Connection weights
Earlier iterations of the Village Link project envisaged the construction of social graphs that have relationships that act like the connections in a neural net - with weights attached to each connection. The concept was that these weightings would then become part of the village thinking machine.
As the project has matured a bit, it has become obvious that while actors may choose to put weights on their private, and maybe even their public assessments of relationship strengths, this is not in scope for the project. Such a weighting system, if it evolved, is something that the standard would support, but which the project does not have to build.
This could be stated in a stronger way: A neural net system that runs between actors is an example of a set of norms. If the Village Link project built this thing, it would be getting in the road of villages doing what they are designed to do. In theory, a village could develop a set of norms that is even more sophisticated than a neural net, and the job of the project is to stay out of the damn way.
Earlier writing ...
One of the possibilities of this project is that we will uncover a way that communities 'think' that is similar to the way that brains think. The Architecture section of the project readme page refers to this possibility.
The core idea is that the architecture of both brains and villages have processing units, connections, and connection strengths. In brains, the processing units are neurons. In villages, the processing units are people.
In both cases, thinking and learning is a process of ingesting new information and pushing it through the existing channels. There are feedback loops and reward functions such that the new information might change connection weights, form new connections, or sever connections that have passed their use-buy date.
In terms of the programming task, this suggests that connections and connection weights should be quite central to the types of reputational claims that our agents make. We can imagine each agent maintaining a list of connections to other agents. Each connection could have a connection weight ... and also a set of rules about which other agents that information will be published to ...
... and even what spin might be put on that information for different audiences. We are good at this stuff. The trick will be to build functionality where prestige strategies can evolve without bound.
In conversation with Steve Vitka (of Kwaai and the Decentralized Trust Graph Working Group) I found myself, on the fly, reaching for an analogy to exemplify that fact that it might not be just connection weights that matter, but also the first derivative of connection weights with respect to time. The example I landed on was this:
Consider the relationship, between the US and the UK under Reagan and Thatcher in the 1980s. The relationship was strong and getting stronger. Scroll forward to the same relationship in 2026 under Trump and Starmer. The relationship is strong and getting weaker.
The strategic implications are very different.
I concocted the example to make the point about the first derivative ...
... but it also exemplifies something that I had perhaps not previously noticed: The damn relationship does not exist in any formal or verifiable sense. It doesn't have a DID, and it's never going to get one. Our graph won't have an object for that relationship in its architecture (let alone a object for the relationship strength, or; god forbid, an object for the first derivative of the relationship strength).
Instead, what is exists is the private assessment that each stakeholder makes about that relationship. China, Russia, Ukraine, Greenland, Denmark, Iran ... all care deeply about that relationship, and are all frantically making private assessments. In doing this they can employ whatever strategies they fancy, (including the taking of first derivatives, if that's they way they roll). Then, of course, they all have desired outcomes in a world where that edge is important ... and they can all make reputational claims about the edge and the two parties.
In this view, the things that exist are (1) the private assessments and (2) the reputational claims.
In this sense, even 'Greenland' does not really exist. It does not have a DID.
'Greenland' is a plausible fantasy ... made more plausible by the reputational claims of 60,000 people who say, "I am a Grønlænder."
