Proposed process
Citizens, analysis, and accountable authority
The proposal separates public input, research, recommendations, and decisions so each can be examined on its own terms. It is designed to preserve uncertainty, disagreement, and responsibility rather than hide them inside one answer.
Proposed process
The detailed public flow
Each stage would create records for later challenge. The stages organize inquiry; they do not turn analysis into democratic consent or legal authority.
A question could begin with citizen submissions, testimony, an institutional request, new evidence, an implementation failure, an appeal, an emergency, or research. Its source and requested action would be recorded.
Participants would document competing framings, the relevant jurisdiction, affected stakeholders, protected rights, local context, and a no-action alternative.
A citation begins as a lead, not proof. Reviewers would retrieve sources, connect claims to evidence, assess quality and context, and track evidence lineage.
Distinct AI systems and human reviewers could independently analyze evidence, construct alternatives, question assumptions, request targeted research, and test failure scenarios. Multiple model instances are not democratic voters.
An evidence assessment, values-and-trade-offs map, AI-assisted recommendation, dissent, and human-authorized decision would remain separate records.
The legally responsible institution could accept, reject, or modify the recommendation. Start proportionately. Where lawful and appropriate, begin with advice, simulation, or a bounded pilot.
Important forecasts and success criteria would be registered before results are known. Outcomes, complaints, appeals, and new evidence could trigger review without automatically changing policy.
What travels through the system
Citizen testimony, public evidence, analytical records, human authority records, and outcomes carry different duties. Provenance should make later conclusions traceable without exposing protected material.
| Information type | Examples | Main protections |
|---|---|---|
| Citizen input | Issues, testimony, priorities, objections, appeals | Consent, privacy, representation limits, anti-capture checks |
| Evidence | Research, law, administrative data, forecasts | Verification, lineage, uncertainty, freshness |
| Analytical records | Framings, options, assumptions, dissent, recommendations | Reproducibility, provenance, procedural audit |
| Authority records | Authorization, rejection, modification, rationale | Legal authority and accountability |
| Outcomes | Implementation effects, complaints, new evidence | Predeclared metrics, appeal, review, version history |
Illustrative example
Urban planning
This fictional case uses no real jurisdiction, resident data, measured NousPolis results, precise probabilities, or actual policy recommendation.
Residents, businesses, service providers, and underrepresented groups could contribute testimony alongside planning, legal, and infrastructure constraints. Competing framings might emphasize housing supply, displacement, accessibility, land value, regional transport, or neighborhood continuity.
Reviewers could examine rents, vacancies, construction, transit use, accessibility, infrastructure capacity, emissions, displacement, and transfer limits on evidence from other cities. Alternatives would include no change as well as phased or combined approaches.
Evidence, trade-offs, recommendations, minority reports, and the human city decision would remain distinct so disagreement is visible and analytical output is never mistaken for authorization.
Predeclared measures could cover housing delivery, rent burden, displacement, accessibility, infrastructure load, travel behavior, emissions, and unequal effects. Crossing a condition starts review; it does not automatically change policy.
Illustrative example
Climate policy
A jurisdiction-neutral case would examine household burdens, worker and industry transition, rural and urban differences, public health, utility constraints, infrastructure, fiscal capacity, and law.
Competing framings could emphasize emissions, affordability, energy security, health, employment, adaptation, rights, or intergenerational responsibility. Analysts would test central, adverse, tail-risk, and implementation-delay scenarios without inventing precise probabilities under deep uncertainty.
Policy portfolios could combine pricing, retrofits, clean generation, grid investment, transit, industrial policy, or resilience spending. Evidence assessment, values and trade-offs, AI-assisted recommendation, and human-authorized decision remain separate.
Some reversible interventions might support a bounded pilot. Irreversible or population-scale action would need stronger scrutiny and explicit human and legal authorization.
Proposed process comparison
Different governance questions
NousPolis is primarily a process for researching, challenging, authorizing, and revising public policy. Prediction markets may inform forecasts but do not govern. Architecture v0.1 does not define proxy delegation or a citizen voting system, and model-role separation is not institutional independence.
AI agents are not citizen representatives or democratic voters. The architecture requires neither blockchain nor token voting because neither technology by itself supplies legitimate participation, independent oversight, or public authority.
Robin Hanson's Futarchy proposal is a primary Futarchy reference. Chiara Valsangiacomo's scholarly definition discusses Liquid Democracy, while The Principles of LiquidFeedback describes one implementation.
Today
Limits and unresolved questions
There is no deployed public-policy workspace and no public-policy authority. N1 remains blocked until every mandatory safeguard passes. Rules for citizen standing, identity, representation, voting, delegation, and ratification remain constitutionally unresolved. Participation infrastructure still needs anti-bot and anti-capture protections, privacy, moderation, accessibility, and due process. The full process has not proved its cost, speed, scalability, or real-world performance.
Today
Evidence and next steps
Read Architecture v0.1, the N1 enforcement requirements, the dated N1 evidence snapshot, the independence rules, and the adversarial review disposition.