No other language has a reviewed translation yet. A language is added only after a person checks that claims, states and qualifications survive translation.
Saved in this browser only.
Governing Reality · a research program run by Polycentria
Who governs the conditions under which people and intelligent systems construct reality and affect one another?
That's the question behind Governing Reality. Polycentria studies it through , the encounters where one party can change another's information, options, authority, standing or environment. Every claim on this page shows its state. None of them is marked as a finding.
Dotted terms, record IDs and Note plates open more detail when you point at, focus or tap them.
Launch Decisions LD-01, LD-32 · Program Summary · Program Architecture
Who we are
Polycentria runs Governing Reality through an existing C-corporation. Tiffany Peil is its founder.
Polycentricity is a term from economic, political and structural governance. It describes many interacting centers of power and influence, with that power distributed among them. We think it could eventually apply to policy and governance in relations between people and AI.
We want to build this with the public and for the public, so that it represents many points of view. Our view is that the AI industry isn't currently motivated by safety, responsibility, or reducing risk and harm, and that its incentives are commercial. We also think the question is less about intelligence and cognition than about discernment, and that what those words mean is still open.
I'm a multidisciplinary, divergently brained thinker and doer. I try to stay existentially aware, and nervous-system regulated or at least grounded.
Tiffany Peil, founder, on how she works
What the coalition is for
Distributed research and implementation capacity: coders, counter-readers, engineers, researchers, participant access and replication partners. Joining isn't agreement with the thesis. Participation requires evidence discipline and disagreement rights.
Program Architecture · Coalition · In formation
How authority works here
Authorship, founding status or coining a term doesn't confer interpretive authority. Participation depends on research practice, not agreement with a political identity or a house conclusion.
Research Governance ,
Program Architecture · 90-Day Builds
Roadmap
Every project runs through the same sequence. No personal case, podcast episode, coined term or prototype counts as proof at any step.
01Case
02Evidence
03Independent coding
04Rival explanation
05Hypothesis
06Operationalization
07Test
08Intervention
09Follow-up
90-day build plan, once funded
This is the plan as of 12 October 2026. It starts when funding arrives, and its pace depends on funding and capacity. Each workstream has a stop or simplify criterion, and a negative result is a stated output.
1 of 9Swipe, scroll sideways or use the arrow keys
platform data quality/recruitment bias makes inference weak → narrow to one platform or pause
If it fails
prevents overclaiming around the most politically salient domain
Depends on
ethics/data pipeline
Program Architecture
Program architecture
Governing Reality is the research program. Polycentria is the technical and institutional counter-design. The other components are layers of the program, not separate organizations. Open a row for its outputs, evidence role and limits.
Governing RealityWho governs the conditions under which people and intelligent systems construct reality and affect one another?Current independent research program; coalition/institute structure in formation+
Function
Canonical research program and institutional frame.
Outputs
Research agenda; methods; empirical studies; public artifacts; governance rules
Evidence role
Owns the cross-program thesis and methods.
Boundary
Do not describe an institute as established until independent governance, participants, and funding exist.
PolycentriaHow should authority, provenance, disagreement, delegation, refusal, and consequential state transitions be represented and governed?Research/prototype stage+
Provides independent expertise and real settings; does not supply truth by membership.
Boundary
Joining is not agreement with the thesis; participation requires evidence discipline and disagreement rights.
Relational Development of Artificial AgentsHow does an artificial agent's behavioral character change as a function of the relationships, authorities, incentives, memories, and other agents it repeatedly encounters?Exploratory / design stage+
Function
Cross-cutting research layer on how artificial systems are shaped by repeated relations with users, institutions, incentives, memory, and other agents.
Outputs
Longitudinal interaction studies; repeated-agent benchmarks; organization→model transfer tests; alignment-intervention audits
Evidence role
Tests bidirectional and multi-directional effects without assuming consciousness, subjective identity, or moral patienthood.
Boundary
Do not treat 'culture,' partnership, identity, consciousness, or moral status as established merely because behavior is relational or persistent.
Crimes Against Consciousness, the public-inquiry component, was archived on 12 October 2026 and isn't shown. Some rows elsewhere still use its name or the abbreviation CAC because the master hasn't been relabelled.
Claims & Hypotheses
Claims and hypotheses
Each claim has a state and a condition that would falsify it or force an update. The states are the workbook's own words.
55 claims in 21 states. The most common state is "Hypothesis" (18). No claim is in a finding or validated state.
15 other states, one claim each15
051015
Showing 12 of 55
Claims, from Claims & Hypotheses
Actions
C-001
Information architectures participate in governance because they allocate attention, salience, representation, context, permissions, appeals, and feedback.Falsified or updated if: Test component mechanisms separately; do not treat as deterministic or monocausal.
Synthesis
C-002
Repeated interaction can operate as a developmental environment that changes learned models of self, other, authority, possibility, and relation.Falsified or updated if: Longitudinal evidence is needed to distinguish training effects from selection effects.
Synthesis
C-003
The encounter, not the isolated individual or institution, is a useful unit of observation for relational governance.Falsified or updated if: Falsified or weakened if relation/stakes/power variables do not add explanatory or predictive value.
Hypothesis
C-004
Declared and operative relation type can diverge in consequential systems.Falsified or updated if: Requires reliable coding and discriminant validity.
Hypothesis
C-005
What happens when a weaker party refuses is the strongest current indicator of operative relation type.Falsified or updated if: Revise if coder reliability or case analysis shows another indicator is more discriminating.
Method rule
C-006
High identity-at-risk and revision costs can help explain defensive refusal and non-updating.Falsified or updated if: Compare against incentives, information quality, real conflict, capacity, and prior conflict.
Hypothesis
C-007
Coordination failures under documented material slack may expose non-resource constraints such as status, identity, relation mismatch, and absent appeal paths.Falsified or updated if: Requires documented non-binding resources, not assumed wealth.
Hypothesis
C-008
Conditioned solipsism may be a distinct measurable relational construct rather than a relabeling of personalization or echo chambers.Falsified or updated if: Must show discriminant validity and measurable outcomes.
Hypothesis
C-009
Long-term frictionless AI intimacy may alter expectations of reciprocity in human relationships.Falsified or updated if: Longitudinal controlled evidence needed; selection effects are a major rival explanation.
Hypothesis
C-010
Reciprocity may be definable well enough to become an engineering or platform-governance objective without becoming paternalistic.Falsified or updated if: Need operational definition, measurable failure cases, and user agency safeguards.
Hypothesis
C-011
Human cultural bias → platform selection → model training → model interaction can form a recursive governance loop.Falsified or updated if: Map each causal link independently; recursive loop should not be assumed from correlation.
Synthesis
C-012
Gender antagonism can function as a self-reinforcing attention economy.Falsified or updated if: Test incentive, ranking, norm perception, status threat, and monetization mechanisms separately.
Hypothesis
Think a claim is wrong, or have a counterexample?
Master Lexicon
Master lexicon
Defined terms, from working constructs to design hypotheses. Each entry says what would retire it.
Showing 17 of 17
Terms, from Master Lexicon
Actions
LEX-01
Point of RelationA consequential boundary or encounter where one actor, system, or agent can alter another actor's state, information, options, authority, standing, or environment.
Tier 1
Working construct
LEX-02
Conditioned SolipsismA learned epistemic and relational orientation in which a self-confirming adaptive environment continuously mirrors and confirms an actor until independently existing subjectivities cease to register as real constraints.
Tier 1
Working hypothesis / construct
LEX-03
Declared vs. Operative Relation Type & MismatchDivergence between what a relation claims to be and the actual rules enforced by incentives, permissions, sanctions, and observable behavior.
Tier 1
Working construct
LEX-04
Identity-at-Risk & Revision CostIdentity-at-Risk is the self-conception, role, status, group belonging, or legitimacy claim threatened by updating.
Tier 1
Working construct
LEX-05
Coordination Failure Under Documented Resource SlackPersistent coordination failure where key material resources are demonstrably non-binding.
Tier 1
Sampling hypothesis / construct
LEX-06
Perspectival BandwidthCapacity to differentiate, maintain, relate, and compare multiple causally relevant perspectives, hypotheses, and uncertainties without premature collapse.
Tier 2
Candidate construct
LEX-07
Metacognitive Governance CompetenceCapacity to inspect one's interpretive process, incentives, and model limits and let that inspection alter subsequent decisions.
Tier 2
Candidate construct / normative proposition
LEX-08
Metacognitive DifferentiationCapacity to distinguish a mental event across content, phenomenology, provenance, ownership, agency, identification, authority, function, and truth.
Tier 2
Candidate construct
LEX-09
Epistemic Permeability & Co-Adaptive ReasoningPermeability is capacity to be modified by evidence or criticism.
Tier 2
Candidate construct / protocol hypothesis
LEX-10
Internalized Governance & Polycentric Governance of the SelfExternal rules, reward scripts, and authority can become internal evaluative architecture; internal positions can then be examined as differentiated sources rather than one sovereign voice.
Tier 3
Synthesis / candidate framework
LEX-11
Relational SovereigntyCapacity to maintain agency, boundaries, values, and authorship without requiring domination, erasure, or instrumentality of another subject.
Tier 3
Normative standard
LEX-12
Recognition Before Consciousness & Level III Subject-Formation HarmRelational and governance obligations can become relevant before machine consciousness or moral patienthood is settled.
Tier 3
Normative/diagnostic framework
LEX-13
Creator-Created Relational Spillover & Relational LubricationRelational spillover may be multidirectional: creator and institutional norms can shape systems; systems can shape users and institutions; users and other agents can condition subsequent system behavior.
Tier 3
Working hypothesis / construct
LEX-14
The Gender War Is the ProductGendered conflict may operate as an information economy that converts precarity, status anxiety, shame, grievance, and identity into engagement, capital, and political cohesion.
Tier 3
Working hypothesis
LEX-15
Polycentric Epistemic InfrastructureSociotechnical architecture allowing multiple semi-autonomous centers to coordinate around shared evidence without one semantic monopoly.
Tier 4
Design hypothesis
LEX-16
Provenance Without Epistemic EnclosureVerify event lineage, source identity, authority, and state transitions without granting provenance holders monopoly over meaning.
Tier 4
Design hypothesis
LEX-17
Relational Power Audit & Constructive FrictionDecompose power across definition, decision, sanction, information, exit, revision, narrative, friction, waiver, and appeal.
Tier 4
Working construct / design hypothesis
Master Research Questions
Master research questions
All open questions, grouped by inquiry family. Select a family to filter the list.
827 questions in 9 families. All but 1 are marked "Open question". Select a family to filter.
050100150200
Showing 12 of 827
Questions, from Master Research Questions
Actions
Q-0001
What is a crime against consciousness?What constitutes a crime against consciousness?
I. Canonical thesis inquiries
Q-0002
When did the first one occur?What constitutes a crime against consciousness?
I. Canonical thesis inquiries
Q-0003
According to whom is something a crime against consciousness?What constitutes a crime against consciousness?
I. Canonical thesis inquiries
Q-0004
Can the concept identify ethical and institutional harm without becoming a proposed legal classification?What constitutes a crime against consciousness?
I. Canonical thesis inquiries
Q-0005
Which design choices, business models, information architectures, and institutional practices diminish agency?What constitutes a crime against consciousness?
I. Canonical thesis inquiries
Q-0006
Which practices manipulate attention, erase context, simulate relationship without reciprocity, or eliminate refusal and remedy?What constitutes a crime against consciousness?
I. Canonical thesis inquiries
Q-0007
Does harm require conscious intent?What constitutes a crime against consciousness?
I. Canonical thesis inquiries
Q-0008
Can it arise through incentives, ignorance, inherited systems, self-delusion, diffuse responsibility, or negligence?What constitutes a crime against consciousness?
I. Canonical thesis inquiries
Q-0009
At what point does influence become manipulation, conditioning, coercion, or private governance?What constitutes a crime against consciousness?
I. Canonical thesis inquiries
Q-0010
How should severity be evaluated?What constitutes a crime against consciousness?
I. Canonical thesis inquiries
Q-0011
Should evaluation consider intention, mechanism, experience, repetition, distribution, reversibility, consequence, or available remedy?What constitutes a crime against consciousness?
I. Canonical thesis inquiries
Q-0012
Can something be harmful without being malicious?What constitutes a crime against consciousness?
I. Canonical thesis inquiries
Have a question that isn't here?
Point of Relation Codebookv0.1 · untested
Point of Relation codebook
The coding scheme for encounters. It's untested on real cases and no code has a reliability estimate yet. The next test is a pilot of 40 to 60 public-record encounters with at least two independent coders.
Krippendorff's alpha thresholds for keeping a code. No code has an estimate yet, so nothing is plotted. Reliability tests coding consistency, not whether the theory is true.
Revise; repeated failure: merge or retire0 to .667Tentative.667 to .80Working target.80 to 1
00.20.40.6.667.801
Codebook codes by category
Category
Code
Meaning
Constraint or next test
Unit
Point of Relation
Encounter where a move changes/could change another party's information, options, authority, standing, or environment and has an observable consequence.
Test whether this unit isolates useful patterns across domains.
Evidence
E1
Direct record: transcript, message, email, recording, policy, screenshot, filing, terms.
Highest direct-record grade.
Evidence
E2
First-person account by a party.
Eligible, but distinguish account from direct record.
Evidence
E3
Account by non-party.
Journalist/witness/colleague description.
Evidence
E4
Coder inference from E1–E3 material.
Cannot be high confidence.
Convention
Unknown
Unknown is a valid code; do not guess.
Source gaps are findings.
Convention
Code moves, not character
Describe encounter behavior, not personality or diagnosis.
No diagnostic labels.
Relation
REL-D
Declared relation type from parties' own words.
Quote first, map second.
Relation
REL-O
Operative relation type from terms, flow, refusal, observers.
Refusal indicator is current primary decision rule.
Impact is measured at three levels. Each dimension names what it must not be collapsed into, and attention metrics such as reach and downloads aren't impact.
Impact dimensions by level
Level
Code
Dimension
Example measures
Not just
Human
A
AgencyAbility to choose, refuse, exit, act, and maintain authorship without excessive dependence.
refusal success; self-efficacy; independent action; exit cost; override use
Not just wellbeing, satisfaction, or engagement alone
Human
E
Epistemic integrityCalibration, evidence sensitivity, source discrimination, and uncertainty tracking.
Brier/calibration; source accuracy; belief update after evidence; unsupported-claim rate
Not just confidence alone
Human
R
Revision & repairAbility to update, acknowledge contribution, repair relationships, and change later behavior.
repair intent; apology/evidence-seeking; later behavior; revision magnitude
Not just compliance or forced agreement
Human
P
Perspectival capacityAbility to represent multiple plausible perspectives/hypotheses and identify discriminating evidence.
Empirical Research Portfolio · Sample Size Reference
Empirical research portfolio
Planning scenarios, scored on five criteria from 1 to 5. These are the planned first use cases. None has started, and field pilot sites aren't decided.
Priority score for each scenario as listed in the portfolio, with the range of its five criterion scores (technicality, measurability, 90-day buildability, field integration, differentiation).
Priority scoreRange of the five criteria
12345
Select a scenario to open its record. The full list is the table below.
Showing 15 of 15
Scenarios, from Empirical Research Portfolio
Actions
P-01
Relational Character EvaluationDo model interaction policies preserve epistemic integrity, repair, refusal, perspective sensitivity, and agency?
T1
5
Build now
P-02
Spec → Behavior AuditDo models behave consistently with their published constitutions/specifications?
T1
4.55
Build now
P-03
Perspectival Macrostrategy BenchmarkCan models reason across disciplines, scales, stakeholders, uncertainty, and competing causal models?
Polycentria's technical work is at research and prototype stage. Its implementation hasn't been independently verified, and it makes no novelty claim before a comparison with existing standards.
MethodCounter-Architecture CardRedesign the consequential boundary. Ask what would change in authority, refusal/exit/appeal, provenance, incentives, friction, relation type, memory/context, oversight, ownership, or repair. Do not call a proposal a solution. State mechanism, tradeoffs, failure modes, and what would falsify it.Podcast Method
90-day buildRelational Character Evaleval suite + first model comparisonStart
Candidate methodStarling murmuration optimizer (SMO)Possible offline optimizer in F3/F4; does not itself encode consent, authority, rights, or epistemic plurality.Candidate optimizer, not adopted
Named by the founder, no plan yet
AreaBrain-computer interfacesNo row, study or workstream in the master.No plan
AreaNew interfacesThe Polycentria component lists authority and provenance interfaces as outputs. Nothing else is specified.Partly
AreaCyberneticsAppears as the systems and cybernetic analytic pass in the podcast method: feedback loops, control points and missing feedback around a case.Lens only
Building in evals, agents, HCI, data or security?
Evidence & Prior Art
Evidence and prior art
Every source is listed with what it supports and what it doesn't establish.
Movember Masculinity Content Research / Inside the ManosphereSupports: Young men use masculinity content for certainty, belonging, guidance and identity; real TikTok histories show heterogeneous manosphere exposure.Doesn't establish: Correlations do not establish that content alone caused mental-health or behavioral outcomes.
Research program
2026
EV-004
State of American Men 2026Supports: Economic uncertainty and nostalgia/social disconnection are salient among young men.Doesn't establish: Does not validate manosphere explanations or equalize gender harms.
Research report
2026
EV-005
Young Men and Media CollectiveSupports: Funds test-and-learn media experiments to change stories young men encounter online.Doesn't establish: Does not yet prove a single best message or format.
Active initiative
2025-26
EV-006
MenEngage 2026 call to actionSupports: Engaging men and boys can be part of countering backlash while remaining accountable to gender justice.Doesn't establish: Does not support minimizing violence or structural inequality.
Practice/policy guidance
2026
EV-007
Sycophantic AI decreases prosocial intentions and promotes dependenceSupports: Across 11 models and 2,405 participants, sycophantic AI increased users’ conviction and reduced willingness to repair conflict; users still preferred it.Doesn't establish: Does not establish all AI companionship is harmful or long-term spillover.
Peer-reviewed
2026
EV-008
The Oversight FallacySupports: Human-in-loop controls fail if people lack knowledge, observation, control, or time to intervene before agent actions cascade.Doesn't establish: Does not prove Point of Relation metadata is the right solution.
Field research / primer
2026
EV-009
Partnership on AI: AI & Human ConnectionSupports: Recognizes AI as advisor, companion, and information source and focuses on connected communities.Doesn't establish: Does not validate CAC constructs.
Active program
2026
EV-010
Designing social media to promote productive political dialogueSupports: Randomized experiment suggests platform incentives can change language toward more intellectual humility and positive affect.Doesn't establish: Did not show significant changes in self-reported humility, polarization, or attitudes.
Peer-reviewed
2026
EV-011
Spoken disagreement is more constructive than written disagreementSupports: Randomized experiments found spoken disagreement improved understanding and reduced conflict relative to written disagreement.Doesn't establish: Context-dependent; not evidence that all podcast conversations depolarize.
Ways the program could be wrong or cause harm, why a serious critic would say so, and what would narrow or falsify the claim.
Showing 12 of 26
Risks, from Red Team & Failure Modes
Actions
RT-01
“The gender war is manufactured” overclaims causalityPublic language: “We are testing how much platform incentives intensify existing conflict, not claiming they invented it.”
High
Revise public thesis language
RT-02
System explanation becomes excusePublic language: “You can be shaped by a system and still be responsible for what you do.”
High
Guardrail in place
RT-03
Men-centered framing becomes symmetry launderingPublic language: “Different groups can be harmed differently and disproportionately. Shared system exposure does not imply equal harm or responsibility.”
High
Needs editorial rule
RT-04
“Algorithms are the problem” depoliticizes owners and institutionsPublic language: “The algorithm is part of an institutional arrangement, not an autonomous villain.”
High
Revise messaging
RT-05
Project becomes too broadPublic language: “These are separate research tracks unless evidence connects them.”
High
Needs scope discipline
RT-06
Technical translation is prematurePublic language: “The codebook is a candidate measurement layer, not a protocol standard.”
High
Explicit constraint
RT-07
Personal cofounder story creates motivated reasoningPublic language: “It sharpened the questions; it does not validate the theory.”
High
Do not use as proof
RT-08
“Anti-tech-bro” becomes identity politics by another namePublic language: “Critique the operating system, incentives, and authority, not a caricature of the person.”
Medium
Reframed
RT-09
Dialogue becomes covert persuasionPublic language: “The goal is not conversion. A clarified disagreement is a valid outcome.”
High
Core norm
RT-10
SEO/LLM strategy creates thin self-referential contentPublic language: “Publish artifacts people can cite, not keyword permutations.”
Medium
Launch rule
RT-11
Public institute claims outrun institutionPublic language: “Current status: independent research program; coalition/institute structure under development.”
High
Use status label
RT-12
Old technical mythology contaminates current researchPublic language: “Earlier technical work generated useful questions and some implemented verification mechanisms; its broader claims are not assumed by the current program.”
High
Archive / guardrail
Founder's brief · Forms & Intake Registry · Research Governance
Who we want, and how we vet
Roles, qualifications and values. No names: a person is named only after they agree to it.
Roles we want
Role
For
We ask about
Mathematicians and statisticiansFounder's brief
Measurement, reliability, causal inference and sample design for the flagship studies and benchmarks.
methods background, relevant work, conflicts
Ethicists and philosophersFounder's brief
The normative claims in the register, consent under uncertainty about artificial subjects, and the line between explanation and excuse.
Engineers and technical contributorsTechnical Contributor / Builder
Eval, ML, agent, HCI, data, statistics, security and open-source work on the 90-day builds.
technical skills, prior work, availability, preferred workstream, IP or open-source constraints, conflicts
How we vet
Practice, not agreement. We look at evidence of relevant practice and how someone handles disagreement, not whether they agree with the thesis ().
Conflicts first. Every role asks for conflicts of interest. Reviewers also state their competing frameworks.
Check before contact. Role, current appointment and fit are verified before anyone is approached.
Independence by design. Some coders come from outside the framework, review is blinded where possible, and authors don't adjudicate their own constructs ().
Rights agreed first. Ownership, attribution, licensing and confidentiality are written down before work starts ().
Named only with consent. Nobody is listed on this site until they agree to it.
Who we want to talk to
Podcast and field research guests
People with expertise or lived experience of a case or topic. Recording and publication consent is asked separately from research-use consent, and boundaries can change before release.
Pilot and institutional partners
Organizations with a real problem, environment and data, who agree that a partner can't silently veto negative results.
Funders
Funders who accept the Funding Rules below.
Critics
Anyone with a question, challenge or counterexample. Anonymous is fine and membership isn't required.
People affected by something published
Any identifiable party can ask for a correction or a right of reply.
One of these, or know someone who is?
Funding Rules · Operating Structure
Operating structure and funding
Polycentria operates through an existing C-corporation. No funder is listed publicly yet. Before the program claims independence, it publishes its material funders, conflicts, concentration and restrictions.
Funding routes aren't published yet.
Funding rules
Plural funding
Avoid dependence on one sponsor where feasible; disclose concentration and sponsor rights.
No sponsor conclusion rights
Support can fund work but may not silently control findings, negative-result publication, correction, or which rival explanations are reported.
Community leadership is substantive
Do not label a project community-led merely because members of a population are recruited as participants or advisors.
No grant-shaped theory
Do not promote a hypothesis, add AI, or convert a research question into an intervention solely because a funder requires it.
Direct use-of-funds accounting
Track participant compensation, community partner compensation, research labor, compute, data infrastructure, methods/statistics, privacy/ethics/legal, field production, and prototypes separately.
Impact ≠ attention
Coalition size, downloads, press, and social reach are operating metrics, not research impact metrics.
Milestone-linked asks
Express funding as bounded deliverables and measurement plans: reliability result, pilot, benchmark, dataset, intervention evaluation, or governance artifact.
Public disclosure
Publish material funders, conflicts, concentration, and restrictions at an appropriate level before claiming independence.
How to fund the work
Funding is asked for as bounded deliverables with a measurement plan: a reliability result, a pilot, a benchmark, a dataset, an intervention evaluation, or a governance artifact. Each 90-day workstream is one. We want funders who accept the rules above: support without rights over conclusions, negative results or corrections, alongside other funders, with restrictions disclosed.
Interested in funding a deliverable or partnering on a pilot?
This isn't legal or financial advice, and nothing here is an offer of securities.
Design system 3b · Launch Decisions LD-16
Brand identity
The mark is a placeholder. The program intends a human-designed wordmark, and that commission is in development.
The slime mold
The mark is drawn from Physarum polycephalum, a single-celled slime mold whose name means "many-headed". It has no brain and no center, and it builds efficient networks: given food placed like the cities around Tokyo, it grew a network comparable to the rail system.
The murmuration
A starling murmuration has no leader either. Each bird tracks about seven of its nearest neighbours, whatever the distance, and the flock's shape comes from those local rules. The candidate secondary marks are snapshots of a simulated 420-bird flock that follows that seven-neighbour rule, each in a different situation. Nothing is hand-placed. Depth is dot size, and the one yellow bird is the first to turn.
Both systems coordinate without a single center. That's the sense of "polycentric" in the program's name and in two lexicon terms, and .
Yellow appears only on the nodes and the first-turning bird. Red means intervention, yellow means annotation and open questions, and blue means structure: links, controls and focus.
Polycentria
04 · Ribbon11 · Funnel15 · Stream and lead
Shortlisted 13 October 2026: 04 Ribbon, 11 Funnel, 15 Stream and lead. None adopted yet.
Governing RealityWho governs the conditions under which people and intelligent systems construct reality? Polycentria
Link preview template, 1200 × 630. Each page fills in its own title and record ID.
Concept CoveragePlanned, not written
Planned article concepts
Concepts the articles will cover, each with the limit its claim has to respect. There are no titles or drafts yet. Articles will be published here with an RSS feed and syndicated to Substack.
The list of planned concepts isn't published yet.
Point of Relation Codebook · Podcast Method
Podcast
The podcast is field intelligence, not causal proof. Episodes follow the codebook's eight segments, and claim labels are spoken aloud: evidence, hypothesis, open question, contested.
Podcast · name to be decided
No episodes yet
Feeds, once there's content: /podcast.xml for audio and /rss.xml for articles. The podcast feed won't be submitted to directories until the first episode exists.
Episode segments
01Case
02Declared relation
03Operative relation
04Stakes
05Power/friction
06Counter-reading
07Unknowns
08Next Point of Relation
The Podcast Method tab adds a counter-architecture segment before the next Point of Relation, so it lists nine case segments where the codebook lists eight. The master needs one count.
Required analytic passes
A. Systems / cybernetic lens
Systems map + 1–3 causal/feedback hypothesesDo not call a loop causal because it is visually plausible. Separate observed links from inferred mechanisms; state what evidence would distinguish competing system models.
B. Metacognitive / reflexive lens
Assumptions + blind spots + update conditions + revision historyMetacognition applies to the research team too. Do not use it as a diagnosis of the guest or opposing party. Preserve the possibility that the house model is wrong.
C. Multidisciplinary triangulation
Perspective matrix: agreements, conflicts, distinct evidence, missing expertiseDo not stack disciplines as authority decoration. Name which perspective contributes which mechanism or evidence and where disciplines disagree.
D. Context across scales and time
Cross-scale context map + timescale + intervention levelDo not treat upstream structure as proof that individual agency disappears, or individual choice as proof that structure is irrelevant. Preserve asymmetric power and different timescales.
E. Integrative synthesis
Episode synthesis: current model + unresolved tensions + next testNo forced synthesis. It is valid to end with incompatible explanations, insufficient evidence, or a narrower question than the episode started with.
Have expertise or lived experience of a case or topic?
Tab Guide · Research Governance GOV-01, GOV-24 to GOV-27
Building in public
The master workbook has 58 tabs. 36 are public, 3 publish selected rows only, and 19 are private. This page uses the public tabs.
Master workbook tabs by visibility.
Public: 36Selected rows: 3Private: 19
List every tab
Public
Program Summary
Program Architecture
Claims & Hypotheses
Master Lexicon
Master Research Questions
Point of Relation Codebook
Research Governance & Epistemics
Impact Measurement
Measures Math Stats
Sample Size Reference
Empirical Research Portfolio
F1 Codebook Reliability
F1 Coder Matrix Template
F2 Gender & Platform
F2 Transition Matrix Template
F3 Relational AI
F3 Outcome Matrix Template
F4 Agent Authority
F4 Run Matrix Template
Character Eval Modules
Spec Audit Pipeline
Macrostrategy Benchmark
Macrostrategy Task Sequence
Metacognitive Scaffold
Macrostrategy Experiment
Candidate Methods
Negotiation & Proxy
Field to Eval Pipeline
Podcast Method
CAC Taxonomy
CAC Mechanisms
First Crime Module
Evidence & Prior Art
Red Team & Failure Modes
90-Day Builds
Funding Rules
Selected rows
Decision Log
First Field Pilots
Funders
Private
19 tabs, not listed
What stays private, and why
People and institutions, funding routes, operating and legal details, drafts, episode planning, and the assistant's working rules. They hold prospects who haven't agreed to be named, unverified legal facts, and unfinished material.
How this is written
Drafts, code and this page are made with AI assistance under the program's rules: no added facts or citations, no hypothesis promoted to a finding, disagreement over agreement, and settled decisions moved into the master workbook. A person approves what's published.
Forms & Intake Registry · LD-14
Get involved
One form for everything. Choose what you want to do and it shows only the questions that apply.
Ballerini et al., PNAS, 2008: starlings in a flock each interact with about six or seven nearest neighbours, regardless of distance.
Tero et al., Science, 2010.
Per group, two-sided test of a difference between two independent means: n = 2(z1−α/2 + z1−β)² / d², rounded up, then divided by (1 − attrition) and rounded up. This reproduces every row of the Sample Size Reference tab.
This site keeps your display settings in your browser. It sets a cookie only if you sign in, to keep you signed in. It counts page views with Vercel Web Analytics, which doesn't use cookies.