Does a Society Think and Feel?
Testing the CAMS5 distributed-cognition thesis on 229 society-years of institutional history
The CAMS framework models a society as eight institutional nodes whose capacity to "think" (Abstraction × Coherence) is gated by how they "feel" (Capacity − Stress). This report puts that interpretation through a five-part statistical battery. The verdict is graded: the framework's emotional core is empirically solid; its cognitive claims are coherent but not yet load-bearing.
In plain language (2,791 characters)
Can a society think? Testing a model of the national mind
The CAMS framework pictures a society as eight interacting institutional organs: leadership (Helm), defence (Shield), memory (Archive), meaning-making (Lore), administration (Stewards), production (Craft), labour (Hands) and exchange (Flow). Each organ is scored on four dials — Coherence, Capacity, Stress and Abstraction. From these the theory builds two bold ideas. First, an institution's "thinking" can be approximated by its capacity to model the world (Abstraction) multiplied by how well it holds together (Coherence). Second, its "feeling" is the margin between Capacity and Stress: surplus feels like confidence, deficit like exhaustion. Multiply thought by feeling and you get a "pulse" — how much of what an institution knows can actually be put to use.
We tested these ideas on two long national histories: Australia (1875–2026) and Argentina (1950–2026).
The emotional thermometer works. In both countries the capacity-minus-stress margin plunged exactly when history says it should: Australia's 1890s banking crash, the Great Depression, the Pacific War, the 1970s stagflation and the pandemic; Argentina's coups, hyperinflation and the 2001 collapse. Across two centuries of combined data, crisis years sit on average in the bottom tenth of the scale. The model survives a harder test: the relationship shape learned from one country predicts the other's pattern almost perfectly — though each nation needs its own calibration, like two instruments tuned to the same musical scale.
The thinking side is shakier. The "thought" measure adds almost no predictive power once you already know how stressed an institution is: feeling dominates the pulse, and knowledge rides along. And the theory's most intriguing claim — that under pressure societies shift from reflective organs (memory, meaning, planning) to reactive ones (defence, labour, trade) — held in chronically stressed Argentina, but reversed in wealthy Australia, where depressions crushed labour and trade while government grew. Stress does not universally shorten a society's mind; the response depends on the kind of society and the kind of crisis.
Nor could we tell which comes first. Does stress suppress reflection, or does failing reflection invite stress? The two move together, and 150 years of data refuse to say who leads.
The verdict, then, is neither dismissal nor coronation. The framework's emotional core is real and historically literate; its cognitive claims remain promising but unproven. The next test is whether these measures can predict things outside the model — growth, unrest, recovery — better than stress alone. Until then, "societies think and feel" stands as a serious hypothesis: one pillar solid, one under construction.
General-audience summary · 2,791 characters
Verdicts on the seven graded conclusions
The test battery
T1 — Does A·C add predictive value beyond (K−S)?
Expanding-window, one-step-ahead forecasts of next-year aggregate pulse (minimum 25 training years), compared by paired circular block-bootstrap on squared errors:
| Society | M0 mean | M1 (K−S) | M2 (K−S)+A·C | M3 pulse | M2 > M1? |
|---|---|---|---|---|---|
| Australia | −0.03 | 0.567 | 0.574 | 0.658 | No (56% bootstrap agreement) |
| Argentina | −0.03 | 0.465 | 0.444 | 0.572 | No (point estimate negative) |
M3 (pulse) beats M1 at 98% bootstrap confidence in both societies — but M3's out-of-sample R² is identical to an AR(1) on the target. The edge is target autocorrelation. Forecasting next-year change in (K−S) is near-hopeless from current CAMS state (all OOS R² ≤ 0.09), and adding A·C makes it worse.
T2 — Does stress switch computation from deliberative to reflexive nodes?
Deliberative D = {Helm, Archive, Lore, Stewards}; reflexive R = {Shield, Hands, Flow}; balance = D-share of total A·C. Associations tested against 5,000-draw circular-shift nulls that preserve each series' autocorrelation:
| Cut | Australia | Argentina |
|---|---|---|
| Full series, ρ(KS, D-share) | −0.16 (p = 0.39) | +0.39 (p < 0.001) |
| Overload years only (KS < 0) | −0.68 (p < 0.001) · n = 25 | +0.45 (p < 0.001) · n = 52 |
| Surplus years only | −0.22 (n.s.) | +0.25 (n.s.) |
| Tercile contrast, high − low KS | −0.007 (p = 0.55) | +0.036 (p < 0.001) |
The switching relation is sign-inconsistent across societies and cancels when pooled (ρ = 0.04, n.s.). Argentina behaves as the thesis predicts; Australia's acute crises show the opposite morphology — reflexive economic nodes (Hands, Flow) collapsed further than the deliberative state (1890s, 1930s), which is historically intelligible but contrary to a universal stress-shortens-pathways law.
T3 — Does the model transfer across societies?
| Model | Fitted on → applied to | In-sample R² | OOS R² | Rank ρ(pred, actual) |
|---|---|---|---|---|
| pulse ~ (K−S) | Australia → Argentina | 0.976 | −2.66 | 0.976 |
| pulse ~ (K−S) | Argentina → Australia | 0.816 | +0.25 | 0.991 |
| D-share ~ (K−S) | either direction | 0.01–0.17 | < 0 | 0.15–0.39 |
The monotone form generalises almost perfectly; the calibration does not (Australia's pulse scale is roughly 4× Argentina's). No significant common international component: ρ(KS_AU, KS_AR) = 0.27, shift-null p = 0.20. Flagged anomaly: deliberative shares are anti-correlated (−0.43) across the two societies.
T4 — Which leads: feeling or thought?
Cross-correlation profiles of D-share vs (K−S) peak at lag k = 0 in both societies. Argentina's raw asymmetry hints that deliberative balance leads the energetic margin by ~1 year (ρ(k=+1) = 0.38 vs ρ(k=−1) = 0.23) — the reverse of the stress-gates-cognition arrow — but neither the levels asymmetry (p = 0.34) nor the differenced lead (ρ = 0.16, p = 0.23) survives the nulls.
T5 — Is the energetic axis historically literate?
| Society | Crisis years | Mean (K−S) percentile | Null p | Mean Stress percentile |
|---|---|---|---|---|
| Australia | 1893, 1930–32, 1942, 1974–75, 2020 | 9.4% | < 0.0002 | 92.2% |
| Argentina | 1955, 1976, 1982, 1989, 2001–02, 2020 | 10.5% | < 0.0002 | 91.7% |
Sensitivity summary
| Parameter varied | Result | Impact |
|---|---|---|
| Craft moved into deliberative group | AR ρ 0.33 (p = 0.023); AU −0.23 n.s. | None |
| Flow removed from reflexive group | AR ρ 0.60 (p < 0.001); AU +0.16 n.s. | AR effect strengthened; AU still null |
| Pulse-share instead of A·C-share | AU −0.32 n.s.; AR −0.09 n.s. | Switching visible only in the A·C margin |
| Australia pre-1950 / post-1950 | −0.43 (p = 0.086) / +0.15 (n.s.) | AU sign instability across eras |
| Argentina post-1980 | ρ 0.30 Spearman n.s. / 0.45 Pearson (p = 0.04) | Weaker in modern era, same sign |
| Pearson vs Spearman | Same signs, similar p-values | Estimator-robust |
| Node-level stress slopes (overload years) | AU gap +0.01; AR gap −0.06 | Selective sensitivity small, sign-unstable |
Data findings that matter for the pipeline
1 · Legacy formula identified. The Australia file's Node Value
column is exactly C + K − S + 0.5·A (OLS R² = 1.0000) — a linear
legacy aggregate, not the canonical pulse. It correlates with σ at ρ = 0.98 and must not
be mixed with canonical pulse in downstream pipelines.
2 · Bond Strength is not node-reconstructible. Best fit from node-level C/K/S/A reaches R² ≈ 0.87, consistent with a network-level bond equation whose inputs are absent from this file. It attenuates with Stress (ρ = −0.74), as the canonical equation implies, but exact verification needs edge-level data.
3 · No independent outcomes. These files contain only CAMS variables. The thesis's stated next threshold — incremental prediction of independently measured outcomes — could only be tested internally here (future CAMS states from current ones).
Revised graded conclusions
- (Kᵢ−Sᵢ) is a validated descriptive-behavioural measure of institutional energetic condition — historically literate in both regimes, rank-transferable across societies, and the framework's workhorse.
- (AᵢCᵢ) is formally coherent but currently predictively redundant given (K−S); its value must be sought against independently measured outcomes, not internal dynamics.
- Pulse σ behaves exactly as designed and is empirically ≈ a monotone transform of (K−S) at node level: a state summary, not added information.
- Deliberative–reflexive switching is real but regime- and society-specific. Argentina (chronic overload) shows thesis-direction suppression; Australia (surplus society, acute crises) shows the opposite sign. The law needs a regime qualifier or a nonlinear threshold form before it can be stated universally.
- Causal direction between energetic margin and deliberative balance is unresolved: only contemporaneous coupling is established.
- Society-specific calibration and sign-inconsistent switching support the "slow structural tendencies" reading of national profiles — and caution against universal dynamic laws fitted on a single society.