Architecture v0.1 · N0
/
Building toward the N1 Research Laboratory
Governance for the intelligence age
A modern operating system for governance.
AI helps society think. Humans retain sovereignty.
NousPolis explores how societies can make better collective decisions using evidence, continuous learning, citizen participation, institutional safeguards, and diverse artificial intelligence — without handing political sovereignty to AI.
Participatory
Evidence-rich
Adaptive
Human-authorized
Living governance network
IllustrativePublic issue
Housing affordability
Regional policy review
4framings
12AI roles
2human reviews
Deliberation in progress
Decision authority
Recommendation only
NousPolis currently has no autonomous public-policy authority. That limit is structural, not cosmetic.
Why governance needs an upgrade
Our institutions were built under different constraints.
Modern governments must navigate complex, interconnected problems with limited research capacity, fragmented expertise, slow feedback, and public participation that is often episodic. AI creates new capacity — but only if the institutions around it are designed well.
Conventional cycleOften linear
Issue
→Political process
→Policy
→Implementation
→Occasional review
Important decisions can become difficult to revisit once institutions, budgets, incentives, and political identities form around them.
NousPolis cycleDesigned to learn
Legitimate
decisionhuman authority
Frame
Evidence
Deliberate
Forecast
Measure
Challenge
decisionhuman authority
Public questions can move through explicit reasoning, authorization, measurement, challenge, and revision rather than ending when a policy is adopted.
What governance could become
More capable without becoming less human.
NousPolis is not primarily about automating bureaucracy. It asks what governing institutions could look like when rigorous analysis, structured participation, forecasting, and institutional memory become much cheaper.
Reactive→Anticipatory
Forecast consequences and stress-test policy options before large-scale implementation.
Opaque→Inspectable
Expose evidence, assumptions, disagreement, uncertainty, authority, and revision history.
Static→Learning
Treat policies as revisable interventions whose outcomes can be measured against expectations.
Expert-limited→Intelligence-rich
Give difficult public questions access to broad multidisciplinary research and diverse machine reasoning.
Periodic input→Continuous participation
Create structured roles for citizens and stakeholders between elections and occasional consultations.
One-size-fits-all→Context-sensitive
Adapt decisions to jurisdiction, culture, institutions, and local conditions while preserving higher-order constraints.
How NousPolis works
From public question to institutional learning.
The public-facing lifecycle can be understood as eight linked functions. The underlying architecture applies stricter classifications, provenance, permissions, review rules, and maturity gates beneath them.
Understand
Map the question, jurisdiction, stakeholders, and competing ways of framing the problem.
Investigate
Build evidence states, assumptions, causal explanations, provenance, and explicit uncertainty.
Deliberate
Assemble diverse models, methods, expert roles, challengers, and minority perspectives.
Compare
Evaluate scenarios, forecasts, trade-offs, distributional effects, and risk.
Authorize
Route decisions to legitimate human institutions according to their actual authority.
Implement
Produce advice, pilots, experiments, or implementation only when the relevant gates permit it.
Measure
Compare outcomes with precommitted expectations and track unintended effects.
Learn
Challenge, appeal, revise, supersede, regress, or scale as new evidence arrives.
Constitutional boundaryNothing becomes binding merely because AI models agree.
Make governance inspectable
See the reasoning, not only the decision.
A mature governance system should let people inspect what entered the process: which framings were considered, where the evidence is weak, what citizens prioritized, how forecasts differ, who dissented, and which institution actually holds authority.
- 01Competing framings remain visible.
- 02Evidence quality and gaps are explicit.
- 03Forecasts can be checked against outcomes.
- 04Dissent and appeals remain part of the record.
Illustrative future interface. This visualization describes the intended direction of later maturity levels; it is not a claim that the N0 architecture currently provides these executable capabilities.
ISSUE NP-0047
Urban housing affordability
Regional review · recommendation track
Deliberating
87evidence items6 contested
4competing framingsall retained
12model roles2 challengers
4.2kcitizen signalsweighted by standing
Citizen prioritiesTop signals
Affordability92
Supply74
Stability63
Transit access57
Competing framings4 active
AHousing supply constraint
BLand-value distribution
CCredit & investment distortion
DRegional transport mismatch
Forecast comparison12-month horizon
Policy A−7%
Expected rent effect · medium uncertaintyPolicy B−3%
Lower rent effect · stronger distributional benefitResidents
Stakeholders
Local groups
Structured participationPublic reasoningagenda · framing · priorities · appeal
→
Public explanationHow input affected the process
Citizens are part of the system
Governance should not happen only between institutions and machines.
The long-term NousPolis vision includes structured channels for people to shape agendas, surface local knowledge, challenge framings, express priorities, join deliberative processes, and appeal decisions.
Agenda
Citizens can surface issues that institutions overlook.
Framing
Different communities can contest how a problem is defined.
Deliberation
Public input can be structured rather than reduced to raw popularity.
Accountability
Institutions should explain how participation affected the outcome.
Participation should influence a governed decision process — not become an unmoderated popularity contest.
Governance at the right scale
One architecture. Many levels of governance.
NousPolis is designed to respect jurisdiction, culture, law, rights, institutional structure, and local context rather than impose one centralized answer everywhere.
AI's role
AI expands governing capacity. It does not become government.
AI can help+
- Search and synthesize evidence
- Generate competing interpretations
- Identify assumptions and evidence gaps
- Model scenarios and create forecasts
- Stress-test proposals and dependencies
- Monitor outcomes and support deliberation at scale
AI cannot grant itself—
- Democratic legitimacy
- Legal or constitutional authority
- Coercive public power
- Permission to experiment on people
- The right to override rights or due process
- The authority to certify its own institutional readiness
Govern the process itself
Power hides in more places than the final vote.
NousPolis explicitly separates three kinds of power so that invisible procedural choices do not quietly become sovereignty.
●
Substantive power
Who decides the answer?
Competing views and precommitted aggregation reduce oracle-like authority.↔
Procedural power
Who frames, routes, selects, aggregates, and stops?
Procedure is governed as a real institutional power.◇
Definitional power
Who decides which safeguards apply?
Materiality, credibility, urgency, and catastrophic-risk labels cannot silently switch protections off.Constitutional constraints
Modern intelligence needs enforceable boundaries.
NousPolis treats governance safeguards as institutional rules that must become machine-enforceable as maturity increases — not as good intentions buried in prompts.
“A rule is not safely enforced merely because an LLM prompt says it should be obeyed.”
Intake, classification, framing, routing, evaluation, review, and authorization remain distinct functions.
Claims retain source, method, dependence, assumptions, and evidence-state information.
Minority reports, appeals, external scrutiny, and review separation protect dissent.
AI recommendations do not silently become implementation authority.
Forecasts, releases, state changes, and governing commitments become reconstructable and tamper-evident.
Maturity is not a one-way ladder; serious failures can require suspension or regression.
Maturity & authority
Capability does not grant authority.
NousPolis advances only through explicit institutional gates. Better models or stronger benchmarks cannot substitute for law, legitimacy, security, ethics, independence, accountability, or public consent.
N0
NowArchitecture
Institutional design and hard entry criteria.
N1
BuildingResearch Laboratory
Executable safeguards; non-binding research.
N2
FutureAdvisory System
Structured voluntary recommendations.
N3
FutureExperimental Policy
Bounded authorized real-world pilots.
N4
FutureParticipatory Governance
Legitimate civic participation infrastructure.
N5
FuturePublic Decision Infrastructure
Integration with legitimate public institutions.
≠
Technical success is not political legitimacy.
No maturity transition is automatic, and AI cannot be the sole certifier of its own readiness.
Why now
Deep institutional intelligence may become abundant.
Governments have always faced hard limits on research capacity, expert availability, policy simulation, administrative analysis, continuous monitoring, and large-scale deliberation.
AI can change the economics of institutional intelligence. The opportunity is not merely to automate existing bureaucracy — it is to rethink governance around much greater reasoning capacity.
That requires institutions capable of governing the intelligence itself.
Scarce institutional intelligenceAbundant analytical capacity
More intelligence requires stronger institutional constraints.
An open institutional design problem
Help design governance for the intelligence age.
NousPolis needs criticism as much as support. We welcome governance researchers, political scientists, economists, philosophers, constitutional scholars, AI researchers, security engineers, civic technologists, policymakers, institutions, foundations, funders, and serious skeptics.