Research evidence · v1.2.2
Research / Methods
Compare mechanisms under assumptions. Then challenge the assumptions.
Three layers of evidence
| Layer | What it can establish | What it cannot establish |
|---|---|---|
| External sources | What a study, primary text, or author reports | Future human–synthetic outcomes |
| Targeted stress models | Behavior of specified rules under artificial inputs | Real-world frequencies or complete constitutional validity |
| Internally reproduced long-horizon program | The supplied code regenerates the restored campaign’s aggregate results | Empirical forecasts, external validation, or independent support for assigned coefficients |
Targeted archive
The public research archive contains sanitized CSV data, model scripts, assumptions, and historical reports. Historical names and older rules are retained where needed to understand the experiments. They do not supersede the v1.2.2 Charter. Duplicate nested archives and document-generation scripts are excluded.
Download research archive · Inspect file hashes and inventory · Read base-case accounting
Inspect a model before running it
The archive explores weighted elections, candidate acceptability, thresholds, correlated recognition error, guardian independence, cognitive changes, resilience, integration networks, hierarchy, and durable-wealth dynamics. The preserved scripts specify their own seeds, distributions, sample sizes, and outputs where recoverable. Some historical scripts use fixed working paths or expect companion files; the reproduction guide identifies this rather than claiming a one-click suite.
The network model imposes cross-constituency ties to examine structure. An imposed edge is a modeling device, not a constitutional authorization to compel association. The judicial sensitivity model assumes a latent “correct” outcome; its absolute error rates are not measured judicial truth.
Reproduce the long-horizon campaign
The restored archive includes four model scripts, 24 CSVs, architecture/scenario/value parameter tables, three reports and seven figures. The public edition preserves code and data and supplies a portable runner. Use the v2 baseline, skeptic and ablation files; their sample sizes differ from the original scripts’ defaults. Earlier non-v2 outputs remain labeled historical.
| Experiment | Aggregate result file | Samples per row |
|---|---|---|
| 20 worlds × 7 architectures | main_scenario_architecture_results_v2.csv | 8,000 |
| 19 benefit multipliers × 20 worlds | pulse_skeptic_sweep_v2.csv | 800 |
| Five ablations + full PULSE × 20 worlds | pulse_doctrine_ablation_v2.csv | 1,200 |
| Capacity / homophily / private concentration | institutional_capacity_breakpoint.csv; homophily_integration_sweep.csv; private_concentration_sweep.csv | 4,000 / 3,000 / 3,000 |
| Six compound worlds × 7 architectures | compound_extreme_scenarios.csv | 5,000 |
| Lite capacity / two compound tests | pulse_lite_capacity_test.csv; pulse_lite_extreme_test.csv | 4,000 / 6,000 |
| Reduced-form check | reduced_form_markov_crosscheck.csv | 15,000 |
The runner preserves seed 20260913 and the original experiment-specific seed offsets. It redirects fixed working paths, leaves equations and supplied data untouched, and records comparison tolerances and differences. Verified with NumPy 2.3.5 and pandas 3.0.1. The older baseline sample did not match every numeric cell, so reproduction claims apply to the explicitly listed restored campaign.
Read reproduction instructions · Download the complete public edition
What the scores mean
Baseline summaries weight all 20 scenario archetypes equally. The six scoring systems share architecture inputs, positive integration weights and common failure penalties. Ranking well across them does not establish robustness to every moral priority or model structure. The model’s free-and-integrated outcome also requires liberty, individuality, integration, anti-domination and legitimacy thresholds, as well as avoiding catastrophic failure and bifurcation.
The capacity sweep changes PULSE-aligned score order between sampled index values 0.50 and 0.55; it does not estimate a real-world administrative threshold. The skeptic sweep crosses a fixed liberal baseline between multipliers 0.80 and 0.85 on the mean of six scores. PULSE leads four of six compound worlds by mean cross-value score and loses the two low-capacity extremes.
The reduced-form model ranks PULSE first by mean quality, but reuses the same hand-assigned institutional parameters. Its different mathematics does not constitute independent empirical validation. PULSE-Lite was designed after observing the weak-capacity failure and is exploratory, not an adopted constitution.
Model assumptions still need challenge
The integration variable bundles capability, continuity and cohesion. A v1.2 model cannot validate the exact v1.2.2 boundary against compelled private association. Shared seed schedules make starting draws comparable, but conditional random calls can disrupt perfectly paired later shocks. A PULSE-only illicit-origin cap is another structural assumption; its name-gated rule does not apply identically to the exploratory Lite variant.
Human studies motivate possible mechanisms. They do not calibrate future synthetic behavior or establish that the assigned coefficients are plausible. The archive’s human-source notes are historical and are not an independently verified parameter-fitting pipeline. Check citation scope and remaining gaps
Human evidence and its limits
Power-sharing research distinguishes institutional forms and political context; it does not establish that all power sharing promotes survival. [graham-2017] [andeweg-2000]
The OECD survey examines human trust and perceived voice. ILO exposure estimates concern tasks, not observed wholesale job replacement. IEA electricity scenarios show infrastructure-planning uncertainty. None supplies synthetic-citizen parameters. [oecd-2026] [ilo-2025] [iea-2025]
How to challenge the research
Change model families as well as parameters. Test whether assumed institutional benefits predetermine rankings; examine correlated errors, transition costs, emergency abuse, resource scarcity, strategic rejection, and coordination costs. Keep base worlds fixed when comparing rules. Report uncertainty and unfavorable cases. Identify which findings reproduce and which depend on irrecoverable provenance.
Source categories
| Category | Meaning |
|---|---|
| External empirical evidence | Measured observations about existing systems; transfer limits stated |
| Constitutional / legal precedent | Existing law or institutional practice; analogy is not identity |
| Governance and institutional scholarship | Research and argument about institutional design |
| Philosophical / conceptual literature | Conceptual or normative arguments; no empirical proof implied |
| AI consciousness / welfare research | Scientific indicators, welfare assessment and uncertainty |
| Engineering / technical guidance | Planning and design guidance, distinct from measured effects |
| Related contemporary framework | An adjacent public proposal, not an endorsement |
| Legal / licensing reference | Copyright and reuse sources, not constitutional precedent |
| PULSE internal primary source | The Charter, project briefs, model code and data |
Claim status is separate from source category
A project assumption stipulates a future capability or causal mechanism. A PULSE simulation result reports model behavior with its provenance limits. An interpretation explains a provision or comparison. Verification notes distinguish full-text checks from abstracts, publisher summaries and bibliographic records.
Trial accounting and distinct units
What the numbers count
The long-horizon counts below are verified from the restored campaign; targeted counts retain their distinct provenance limits. They are different units and must not be added into a single “civilizations simulated” total.
| Measure | Count | Meaning |
|---|---|---|
| Targeted model / stress-test families | 48 | Distinct micro/constitutional families through v1.2 |
| Independently randomized targeted base scenarios/trials | 43,415,600 | Conservative recoverable count |
| Targeted comparative rule evaluations | >60.6 million | Many rules applied to the same base worlds for apples-to-apples comparison |
| Explicit multi-voter election worlds | 31,960 | Narrow election-world count |
| Long-horizon dynamic architecture trajectories | 2,314,000 | 300-year scenario trajectories across baseline, sweeps, ablations, extremes and transition tests |
| Five-year long-horizon state transitions | 138,840,000 | Nominal 60-period loop count; not independent civilizations |
| Reduced-form hazard parameter samples | 2,100,000 | Randomized inputs with analytically accumulated hazards; not simulated Markov histories |
| Baseline long-horizon archetypes | 20 | Demography, scarcity, integration, catastrophe, conflict, wealth, etc. |
| Compound extreme worlds | 6 | Multiple severe assumptions combined |
| Primary governance architectures compared | 7 | PULSE plus six materially different comparators |
| Value systems used for scoring | 6 | Six model weightings sharing inputs and failure penalties; not independent moral validation |
Source: TESTING_METRICS.csv · Current Testing and Validation Brief
The dynamic campaign comprises 2,314,000 architecture trajectories and 138,840,000 nominal five-year periods. The reduced-form check uses 2,100,000 randomized parameter samples with analytic cumulative hazards. The targeted provenance accounting totals 43,415,600 base cases; this accounting is not a rerun of every historical model. The narrow election-world count and more than 60.6 million comparative evaluations retain partial archival provenance.
Capacity sensitivity depends on how outcomes are scored
The PULSE-aligned capacity crossover is 0.50–0.55, whereas the mean-six-value crossover is 0.55–0.60. These are scoring-dependent sampled model intervals, not empirically calibrated administrative thresholds.
Assigned effects are hypotheses
Architecture-effectiveness coefficients are explicit model inputs. PULSE’s high baseline rights, checks, integration, resilience and restoration values are hypothesized institutional effects, not independent estimates.
| Assigned dimension | PULSE | Liberal comparator |
|---|---|---|
| Rights | 0.96 | 0.88 |
| Checks | 0.92 | 0.84 |
| Integration | 0.92 | 0.34 |
| Resilience | 0.95 | 0.58 |
| Restoration | 0.96 | 0.38 |
| Complexity cost | 0.46 | 0.22 |
What the 15–20% discount means Here the percentage describes moving selected institutional-effect coefficients toward the model’s neutral setting, not reducing all assumptions. It is not a measured advantage, safety margin or empirical confidence interval.
For selected coefficients, the transform is clip(0.5 + (coefficient − 0.5) × multiplier). PULSE-specific increments above the 0.5 reference are discounted; gridlock, complexity, salience, technocratic-governance and emergency settings remain unscaled.
The 15–20% wording describes the average across six scoring systems. PULSE-aligned, Stability-first and Individuality-first cross at sampled multipliers 0.75–0.80; Capability-utilitarian at 0.80–0.85; Classical-liberal and Procedural-democratic at 0.85–0.90. These ranges describe this model only.
Recovered original archive: directly checked
The original archive is available. Its 28 corresponding code and data files are byte-for-byte identical to the public research archive already restored to the site. We recomputed 3,120 stored scores directly from those tables and the published scoring rule; differences were no larger than 3.33 × 10⁻¹⁶. This verifies the arithmetic, not the assumed institutional effects.
The original v1.2.2 testing brief reported second place in six comparisons. The actual tables say fourth; the current explanatory edition corrects that wording. Its reduced-form “trajectories” are randomized parameter samples with analytically accumulated hazards. These source errors are disclosed; the Constitution is unaffected.
How the work was done
The framework emerged through iterative systems design rather than from a predetermined constitutional theory: proposed mechanisms were repeatedly retained, revised, or rejected against the project's normative commitments, adversarial counterexamples, relevant prior work, and targeted simulations.
The steward supplied normative judgments, design goals, architectural choices, acceptance/rejection decisions, most adopted thresholds and substantive corrections. AI systems assisted with drafting, literature discovery and synthesis, adversarial critique, consistency checking, simulation design, code generation, formalization and document production. The paper's prose is heavily AI-generated. Internal model critique is not independent peer review.
Working paper 1.0: verification scope
The paper’s recorded verification re-executed the 140-row v2 baseline, recalculated 100 guardian configurations and re-executed nine acceptance settings. Ranks and skeptic/ablation summaries were recomputed from archived tables. It does not claim that every historical sweep was newly executed. Earlier reproduction records remain separate.
The Functional Integration ablation changes both the integration coefficient and identity salience. Its free-integrated outcome also requires integration above 0.50. The contrast therefore does not isolate a causal effect of teamwork.