STStratosGovernance diagnostic

Stratos: a regime-aware governance diagnostic

Working name. To be finalised before public launch.

Status: pre-MVP. Public repository. Contributions welcome.


What this is

Stratos is an open-source diagnostic dashboard that evaluates how effectively governments convert their resources, institutional authority, and structural inheritance into human flourishing, resilience, fairness, and adaptive capacity. It is built on publicly available data, with full methodological transparency and code reproducibility.

It is not a ranking. It is not a single index. It does not produce a leaderboard.

It produces country-level diagnostic profiles with explicit uncertainty bands, regime-aware comparisons, structural-constraint adjustment, and clear flags on data quality. The output is conditional: "given this country's regime classification, its structural constraints, and the validity of its statistical reporting, here is how it is performing on five governance dimensions relative to its actual peers."

The killer feature is not the dashboard. It is the diagnostic narrative engine that reads each country's data and surfaces the meaningful comparisons, the genuine performance gaps, the trajectory direction, and the limits of what the data can tell us.

Why this project exists

Three problems with existing governance dashboards drove this design.

The first is construct confusion. Most governance indices collapse incommensurable values into a single ranking. Democracy, state capacity, economic management, and human welfare are treated as ingredients in one number. Different value systems produce different rankings, and users have no way to see which value choices are driving the output.

The second is the comparison problem. Comparing Norway with Niger as if they face equivalent constraints is a category error. Existing dashboards mostly ignore structural inheritance (geography, demography, resource endowment, historical institutional legacy, conflict exposure) or treat it as a footnote rather than as the foundation of valid comparison.

The third is measurement opacity. Official statistics are produced by the entities being measured. Argentina falsified inflation data for over a decade. China's provincial GDP totals routinely exceed national GDP. Authoritarian regimes report fictional turnout and completion statistics. Any dashboard built on these inputs without explicit cross-validation and uncertainty quantification inherits the credibility problems of its sources.

Stratos addresses all three by design. Values are explicit and adjustable. Comparison happens within regime cluster and structural-similarity cohort first. Data quality is a first-class output, not a footnote.

Who this is for

Three audiences, in priority order.

Citizens and journalists who want to understand whether their government is performing well given the constraints it inherited, and how it compares to peers facing similar conditions. The diagnostic narratives are written for this audience.

Policy researchers and governance scholars who want a reproducible, transparent platform for cross-national governance analysis with proper uncertainty handling. The methodology and raw data layers serve this audience.

Civil society organisations and NGOs working on governance reform who need defensible diagnostic baselines for advocacy and program design. The trajectory and structural-constraint layers are particularly relevant here.

We are explicitly not building this for governments themselves, who have access to better internal data than any public dashboard can offer, nor for investors or sovereign-risk analysts, who already pay for proprietary alternatives.

What makes Stratos different

Four substantive choices distinguish this project from prior governance dashboards.

1. Regime classification before scoring

Every country-year is classified into a governance regime cluster before any comparison happens. The clusters are derived from V-Dem regime types and capacity indicators, and they reflect a theoretical position: different governance regimes have different objective functions, so scoring them against a single rubric mislabels both successes and failures.

Within-regime comparison is the primary view. Cross-regime comparison is available but framed as comparing different operating models, not different performance levels on a shared scale.

2. Structural-constraint adjustment

Before scoring, each country is conditioned on what its current government did not choose: geography, demographic structure, resource endowment, historical institutional legacy, conflict exposure, baseline disease burden. The dashboard shows both adjusted and unadjusted views. Adjusted: "how is this government performing given what it inherited." Unadjusted: "how is this country doing in absolute terms." Both are legitimate questions.

3. Errors-in-variables throughout

Governance indicators are textbook measurement-error problems. GDP per capita is measured with error. HDI components are measured with error. The relationships between inputs and outcomes are estimated with error. Stratos applies Probability Cloud Regression to the conversion-efficiency dimension to produce confidence bands rather than point estimates, and propagates that uncertainty through to the diagnostic outputs.

This is a substantive methodological contribution. Most governance dashboards report point estimates as if the underlying indicators were measured precisely. They are not.

4. Diagnostic narratives, not rankings

The headline output for each country is a written diagnostic that names the meaningful comparisons, the largest performance gaps, the trajectory direction with confidence, the structural constraints shaping the result, and the data-validity concerns. The numbers exist to support the diagnostic, not the other way around.

This is the move that makes Stratos useful rather than ornamental. A leaderboard tells you who is at the top. A diagnostic tells you what is actually happening.

Theoretical foundation

Stratos rests on a substantive theoretical claim: governance is fundamentally a problem of uncertainty navigation under structural constraint. States face exploration-exploitation trade-offs at every level. Policy experimentation versus standard delivery. Regulatory updating versus stability. Federal versus subnational coordination. Long-horizon investment versus short-horizon response.

Different regime types embody different solutions to this trade-off. High-capacity liberal democracies push exploration into civil society and the legislative process while exploiting through stable bureaucracy. Closed autocracies centralise both exploration and exploitation, producing high variance in outcomes that depends heavily on individual decision-makers. Fragile states often lack adaptive capacity entirely.

Evaluating governments requires understanding which trade-off they face and how well they navigate it. The five Stratos dimensions (execution capacity, conversion efficiency, distributional reach, accountability infrastructure, adaptive capacity) map directly onto this framing. The first two are exploitation-side. The last two are exploration-side. The middle one is the distributional question that exploitation alone cannot answer.

The detailed theoretical argument lives in the methodology specification and (eventually) in a separate paper.

Scope: what Stratos is and is not

Stratos is:

  • A diagnostic tool for understanding governance performance under explicit assumptions
  • A platform for reproducible cross-national governance analysis
  • A demonstration of regime-aware, constraint-adjusted, measurement-error-aware methodology
  • An open-source public good with permissive licensing for code and data

Stratos is not:

  • An investment-grade sovereign-risk product
  • Political analysis or commentary on specific officials or parties
  • A predictor of regime transitions or political crises (though it may inform such predictions as inputs)
  • A substitute for in-depth country expertise
  • A judgement of regime legitimacy in any normative sense
  • Affiliated with any government, international organisation, or political movement

Public-facing posture

Stratos is built and maintained transparently. Three commitments shape the public posture.

Methodology transparency. Every score on the dashboard is one click away from the raw indicators feeding it, the source and vintage of each indicator, the conditioning variables, the uncertainty quantification method, the full underlying data, and the code that computed it. There is no proprietary scoring black box.

AI-augmented development, disclosed. This project is built by an independent researcher working with AI-assisted coding tools. The methodology is human-designed. The implementation is human-supervised. AI use is disclosed on the methodology page and in the contributor documentation. The disclosure is part of the brand, not a vulnerability.

Open to criticism. The dashboard publishes its methodology and welcomes substantive critique. Issues raising methodological concerns are taken seriously and triaged in the open. The goal is to be a credible platform that improves over time, not a closed system that defends its current state.

Success criteria

Phase 1 success (MVP launch, six weeks from project start):

  • Working dashboard deployed on Vercel covering 25-30 high-capacity democracies
  • Five-dimensional scoring with within-cluster percentiles and confidence intervals
  • Country profile pages with diagnostic narratives
  • Full methodology documentation public
  • Codebase public on GitHub with clear contribution pathway
  • At least three external technical reviewers engaged through public posts

Phase 2 success (regime expansion, four months from project start):

  • Coverage extended to all countries with V-Dem classification (180+)
  • Cross-validation against alternative sources for at least one dimension
  • Structural-constraint adjustment fully operational
  • First external contributor merge accepted
  • Coverage in at least one mainstream-adjacent venue (technical blog, governance newsletter, etc.)

Phase 3 success (full vision, eight to twelve months):

  • All five dimensions with PCR-style uncertainty quantification
  • Trajectory module operational with early-warning signals
  • Methodology paper draft circulated for review
  • Active contributor community with at least three regular maintainers beyond the founding author
  • Cited in at least one external research output

Failure criteria (kill switches) are documented in the roadmap.

How to contribute

Contributions are welcome across methodology, code, data sourcing, documentation, translations, and design. See CONTRIBUTING.md for the contributor workflow, code of conduct, and contributor licence agreement.

The project follows a benevolent-dictator model for now, with the founding author as final decision-maker on methodology and scope. This will transition to a broader maintainership structure as the contributor base grows.

Active areas where help would matter most at this stage:

  • External methodology review from governance scholars
  • Data source validation and bias documentation
  • Frontend development (React, Tailwind, shadcn/ui)
  • Country-expert review of diagnostic narratives
  • Documentation and translation

Licensing summary

  • Code: MIT licence (permissive, allows commercial use with attribution)
  • Documentation and methodology: CC-BY 4.0
  • Derived data and aggregate scores: ODbL (Open Database License) where source licences permit
  • Source data: governed by the original source's licence terms (see DATA_SOURCES.md)

Full open-source governance, contributor agreements, and IP boundary documentation are in OPEN_SOURCE_GOVERNANCE.md.

Acknowledgements

Stratos draws on the publicly available work of the World Bank, V-Dem Institute, UNDP, IMF, OECD, Yale EPI, Notre Dame Global Adaptation Initiative, Transparency International, and others. The methodology builds on Probability Cloud Regression (Soliman, 2026) and incorporates regime-aware diagnostic principles developed in adjacent projects.

This project is built without institutional affiliation.

Project lead

Kareem Soliman, independent AI researcher. Sydney, Australia. ORCID: 0009-0006-7204-1129.

Repository: [URL pending] Live dashboard: [URL pending]