JUNO/CAMS Blind-Analysis Protocol v1.1.1

This page presents results of the JUNO/CAMS Blind-Analysis Protocol v1.1.1, which tested whether decomposing the legacy activation term into distinct cognitive (gᵢ) and affective/energy (eᵢ) components improves morphological classification and identity disambiguation. Three unlabelled blind sets of 8 societies each were analysed using a strict falsification-first two-track design: canonical operators locked first, then cognitive-affective operators computed, keys revealed only at Stage 6.

Relation to JUNO v1.2. This experiment used the v1.1.1 protocol — a cognitive-affective extension that decomposes the activation term, distinct from the v1.2-Final numerical-policy and dead-code fixes. The v1.1.1 operators are not incorporated into the production JUNO v1.2 classifier; this experiment informs future protocol iterations.

Key Results — Pooled (24 Societies)

66.7%

Track 1 Rank-1 Accuracy
(canonical)

83.3%

Track 2 Rank-1 Accuracy
(augmented) +16.6%

0.667

MRR — Track 1 (canonical)

0.833

MRR — Track 2 (augmented) +0.167

Formal result: FAILED. Despite the MRR improvement, the amended Stage 2B criterion (A4) requires non-degradation of confidence calibration. Track 2 increased confidence on false positives for Afghanistan and Iraq (both true family: conflict-state), which were incorrectly relabeled to cognitive-affective split with high confidence. This triggered the calibration degradation penalty.

Set-by-Set Breakdown

Set	Track 1 MRR	Track 2 MRR	Delta	Finding
Set 1	0.750	0.750	0	Honest null — signed operators confirmed canonical; Hands Drag detected in 6/8 societies
Set 2	0.500	1.000	+0.500	High disambiguation value — 4 misclassifications corrected (China, Russia, Poland, Iran → cognitive-affective split)
Set 3	0.750	0.750	0 (net)	Calibration failure on Afghanistan and Iraq; 2 correct relabelings offset by 2 false-positive upgrades

Error Migration Audit

4

Misses fixed by Track 2
(China, Russia, Poland, Iran)

0

New misses created by Track 2
(no canonical correct overturned)

Interpretation

The calibration gate failure reveals a specific limit of the v1.1.1 signed operators: while they correctly decode masked internal stress in authoritarian and transitional states — where high cognitive elaboration coexists with negative affective valence — they misinterpret structural collapse in active conflict states. In Afghanistan and Iraq, extreme viability failure (Vᵢ < 0) combined with negative affective valence was misclassified as a cognitive-affective inversion rather than a Freeze/Collapse event.

Fix identified. The precedence order in the regime classifier must be refined so that extreme negative viability (Rule 1: Freeze/Collapse) strictly overrides any cognitive-affective inversion signals. This would prevent high-confidence cognitive-affective split labels being assigned to societies undergoing fundamental systemic collapse. Implementation is planned for a future protocol iteration.

The sign test (one-sided binomial, H₀: no improvement) returned p = 0.0625 on 4 positive deltas, 0 negative deltas, and 20 tied societies — below the α = 0.05 threshold, primarily because only 4 of 24 classifications changed rank. The improvement is real but not yet statistically significant at the required confidence level given the current sample size.

Figures

All figures generated programmatically from the scored corpus. Click any image to open full-size.

Figure 11: Rank-1 Accuracy comparison Track 1 vs Track 2

Fig 11 — Rank-1 accuracy: canonical (Track 1) 66.7% vs augmented (Track 2) 83.3%

Fig 10 — MRR per society; note Set 2 improvements

Fig 13 — MRR delta; positive = Track 2 improvement

Fig 8 — Disambiguation summary; red bars = Track 2 forced primary family relabeling

Fig 12 — Full classification scoring matrix

Fig 3 — Regime classification timeline (v1.1.1)

Fig 1 — Mean Vᵢ trajectories by set

Fig 5 — Signed activation heatmap sᵢ_signed by node

Documents & Data

Report

JUNO/CAMS Blind-Analysis Protocol v1.1.1 — Full Report

Protocol specification, methodology, all set analyses, Stage 6 scoring · PDF

CSV

Stage 6 Scoring Results

Post-key scoring table: regime, family, MRR, Track 1 vs Track 2 · CSV

CSV

Canonical Society-Year Metrics

Track 1 — V̄, B̄, s_min, λ₂ for all 24 societies across 3,000+ society-years · CSV (448 KB)

CSV

Signed Society-Year Metrics

Track 2 — cognitive (gᵢ) and affective (eᵢ) decomposition; signed activation sᵢ_signed · CSV (637 KB)

Figure

Figure 8 — Disambiguation Summary

Rank-1 identity disambiguation: canonical vs augmented per society · PNG

Figure

Figure 10 — MRR Comparison

MRR per society across both tracks, showing Set 2 improvements · PNG

Figure

Figure 11 — Rank-1 Accuracy

Pooled Track 1 vs Track 2 rank-1 accuracy · PNG

Figure

Figure 12 — Scoring Matrix

Full classification scoring matrix · PNG

Figure

Figure 1 — Viability Trajectories

Mean Vᵢ trajectories across all sets · PNG

Figure

Figure 3 — Regime Timeline

v1.1.1 regime classification timeline across society-years · PNG

Figure

Figure 5 — Signed Activation Heatmap

Mean sᵢ_signed — cognitive/affective alignment and inversion by node · PNG