Occam's Razor and the Anatomy of Civilisations

It began with a conversation with GPT that became unexpectedly productive. I was looking to find evidence that human societies were dangerously indeterminate, so as to strengthen arguments against nuclear weapons.

The eight nodes and four metrics were originally chosen because they were the smallest set that could credibly map a complex adaptive system onto a society without violating first principles. That was just Occam's Razor applied.

Once you take seriously the idea that societies are complex adaptive systems, the thermodynamic grounding almost forces itself on you. Systems that coordinate work, process information, store memory, and dissipate energy become classical objects. Coherence, capacity, stress, and abstraction become the state variables that describe how any human system can remain far from equilibrium without tearing itself apart.

Every society — whether it's the Roman Empire, modern China, or Australia — has to coordinate decisions, defend itself, generate and enforce legitimacy, store and allocate wealth, maintain skilled production, sustain mass labour and reproduction, preserve institutional memory, and move goods and information.

Eight are required because if you merge cultural meaning with administrative memory, you can't tell whether a society is losing its shared reality or its capacity to remember. If you merge skilled craft with mass labour, you erase the pattern where the technically competent parts of a society are functioning while the labour substrate is overheating. If you merge exchange with wealth storage, you miss how commerce can become increasingly abstract and self-referential while the allocative core loses coherence.

Eight is the minimum resolution that can model the tensions we recognise as history: centre versus periphery, capital versus labour, security versus production, narrative versus lived experience, complexity versus coherence. Eight nodes keep the anatomy intact.

The spooky part was how well it worked. Coherence and abstraction became a snapshot of a cognitive axis: how the society "thinks," and whether its thinking is grounded or floating above fragmentation. Capacity and stress became the metabolic axis: whether the society has surplus, or whether it is running hot and reactive.

This taxonomy became a formulation when the nodes became a network. Civilisations look fine right up until the moment their couplings fail — the security function decouples from the coordinating centre, the wealth-holding class detaches from the reproduction of labour, the narrative layer becomes performative, the bureaucratic memory becomes too expensive to maintain and begins to rot. When those bonds weaken, shock propagates.

With eight nodes, the number of possible pairwise relations is large enough for the bond structure to describe a genuine system property and small enough that a human mind can grasp it.

At some point, the model stopped feeling like something we were inventing and started being something we were uncovering. You can still change the labels, tweak the definitions, modify the scoring method; the underlying structures are still seen.

The luck is that this minimal, testable structure appeared that we could use to start treating the problem of history scientifically. Once you have this instrument, you can do the thing historians rarely get to do: you can score, compare, plot, look for invariants, and check whether the same patterns recur across cultures and eras. You can test whether what you are seeing is merely a story you like, or a signature that persists even when you would prefer it not to.

Over the months that followed, the model repeatedly highlighted phase changes the way good physical models do. It flagged the difference between integrated sophistication and brittle sophistication — between high abstraction grounded in coherence, and high abstraction floating above fracture. It distinguished societies that were stable because they had surplus from societies that looked stable because they were dissociated. It clarified why some crises are sticky and hard to reverse: once reactive mode takes over, the system is no longer free to deliberate itself back to health. It must first regain metabolic margins and rebuild couplings.

If there is a moral to the origin story, it is that structured dialogue with AI can act as a kind of modelling wind tunnel. You can iterate quickly, challenge your own assumptions, and keep asking the same hard question: does the structure survive contact with reality? In that sense, GPT was a partner in disciplined reduction — a tool for repeatedly stripping away the unworkable until what remained was functional.

CAMS suggests that civilisations, for all their messy surface detail, have a discoverable anatomy. That a small set of state variables govern their phase behaviour. History is not a narrative; it is dynamics. That does not make the future predetermined. It makes it legible.

The eight-node, four-axis model enables us to see societies clearly enough to argue honestly about them. It produces results that useful instruments do: stable signals, recurring patterns, and warnings you don't get from anecdotes.

It is the kind of solution that arrives when you have been staring at a problem long enough to recognise a usable simplicity when it finally appears.