internet security, TLS, only protects data as it travels over the Internet. The content of every post, every comment is visible inside Facebook, Twitter and every other major social media platform today.
The original use case for this capability was to allow groups of security experts to have confidential discussions of threat and breach situations without the risk of making things worse. While some might consider this degree of security unnecessary for sharing pictures of their children, I know several FBI agents working kidnapping cases who would strenuously disagree.
So in the remainder of this note, I will provide a rough outline of the requirements I believe a new social media system needs to meet and an engineering sketch describing what such a system built with Mesh technology could look like.
For now, I am calling the network Open because that is what I want it to be: A shared social media experience that is genuinely open being built on open specifications and services that can be adapted by each user to meet their personal needs. A network that is neither owned nor controlled by any one single company.
Musk isn’t completely wrong about censorship
Let’s start off by admitting that while much of what Musk is saying is naive at best, his central complaint isn’t entirely wrong even if he expresses it incorrectly.
There is something essentially rotten in the extent to which Twitter and Facebook dominate political debate. Agenda control is after all one of the most effective forms power. It is the power that has allowed Rupert Murdoch to impose his malignant and bigoted prejudices on the English speaking world for the past four decades.
While it is true that neither Twitter nor Facebook is a government, both are considerably larger than certain nation states. It is not entirely unreasonable to ask whether companies whose economic and political power comes from their community of users should be to some degree responsible to that community.
We hear the complaints about Facebook censorship from the far right because they still have access to the establishment media, in particular Fox News, their very own 24/7 propaganda station. Rather less mention is made of the fact that Facebook purged over 10,000 leftist/anarchist accounts in August 2020, in the final weeks of the US General election.
Let us be clear, a private corporation exercising its own editorial judgement over the content they allow on their platform is not ‘censorship’ according to the legal definition. But that doesn’t mean it is acceptable.
A world in which discourse is controlled by a small number of billionaires was precisely the problem I joined the Web project at CERN to solve. Murdoch’s boast in 1992 that his newspaper had chosen the Prime Minister of the United Kingdom, a country he is not even a citizen of was an alarm bell for me.
And for a short time, the Web worked as we hoped. The common person did get a chance at a global voice. But then a new wave of plutocrats took over and the decentralized platform became centralized again at a higher layer. And that is why I cannot see Musk buying Twitter as being the solution. He is merely proposing to replace one king with another in the hope that he will be less awful.
So, Musk does have some valid points. But social media isn’t rocket science. It is much harder because unlike rocket science which is merely a matter of applied physics and some chemistry, social media is all about people. And anyone who has followed Musk’s career will recognize that understanding people isn’t exactly Musk’s strong point. And no, social media isn’t like managing people either: Users are not employees, they do not report to you, they do not do what you tell them.
Musk’s failure to anticipate the backlash against his purchase proposal shows that he does not understand that how legacy platforms are captive to their existing communities. Verizon learned this lesson to their cost when they bought Tumblr for $1.1 billion only to shut down the main type of content people used it to share.
Contra Musk, it is not merely ‘the algorithm’ that is at fault, it is the inputs to that algorithm they choose to allow.
Facebook and Twitter both restrict users to voting for content they like. The only way to object to content on either platform is to post a comment in reply. And replying to a poster with a comment tells they algorithm that you want to see more content from that commentator, not less. Click on the Facebook angry emoji to express your disgust with a post and congratulations, you have just given it a ten times super-like because that is how Mr. Zuckerberg rolls.
Having studied particle physics at Oxford, I joined the Facebook group on Quantum Physics. A small amount of the content was interesting, but most were complete and utter nonsense written by people without even a populist understanding of the field. Facebook does not provide me with any means of telling it that I want to see no more content of this type except to block individual idiot users. If I try to explain that a post is so bad it isn’t even wrong, Facebook will interpret that as a desire to see more content of that type.
Without gatekeepers of any kind, the voices people want to listen to are too easily drowned out by those who merely want to be heard. This effect is well known in Russia where the term ‘Infoshun’ is used to describe the constant stream of obvious official lies whose purpose is to drown out actual news. Anyone wondering why Russia would fund Alex Jones and InfoWars should read about this.
Moderation is a very hard problem, not least because what might appear to be perfectly reasonable standards can quickly become ridiculous in a different context.
What could possibly be problematic with a ban on advocating violence? Well, my Facebook account currently has restricted status after I suggested Ukraine should bomb the Crimea bridge, a statement that I consider perfectly reasonable in the context of a discussion of the Russian invasion. I was also notified that a post of mine with a post of Marine LePen meeting with Putin was removed for breaching their ‘Community Standards’.
Facebook was not always like this. Facebook moderation has lurched from one extreme to another as the company has reacted to the latest wave of complaints. A lesson that is quickly learned in any area of security is that every problem has a simple solution provided that it is the only problem you have to address. But one principle has never changed, whenever the company has faced a conflict between profits and user satisfaction, profits have always, always come first.
So here is my offer to Musk: If you want to create the type of social media network I believe you are looking to turn Twitter into, take my design here and deploy and you could do it for $44 million instead of $44 billion. You don’t need to pay me a dime, just pay for the developers, the servers and so on. It might not turn into a globe spanning community of billions like Facebook or Twitter but it will almost certainly include the community you want to participate in: a community of your peers.
The overarching goal of Open is to put the user’s needs first and to put the user in control.
- A user’s Open account, feeds and all the content they post must belong to the user and not some platform provider.
- A user must be able to publish any material they choose, excepting those prohibited in all jurisdictions (e.g. child abuse, copyright). Publishers may refuse to serve specific customers for any reason or for no reason. But a user whose content is refused by one publisher may switch to another.
- Curation and moderation functions should be limited to selecting content in accordance with criteria specified by the user consuming the content. Consumers are free to choose abuse moderation policies that are as restrictive or as permissive as they choose.
- Open must allow users to protect their privacy. While providing rich social media functionality requires access to meta-data including the identity of a poster, link relationships and link semantics, these functions do not require access to the content itself or to the text of annotations.
- The Open ecology must support a range of business models for publishers, aggregators and content curators including paid subscription. Platforms supported by advertising should not be able to invade the privacy of non-users.
It is equally important to consider non-goals.
First and foremost, for me is that an open social media should not be designed to leverage the network effect to establish a monopoly. Nor should the goal of social media be to impose any form of social conformity or control.
One of the tacit assumptions shared by many working in social media is that they should replace the common experience created by network television. In the days before cable, before the VCR, a substantial portion of every industrial nation watched the same TV programs at the same time.
The notion that society was better when everyone thought alike is a paternalist view I reject, not least because the standards to which everyone was expected to conform were invariably sexist, racist and homophobic. Broadcast media imposed conformity and mediocrity. I value diversity and dissent.
Finally, let us remember one of the key design triumphs of the World Wide Web: Co-opting rather than seeking to replace legacy infrastructure.
When I started using the Web in 1992 there was very little content. I am one of the very few people who can claim to have ‘finished surfing the Web’ which I did in a little over three days of night shifts at DESY. But that didn’t matter because the Web browser also provided a gateway to a half dozen information sources that already existed but had woefully bad user interfaces.
In the same way, Open should be designed in such a way that it can leverage existing social media systems from Facebook to Mastodon to NNTP to Mailing Lists. But that does not mean we should proceed by taking an existing system as the starting point. Lotus Symphony showed why that was such a bad idea. The first thing is to work out where we want to aim, working out how legacy systems will fit into this model must be secondary.
This last caution applies to the Web itself. While it is of course a fact that social media is synonymous with Web/2.0, that does not mean that we should start from the assumption that every user is going to interact with social media through a Web browser. If we build the user experience right, modern Web technologies, in particular Web Assembly will allow a first-rate user experience to be provided in a Web Browser. But that does not mean we should begin the design from within the rather restrictive confines of HTTP, HTML and a single Web portal.
Open is built on the technologies provided by the Mathematical Mesh, an open specification that provides strong cryptographic security protections without impacting the user experience. The architecture of Open divides the functions of a traditional social media platform into three separate parts:
Publishers, who perform delivery of user content and annotations.
Curators, who recommend content to a community of readers.
Aggregators, who provide scalability by facilitating communication between the Publishers and the Curators.
This architecture is designed so that the only part of the system that benefits from network effects is the Aggregators whose role is entirely commodified.
Each user selects at least one Publisher and the curators of their choice. Thus, Alice might use the Open Publisher service bundled with her broadband connection to publish all her content but use separate curation services for her personal and work interests.
From other user’s point of view, Alice’s choice of publisher makes absolutely no difference in the same way that the choice of content delivery network makes no difference to a Netflix or a Disney streaming customer. Depending on what Alice uses social media for, she may care to a greater or lesser degree about capacity, bandwidth limits, reliability and all the other things that users care about in selecting a Web Hosting provider. But her choice makes no difference to the people who consume her content.
The curation services are responsible for selecting content for their customers. Such services may adopt whatever business models and whatever content moderation policies they choose. It is to be expected that some services fail while others thrive.
For example, Bob the broker who is paying $2,000 a month to rent his Bloomberg terminal is likely to be very willing to pay $50 a month each to a half dozen curation services bringing him valuable insights that inform his trades. Meanwhile, Carol the collector might run a free curation service focused on the exciting world of collecting Beanie Babies, Swatch watches and NFTs.
It is probable that at least some curation services will offer a free service supported by advertising. But this would be neither the preferred nor the only business model for curation services. And even the users of free services would have no shortage of providers to choose from. Curators who fail to deliver interesting content, who fail to exclude undesirable content, who spam their feeds with too many ads will lose users to curators who serve their users better.
As previously discussed, the notion of giving users control of the content they consume gives many social media executives apoplectic fits. Users will break off into closed epistemological circles of cult members talking only to those of like mind. And I have absolutely no problem with them doing exactly that. What I do have a problem with is the demand that everyone else must be exposed to those ideas. The right to speak does not entail the right to listen.
The fact that Charles Koch has paid to establish a dozen phony ‘Research’ institutions to propagate lies about the impact of his fossil fuel products on climate does not mean that I am obliged to listen. Nor do I have the slightest interest in hearing another word from the man who tried to steal an election he lost by starting a riot and then praised the dictator of Moscow as ‘a genius’ for invading Ukraine.
The fact that I do not wish to give my time to engage with those who argue in bad faith does not make me narrow minded, nor does it make me uninformed. And in the case of some individuals, the extent of the bad faith is as breathtaking. One was calling me a fool, a dupe and a liar for suggesting Putin was amassing tanks on the border to invade Ukraine one day because ‘it will absolutely never happen’ and then claimed to have been ‘proved right all along’ less than three days later when Putin did exactly what I had predicted.
The aggregators reduce the cost of producing a curation feed by monitoring content from a large number of publishers and delivering it to their subscribers.
Thus, when Alice clicks to ‘like’ a post, this causes one IP packet to be sent from her feedreader to her publisher which is in turn bundled with updates from other users and forwarded to a dozen or so aggregation providers which in turn forward filtered and batched data to the curation services.
Besides allowing the system to scale, the aggregation providers provide a limited control point at which the very most egregious forms of abusive content can be tracked and reported. Death threats are not free speech, neither is the rape of a child, nor are copyright violations.
Disney is not going to be able to continue to make movies like Avengers Endgame costing $350 million to make unless they can make at least 351 million back. Protecting their copyrights is both a legitimate and important interest for them. It is quite possible that the recording and movie industries would provide aggregation services to curators at no charge so that they can be sure their copyright content is being protected.
The aggregators represent a control point albeit one that can only govern through mutual consent. If the aggregation services grossly abuse the trust of the curation services, they will be replaced. As is described in the next section, Open content and annotations are published by appending data to an append only log known as a feed. Open Feeds are encoded in the DARE sequence format which uses a Merkle Tree construction to provide integrity and strong encryption for confidentiality. The upshot is that while an aggregator can suppress content, third-party auditors can determine that content was suppressed. Any suppression of content is thus subject to transparency controls.
A fourth service that Open will make use of but is not part of Open itself is the Mesh Callsign Service which provides users with a conveniently short, portable account name that is theirs for life, with no renewal fees for a modest one-time fee (beginning at $0.10 for callsigns of 9 characters or more).
So Alice can register the name @alice and that is hers. Alice can change her Open publisher and/or her Mesh Service Provider as often as she likes, her Mesh callsign will never change. We have had portable telephone numbers for decades now, why do people find the idea of portable email, messaging and social media accounts so desperately difficult to understand?
What type of content should Open deliver? Should it be short messages like Twitter or allow for longer posts like Facebook does? Should the content be limited to text or include video? Should long format articles like Medium be allowed?
The answer of course is all of the above. Legacy social media platforms have attempted to establish a niche by focusing on a limited range of content with a highly restricted range of interactions. Some of these limitations were originally driven by technical limitations. Shipping large numbers of bits was a major challenge in 2004 and so Facebook didn’t support video until much later. TikTok and Snapchat turned the limitations they applied to the type of content they delivered into a means of establishing a new niche in the social media landscape.
An Open social media must necessarily be open to accepting any type of content from text to images to video to spreadsheets and PowerPoint presentations. Why on earth would someone want to participate in a social media centered spreadsheets and PowerPoint presentations? Well, that is pretty much what being in the C-Suite of pretty much any Fortune 500 company consists of, I used to sell IT security technology to those folk. PowerPoint and spreadsheets are very much their bag.
Of course, an important distinction between Facebook and Fortune 500 C-suites is that the latter require confidentiality. Fortunately, Open can be built on the Mathematical Mesh, a Threshold Key Infrastructure that provides all the necessary functionality to provide end-to-end encryption of content with fine grained control and accountability.
Open makes no restriction on the type of content users publish but it does distinguish between primary content and annotations on content.
Primary content and annotations on content are both published by appending entries to an append only-log known as a feed. Users may keep multiple feeds for different purposes.
Primary content is published by writing a file to a Web server and entering a link to that content in one or more feeds. Annotations are short form publications entered directly into the feed itself that are a reaction to either content or another annotation.
Annotations in turn consist of a lightweight semantic describing the annotation type and optional explanatory text.
The power of lightweight semantics was demonstrated by Hurwitz and Mallory and in their seminal Open Meeting run for Vice President Al Gore in 1994. Participants in the open meeting could respond to proposals in the National Performance Review with one of six typed responses. Besides agreeing or disagreeing with the proposal, participants could qualify the proposal or suggest an alternative. They could also raise questions which other participants could answer.
The range of lightweight semantics should be sufficiently compact that users can make use of them without limiting the range of reactions. One of the chief limitations in the Facebook and Twitter curation and moderation systems is that the only reactions users are permitted to express are approval. There is no Facebook reaction that allows a user to state that a post is erroneous nonsense disproven by facts. It is this intentional limitation that has made Facebook in particular such a powerful and profitable instrument of mass disinformation.
The Devil is in the Deployment
As always with new Internet technology, the devil is in the deployment. Proposing new technology is easy, developing it only a little harder. Changing the status quo is the hard part.
And yet, new Internet phenomenon arise every single week and significant, world changing phenomena appear every year. The Internet is the most diverse and dynamic creative force in human history. The Internet can and does change. The only thing the Internet seems to find difficult is standing still.
The current situation is unstable. As Stanley Baldwin said of the press baron Lord Beaverbrook, the social media incumbents exercise power without responsibility, the prerogative of the harlot throughout the ages. (The latter part of the quote being provided by his friend and confidante Rudyard Kipling.)
Getting billions of people to join a social network ecosystem is hard. But Open need not succeed on the same scale as Facebook, or Twitter or even Tumblr to be considered a success. Every social network that has become a global success has begun small and then grown.
The big difference in Open is that although it is designed to harness the network effect to grow, growth of the network does not substantially alter the balance of power between users and providers. Users of Open will always have the opportunity to choose and change their publishers and curators with absolutely no switching costs. As with the Mesh, an Open account belongs to its user and not to the current service provider.
The Web made it possible for anyone to publish (almost) any material they liked without let or hindrance and that is a valuable asset. I enjoy the opportunity that Facebook and Twitter afford to allow me to potentially interact with any of the five billion or so users of the Internet.
But I am also a person who has degrees from two major universities and have held research posts at multiple world class institutions. I would like to have the opportunity to use the Internet to spend just a little time interacting with my peers.
I am clearly not alone in this desire. There is a nightclub in New York City, only open to owners of a ‘Bored Ape Yacht Club’ NFT where the patrons go to eat bananas. The tokens currently sell for upwards of $750,000, a rather astonishing sum to pay just to get past the velvet rope.
All we need for Open to succeed is for Open to find a small number of communities for which it delivers sufficient value for some in those communities to use it some of the time.
We may try and fail but we will certainly fail if we never try.