# Research corrections and verification — Charter v1.2.2

## Scope and identity

The original long-horizon research archive was compared directly with the public code and result tables. This file-identity and arithmetic check is separate from the recorded baseline re-execution.

Archive SHA-256: `819d1066bc37c57b9d56b31eec045b530240f373fb6478493d2cb548d6f227ab`.

All **28 corresponding code/data files** in the public long-horizon archive are byte-identical to their original ZIP members. Individual paths and SHA-256 values are in `v122-verification.json`. In particular, both the v2 baseline and skeptic tables and the model code are unchanged. The restored archive contains the same v2 numerical campaign.

Recomputed all **3,120 stored value scores** across 140 baseline and 380 skeptic rows from source outcome columns, weight rows and the code's shared penalties. Largest absolute discrepancy: **3.33 × 10⁻¹⁶**, ordinary floating-point rounding. This is arithmetic verification, not a fresh trajectory rerun or external model validation. Existing `long-horizon/verification.json` separately records the prior internal reproduction checks.

## Correction to the Testing and Validation Brief

The v1.2.2 Testing and Validation Brief says:

> PULSE first in 114/120 baseline comparisons; second in 6/120
This is the baseline long-horizon model, where PULSE is assigned high effectiveness on capabilities its institutions are intended to provide. Those effects are assumptions, not independently established results. It is conditional systems analysis, not independent validation. In the skeptic sweep, roughly a 15–20% discount to PULSE-specific institutional benefits removed its aggregate modeled advantage over the strongest liberal-democratic comparator.

The first count is correct; **second is false**. The source CSV ranks PULSE **fourth in each of the remaining six comparisons**, all in `19 Weak institutions`. The source text is preserved as an archival document with an erratum. Current summaries use fourth place.

The superseded concise overview correctly states first in 114/120 with all losses in the same weak-institutions world; it does **not** make the erroneous second-place claim. The Testing Brief’s later prose likewise correctly identifies the common weak-institutions loss.
This is the baseline long-horizon model, where PULSE is assigned high effectiveness on capabilities its institutions are intended to provide. Those effects are assumptions, not independently established results. It is conditional systems analysis, not independent validation. In the skeptic sweep, roughly a 15–20% discount to PULSE-specific institutional benefits removed its aggregate modeled advantage over the strongest liberal-democratic comparator.

## Exact baseline means

Unweighted mean across 20 baseline worlds; mean value score additionally gives equal weight to the six value systems. Each world/architecture row has 8,000 trajectories over 300 modeled years.

| Architecture | Mean six-value score | No catastrophic failure | Free and integrated | Survival without collapse |
|---|---:|---:|---:|---:|
| PULSE | 0.579132959726 | 62.710625% | 56.808125% | 79.113750% |
| Liberal constitutional democracy | 0.524754324171 | 63.091250% | 28.490625% | 81.725625% |
| Majoritarian democracy | 0.440108332274 | 53.460625% | 7.473125% | 77.853125% |
| Minimal-rights market order | 0.425750631123 | 49.745000% | 7.821250% | 76.938125% |
| Species-separated federalism | 0.425077379672 | 50.780000% | 4.901250% | 68.138750% |
| Technocratic guardian governance | 0.400161136435 | 34.077500% | 5.846250% | 66.853125% |
| Strict consociational parity | 0.377232546660 | 45.003125% | 13.916250% | 59.557500% |

Testing Brief's rounded **62.7% vs 63.1%** statement is supported. Liberal is higher by **0.380625 percentage points** in raw catastrophic-failure avoidance. PULSE's score lead does not establish the highest survival or dominance on every metric.

`no_catastrophic_failure` excludes collapse, unrecovered species loss, domination, authoritarian capture and individuality failure. `survival_rate` only excludes collapse. `free_integrated_rate` additionally requires no bifurcation and threshold performance on five average states. These quantities must not be treated as interchangeable or as real-future probabilities.

## What the “15–20%” statement actually means

In `pulse_skeptic_sweep_v2.csv`, means across the same 20 worlds and six scores are:

| PULSE benefit multiplier | Mean six-value score | Difference from fixed liberal baseline |
|---|---:|---:|
| 0.80 | 0.518577381317 | −0.006176942854 |
| 0.85 | 0.533778527887 | +0.009024203716 |

Fixed liberal baseline: **0.524754324171**. Therefore the *sampled aggregate crossing* is between **0.80 and 0.85**, supporting the coarse phrase “roughly a 15–20% discount eliminated the aggregate modeled advantage.” The sampled results do not establish an exact interpolated threshold.

The transform is **`clip(0.5 + (coefficient − 0.5) × multiplier)`** for selected PULSE parameters. It reduces the increment above a generic 0.5 reference, not the entire coefficient. Example: rights 0.96 becomes 0.868 at multiplier 0.80; it does not become 0.768. It is not a measured percentage of real-world performance.

Scaled: `rights`, `checks`, `voice`, `group`, `integ`, `cog`, `resil`, `private`, `adapt`, `restore`, `federal`, `innovation`, `equality`. Unscaled: `grid`, `complexity`, `salience`, `techgov`, `emergency`. Costs remain in the tested PULSE architecture. Competitor scores are held at the baseline parameterization.

Individual scoring systems cross at different sampled intervals:

- PULSE-aligned, Stability-first and Individuality-first: multiplier **0.75–0.80**.
- Capability-utilitarian: **0.80–0.85**.
- Classical-liberal and Procedural-democratic: **0.85–0.90**.

Thus **15–20% applies to the aggregate**, not to every value system individually. Historical shorthand about a common interval does not apply to every scoring system.

## Assigned effects and structural dependencies

These findings are conditional on model assumptions. PULSE is assigned high baseline effects: rights **0.96**, checks **0.92**, group protection **0.95**, integration **0.92**, cognitive protection **0.95**, resilience **0.95**, restoration **0.96**. Liberal comparator values are respectively **0.88**, **0.84**, **0.52**, **0.34**, **0.74**, **0.58**, **0.38**. These are hypotheses about what institutions deliver, not empirical measurements produced by the simulations.

PULSE also has larger assigned complexity (**0.46 vs 0.22**) and gridlock (**0.22 vs 0.18**) than the liberal comparator. This helps explain why weak-capacity settings can reverse comparative performance. It does not validate those exact cost assignments.

The code directly enforces a PULSE-only illicit-origin share cap of **0.30**. That result assumes enforcement rather than independently demonstrating it. Integration's ablation changes both integration strength and group salience; it cannot isolate the capability, redundancy and cohesion components or validate the precise v1.2.2 non-coercion boundary.

All six value systems share outcome definitions and subtract **0.22 × catastrophic-failure rate**, **0.08 × bifurcation rate**, and **0.05 × private-sovereignty rate**. Alternate weights are not independent moral validation.

## Reproducibility language and remaining limits

The original code and aggregate output files are available. Historical statements about missing files describe an earlier evidence state and are superseded.

Remaining limitations:

- Outputs are aggregate CSV rows, not saved individual trajectory logs.
- Original script defaults use 10,000 baseline, 2,500 skeptic and 3,500 ablation trajectories per cell; the v2 stored campaign uses 8,000, 800 and 1,200. Reproduce with the existing portable runner's explicit settings, not unqualified default script execution.
- The Testing and Validation Brief calls **2,100,000 reduced-form samples “trajectories”**. Static code inspection shows randomized parameter draws with hazards analytically accumulated over 60 periods. These are reduced-form parameter samples, not simulated Markov paths.
- **138,840,000** five-year transitions is the nominal total trajectory-count × 60. The loop can exit after all cases collapse; it is not evidence of that many independent civilizations or guaranteed executed state updates.
- Old targeted provenance gaps do not disappear merely because this separate long-horizon archive is available.
- Reproduction reduces uncertainty about arithmetic and execution; it does not validate the assumed effects, behavioral model or synthetic-person psychology.

## Verification data

`v122-verification.json`: archive identity; baseline dimensions, exact means and all 840 rankings; skeptic multiplier means and crossing intervals for all scores; architecture parameters; value weights; score-recomputation error; 28 byte-hash comparisons; source contradictions; experiment counts; remaining limits. The file records a direct comparison with the original ZIP and arithmetic recomputation; this check does not execute model code.

CITATION CORRECTIONS
The superseded concise overview combined the 2016 NBER working-paper DOI 10.3386/w22822 with the 2020 Econometrica article details. The journal DOI is 10.3982/ECTA14835. The journal uses 12 years; the working-paper abstract describes 20 years. These editions are distinguished in the current reference register and scholarly working paper 1.0.
Northern Ireland uses members voting for its 60/40 rule; PULSE uses the full Senate and protected delegations. This is a structural analogy, not an identical constitutional rule.
Constitution comparison: all 344 provisions unchanged; only edition/front matter differs. No constitutional contradiction was found.
