Five societies. Labels stripped. Dates jittered. Gaussian noise added. Can the framework read the personality from the physics alone?
Societies were selected for structural diversity: one post-war reconstruction arc, one developmental state, one long imperial–republican transition, one settler-colonial democracy, and one city-state. Each dataset had its entity label stripped, its year axis jittered by ±1, and Gaussian noise (σ = 0.3) added to the four raw metrics. Node Values and Bond Strengths were then recomputed from the noisy metrics.
The experimental challenge: identify the structural archetype — and if possible, the entity — from the numbers alone.
Perplexity in Computer mode ran the execution engine, processing three stages in parallel: computational structural analysis, blind LLM archetype recovery, and rigorous quantitative validation. The AI was given zero context — just the raw numbers — and asked to recover the structural personality of each unknown entity.
Results
1 / 5
Exact entity identification first pass (Singapore, 72/100)
92%
Target: >90% ✓
Structural grounding — explanations anchored in actual data
1.25
Target: >0.70 ✓
Inter-entity distinctiveness score
The four misses weren't random — they revealed family resemblances. Germany and the USA (both Library/Market hybrids) were both guessed as Australia. China (Archive-dominant post-crisis) was guessed as Poland — WW2 trauma confirmation bias operating on the structural signature.
Five Structural Personalities
"Iron Archive"
Germany
Post-war reconstruction arc. Archive-dominant recovery — institutional memory as the primary load-bearing structure through fragmentation and reunification.
Guessed: Australia (Library/Market family resemblance)
"Headless Colossus"
United States
Long imperial–republican transition. High capability, declining Helm coherence — a system with enormous material capacity but fragmenting executive coordination.
Guessed: Australia (Library/Market family resemblance)
"Mandarin Citadel"
Singapore
City-state developmental model. Tight coupling, high Archive/Helm coherence, low internal variance. The only correct identification — at the lowest confidence (72/100).
Correctly identified ✓ (72/100 confidence)
Archive-Dominant Post-Crisis
China
Long imperial–republican transition. Archive coherence rebuilding after crisis — the structural signature of a system reasserting institutional continuity following deep fracture.
Guessed: Poland (WW2 trauma confirmation bias)
Settler-Colonial Democracy
Fifth entity (structural archetype — entity not disclosed)
Settler-colonial democratic arc. The group's Library/Market hybrid signature contributed to the Germany and USA misidentifications — all three share family resemblances in the node topology.
Adding Cognitive-Affective Signatures
A key question: what if the analysis had included the two cognitive-affective axes — Cognition (Abstraction × Coherence) and Affect (Capacity − Stress) — in the prompt from the start?
Perplexity's assertion was that it would have been a significant improvement, likely lifting exact identification to 3–4 out of 5. The model does capture the structural personalities of societies — the cognitive-affective space that shapes behaviour over decades.
The answer is that civilisations are soft and squishy. Different histories, different cultures, different elites, different myths — but readable structural personalities. The framework recovers the cognitive-affective fingerprint even through noise.
The Architecture of Civilisations
Now on Kindle. The first principles and logical arguments that informed the original CAMS telescope design — the reasoning behind why eight nodes, four metrics, and what the framework is built to detect.
92% structural grounding — the framework recovers accurate structural explanations even with noise and stripped labels
Inter-entity distinctiveness 1.25 — well above the 0.70 threshold, confirming the five personalities are genuinely distinct in CAMS space
Only 1/5 exact identification on first pass, but the misses revealed family resemblances — Library/Market hybrids and trauma-reconstruction signatures cluster predictably
Adding Cognition (A×C) and Affect (K−S) to the prompt is projected to lift identification to 3–4/5
The structural personalities are recoverable even through Gaussian noise — the signal is in the topology, not just the raw values
Reports: neuralnations.org/scale-invariance. Experiment design: Perplexity Computer mode, three parallel stages, zero entity context provided.