Concentric trust circles of collective intelligence

Building collective intelligence from social knowledge graphs

Mathew Lowry


Writing a chapter for a book on Personal Knowledge Graphs made me rethink, and led to a new collective intelligence pilot project with the founders of

This is one in a series of posts derived from my chapter or some of the side-avenues I discovered along the way and couldn’t fit in. Browse all posts.

I originally planned to use the book editors’ deadline to force myself to flesh out detailed user requirements, complete with clickable wireframes. But that was before a co-author invited me to join the Fellowship of the Link (FotL).

I kid you not: that’s actually what we call ourselves, probably only half-ironically. We’re simply a bunch of people interested in collective intelligence, but collectively they taught me something I once knew but somehow forgot: that one of the best ways to think things through and discover new ideas is to simply talk with others with the same interests. Blame the pandemic: I’ve spent too long alone with my ideas.

one of the best ways to think things through and discover new ideas is to simply talk with others

I completely rewrote my chapter as I realised just how many people were already working on different aspects of collective intelligence (TL:DR; the basic tools are taking shape, but they’re not linked together and the users are absent). My chapter now uses as an example of one app in a wider ecosystem, because what really matters is the ecosystem: owned by a data union composed of its users, with revenue streams funding its further development, and thus boosting global collective intelligence.

Collective intelligence

Collective intelligence is commonly understood as a large group of people sharing their knowledge, data and skills to solve common problems. 25 years ago we all thought this would emerge naturally from ever-increasing Web adoption, but our models at the time ignored that collaboration requires more than just information: it requires trust.

Instead, the opposite seems to have happened: we’re effectively stupider as a race than we were when we saved the ozone layer, something we’d probably not manage today despite access to vastly more information.

we’re effectively stupider than when we saved the ozone layer

A lot of people think the rot set in when Google closed its RSS Reader software in 2013. RSS was critical OpenWeb plumbing for underpinning collective intelligence, but although Reader’s closure undoubtedly accelerated the decline, it was a symptom of a deeper problem.

My alarm bells started ringing a few years earlier, as conversations migrated off the decentralised blogosphere and onto centralised walled gardens of surveillance capitalism. The continuous innovation of the Web’s first 10–15 years slowed as the platforms’ grip strengthened and people refocused on maximising their content’s performance on these platforms, rather than publishing their own.

While built on top of open Internet standards, Facebook et al are anything but open, enclosing the vast majority of ideas published today, along with the people who both have them and could use them. Sloshing between this archipelago of walled gardens lies a sea of political disinformation and business-driven crap content, lowering the signal-to-noise to the point where noone can find the solutions we collectively need.

Solving this mess and improving collective intelligence is thus a decentralised knowledge management problem. We’ll need new ideas. The grip of those walled gardens, however, is now a stranglehold on innovation.

Fragmented community, decentralised tools

A lot of people are nevertheless working on solutions. Individually.

It turns out that a lot of people are nevertheless working on solutions, but because collective intelligence is such a multifaceted concept no two solutions look alike. Each person is working on their own innovation in their own corner. Some, like Gordon Brander’s Subconscious or Marc-Antoine’s HyperKnowledge, foresee new protocols, while others (Flancian’s Agora,, IndieWeb) use existing tools and standards as diverse as Webmentions and Git. I’m still trying to get my head around Urbit, which appears to be trying to singlehandedly build an entire ecosystem.

And I hadn’t actually looked at through the specific lens of collective intelligence until the Fellowship conversations, but the basic idea was already there in my first draft introduction:

“The Social Knowledge Graph is thus your personal Knowledge Graph, connected to the Knowledge Graphs of those whose opinion and judgement you respect via a writing-oriented thinking tool”
- From Personal to Social Knowledge Graphs: a vision statement

All solutions have one thing in common: they are decentralised in nature, foreseeing a world where:

  • users publish their own content in their own way using whatever tools they want
  • knowledge interoperability allows this content, published across the web, to be knitted together to help solve problems.

The key problem is that everyone, like me, is focused on their own personal vision, whereas if any of us are to succeed we must work together. After all, our tools must talk to each other.

So it’s beautifully ironic: to overcome the fragmentation separating us we need the collective intelligence toolkit we’re trying to build.

to overcome the fragmentation separating us we need the collective intelligence toolkit we’re trying to build

Bootstrapping upwards

One way forward that emerged from our calls is to bootstrap ourselves upwards via a series of pilot experiments where:

  • the tools we have today are used to link ideas in a specific domain (and the people behind them) together in a decentralised way
  • in the process demonstrating decentralised collective intelligence
  • we learn from the experience to create a new iteration of slightly better tools as a result, along with a slightly bigger audience,
  • rinse, repeat.

I suggested the first experiment should focus on thinking tools, and lo and behold the founders of agreed. We’re currently working on the launch version, to be introduced in my next blog post. If it succeeds, we hope to tackle Pattern Languages in project 2.

Hop over to the Mattermost channel if you want to lend a hand with the project. Otherwise, to get informed when we launch:


This post went live with another, because in parallel with this pilot project I finally took a long overdue plunge into the Fediverse. Decentralised social networks will probably play a key role in the future of and collective intelligence in general, so why not check it out?