Papers & Research

The master essay — a full synthesis of what CAMS has found, where the framework stands, and how the pieces fit together.

A free book and 30-second film explaining the whole project in plain terms. Every organisation as an adaptive human system. CC BY 4.0.

Honest summary: a thermometer, not a crystal ball. What the instrument can and cannot do, stated plainly before you read anything else.

How the instrument was built — wrong turns included. The node names, metrics, and categories that changed along the way.

What the framework claims and doesn't claim — and the specific tests that would prove it wrong.

Why dismissal without engagement isn't refutation — and the pre-registered test (Spearman ρ = 0.70) that would have killed the project if it had failed.

The authoritative definitions of all four metrics (Coherence, Capacity, Stress, Abstraction), Node Value, and every derived composite used across the project.

The case for why exactly these eight functions — not seven or ten — are the right partition for systems of this scale.

A theory of societal pattern, coordination, and collapse. The structural argument for why societies follow recurring templates.

15 predictions sealed before scoring for China and 7 Chinese firms (2021–2024). Result: 6 passed, 9 failed. The kill-switch was not triggered. Written for readers new to CAMS.

The sealed crisis-detection rule applied unchanged to Boeing, Theranos and WorldCom against six matched controls. Interim results — baselines pending.

Blind era-recognition test: could an AI identify a masked society from its metric signature alone? Exploratory and promising — not causal validation, not peer review.

A sealed rule for distinguishing epistemological crises (failure of knowledge and governance) from economic crises (failure of labour and resources). 87% accuracy on canonical data.

The functional attractor coherence study re-run on the full 53-society dataset. A consistency check, not a blind test.

Full panel analysis by Kari McKern and Fable 5 (Anthropic): blinded society classification, scoring, and morphospace placement. Includes PDF.

3 blind sets, 24 societies, mean reciprocal rank improved from 0.667 to 0.833. Calibration gate failure documented. With figures and data.

20 ancient societies from Longshan to the Xiongnu, 22 blind passes from Claude, Grok and GPT. Five pre-registered hypotheses; collapse drains every node at once.

JUNO experiment reveal: 8 masked societies unmasked (Germany, Norway, Russia, UK, Chile, Poland, Argentina, Australia). 1,201 society-years. Core formulas validated at machine precision.

Variance reduction = 1/√N exactly; five independent scorers produce the optimal estimate. Germany and USA 1880–2026. Uncertainty reflects genuine ambiguity, not noise.

Six-part validation: corpus audit, formalism repair, vantage-point bias control, and pre-registered 2027–2028 predictions. 2,532 society-years, 25 entities.

Principal component analysis across 7,880 node-year observations; one factor explains 72.7% of variance. 17 crisis fingerprints matched. 12 falsifiable predictions registered.

36 societies, 32,361 node-years. The bond formula proven algebraically. Stewards confirmed as a leading political-economy indicator (AUC 0.930).

Do different AI models score the same societies the same way? Reproducible coordination signatures across independent AI assessors.

The preserved, verified record of the GPT conversation that produced the book. All 13 files match their hashes. Blind-rater agreement r 0.89–0.93. Zenodo DOI: 10.5281/zenodo.23793977.

Stress across nations became dramatically more synchronised after 2000: one common factor now explains nearly half the global variance. The 1920s–30s parallel, with the debt sign inverted.

Six recurring failure patterns on 37 societies. Late Abstraction Collapse (when memory and meaning-making detach) precedes major falls with p = 0.0002.

44 societies, 4,541 society-years, 1900–2025. 126 coups and crises split into two types. Leave-one-country-out accuracy: 74% for deficit crises.

Full academic paper with Methods/Results/Discussion, adding cultural-grounding path-type results. Japan and Russia reclassified. 4 figures.

Six findings from scoring 238 societies: fixation capacity (the Archive node) correlates r = 0.976 with long-run viability. Four civilisation types as strategies, not a ladder.

229 polities. Coordination problems, geography, and the eight-node architecture. Why different societies face different structural constraints.

Germany 1910–1930 and Late Rome 380–450 CE vs a UK control. When the wealth-stewardship floor breaks, collapse follows — a structural early warning.

Australia and Argentina, 229 society-years. The emotional signal (capacity minus stress) is empirically robust; the cognitive signal (clarity times coherence) is not yet load-bearing.

9-firm corporate panel plus Denmark and France: the same derived metrics work structurally across nations and firms, but the parameters vary. Confirmed by two independent computational audits.

USA 1995–2026. The GFC rescued financial stewardship in two years; the workforce took six. How disaffection became political coordination in 2016 — and again in 2024. 16 charts.

China, Russia, and USA each plotted against their own 20th-century baseline. USA is the outlier: viability and coupling both outside the historical envelope. z = −2.73.

1996–2025. CAMS Stress leads the collapse in Gallup and Pew public trust surveys by 2–3 years — the structural signal comes before the polling does.

States as pods adapted to river valley, coast, continent, and steppe. Why every bloc is sincere and every bloc projects. Simultaneous constriction, not raw power, as the warning sign.

176 annual observations, 1850–2025. 1944 viability trough. Long-run high Archive and Flow; historically low Hands. Low 2025 Helm with critical reflection on recent institutional pressure.

125 years of Australian threat discourse in Trove and Hansard. Three regimes of threat framing; workforce grievances lead security mobilisation.

Five essays on 175 years of Australian strategic dependency: colonial architecture, foreign-owned stewardship, the Anxiety Machine's findings, the trade paradox, and a 2014 counterfactual.

Fertility discourse and institutional stress in Australia 1900–2025. The Archive–Hands nexus as driver of demographic politics. Three-regime typology extended.

Global Slavery Index 2023 crossed with CAMS data for 42 societies. Archive and Stewards are the critical nodes. A four-path structural taxonomy.

Five-scorer ensemble analysis of Apple, Cargill, Citigroup, Coherent, GE Aerospace, and Goldman Sachs, 2000–2026. The same framework, now applied to firms.

Four synchronised signals in the AI sector: threat framing, $38.5B net loss, ratepayer stress from data centres, and 45% S&P500 AI concentration as fragility indicator.

What distinguishes the CAMS approach from standard macroeconomic or political science modelling — and why that distinction matters empirically.

Where the framework sits among Popper, Kuhn, Prigogine, the Santa Fe school, Ibn Khaldun, and Turchin. Each big claim stated alongside the conditions under which it would fail.

First-person log, October 2024 to November 2026. Wrong turns left in. Twelve phases, short commit SHAs throughout. "We don't edit the past, we qualify it."

The founding paper. Five longitudinal datasets, six falsification criteria. The original case for the eight-node, four-metric architecture.

How the thermodynamic coordination formalism fits with extended evolutionary theory. Phase-space diagnostics from first principles.

Two composite signals that track how hard a society is working (operational pressure) and how much it holds together symbolically (symbolic integration). Tested on 2,500 years of Latium.

Does the narrative overlay add information or just decoration? 4,033 complete society-years tested. Three headline renderings are relabelled extrema. Dream–energy alignment: non-redundant but non-predictive.

Simulation report on the stampede-as-phase-transition claim. Explosive Kuramoto hysteresis and branching-process contagion. Conditional support — open code and data.

Five blinded CAMS time-series. Does the same framework work at different scales of institution? A blind inference challenge with downloadable data.

32 falsifiable, probability-weighted claims under the current framework, lodged before outcomes are known. Zenodo DOI 10.5281/zenodo.24046639. Scoring begins when the period ends.

The 12 falsifiable predictions from the June 2026 paper, locked verbatim before any scoring. The instrument is not externally validated until this holdout is scored.

Verdicts for all 15 items under six scorer pools, including per-scorer verdicts and the signed Section 7 file. The raw record behind the results summary.