Collective Thinking Process Protocol

Integrating the Uptake Integrity Protocol with collective agency, governance engineering, and loop engineering

Version: 0.1 Status: Design manual and candidate operating standard Audience: Human teams, AI-agent systems, mixed human–AI collectives, organizations, institutions, federations, and coordinated groups acting toward a shared goal

1. Purpose

Collective action is often represented as a cycle of:

discuss → decide → assign → act → review

That cycle is necessary but incomplete.

It does not reliably distinguish:

a shared observation from a shared interpretation;
agreement about a goal from agreement about what the goal means;
acceptance of a claim from commitment to act on it;
dissent from lack of commitment;
a consequence from evidence;
evidence from the collective procedure that allows it to count;
formal approval from executable authority;
a locally successful action from movement toward the collective ideal;
temporary coordination from durable collective learning.

The Collective Thinking Process Protocol, or CTPP, supplies these distinctions.

Its purpose is to enable a plurality of agents to collaborate toward a shared or overlapping ideal while preserving:

differences in perspective;
differences in knowledge;
differences in affectedness;
differences in capability;
differences in authority;
differences in commitment;
the ability of evidence to defeat the collective model;
the ability of legitimate dissent to revise the goal, means, or governance process.

The protocol treats the collective’s trees as versioned, losable models rather than expressions of a fictitious unified mind.

The resulting architecture has three layers:

Layer	Primary question
Participation engineering	How can heterogeneous agents contribute claims, needs, commitments, objections, and capabilities without being falsely homogenized?
Governance engineering	What roles, records, permissions, safeguards, and decision rules make collective inquiry answerable and executable?
Loop engineering	How should collective observation, modeling, conflict resolution, probing, evidence uptake, revision, and retention unfold over time?

The relationship is:

Participation engineering preserves plurality and makes contributions legible. Governance engineering makes the process authoritative, bounded, and answerable. Loop engineering makes collective learning recur without becoming circular or self-sealing.

Participation without governance produces discussion without reliable action.

Governance without plural participation produces administratively coherent closure.

A loop without either can repeatedly manufacture confirmation for the collective’s incumbent interpretation.

2. Collective thinking as semiotic conduct

A collective does not require a single collective mind.

It may consist of:

individual humans;
AI agents;
organizational units;
committees;
companies;
public agencies;
communities;
professional groups;
autonomous teams;
federated systems;
other collectives represented through authorized agents.

A collective becomes capable of conduct when it possesses a persistent integration of:

distributed interpretations;
shared or interoperable records;
coordinated actions;
recognized commitments;
decision procedures;
revision authority;
means of retaining change.

The formal theory of semiotic conduct describes collective agency without positing mental homogeneity: collective conduct can arise through distributed codes, shared memory, coordinated response, gates, and revision authority.

A collective encounters signs:

events;
complaints;
metrics;
reports;
forecasts;
incidents;
stakeholder testimony;
policy outcomes;
task failures;
messages from other agents;
changes in the environment.

Different participants may interpret the same sign differently because they possess different:

vocabularies;
experiences;
incentives;
values;
models;
responsibilities;
access to evidence;
exposure to consequences.

Those interpretations generate proposals, objections, commitments, and actions.

The actions co-produce consequences with:

other participants;
non-participating stakeholders;
material constraints;
institutions;
technologies;
chance;
adversarial actors;
environmental conditions.

The consequences do not become collective evidence automatically.

They pass through:

selection;
classification;
attribution;
registration;
criticism;
adjudication.

The collective traversal is therefore:

World conditions and stakeholder signs
        ↓
Distributed interpretations
        ↓
Proposals, needs, claims, conflicts, and commitments
        ↓
Collective model and action selection
        ↓
Coordinated and uncoordinated conduct
        ↓
Material and social consequences
        ↓
Plural evidential uptake
        ↓
Collective criticism and adjudication
        ↓
Revision of model, commitments, governance, action, or ideal

The collective is not defined by unanimity.

It is defined by whether differentiated agents can produce coordinated conduct under explicit and revisable terms.

3. The central object: a losable collective claim

A collective process should not begin with:

“We agree that this is the problem and this is the solution.”

It should begin with:

“This is the current shared or provisionally adopted model, these are the unresolved differences, and these are the consequences that could cause the collective to revise it.”

A collective claim is a claim entered into the shared model through a recognized procedure.

Collective entry does not imply that every participant personally believes it.

The protocol distinguishes:

individual acceptance;
collective procedural acceptance;
normative endorsement;
action commitment;
resource commitment;
authority to bind others.

Examples of collective claims include:

Observation claim

Service interruptions occur most frequently during cross-region failover.

Classification claim

These interruptions constitute unacceptable reliability failures.

Causal claim

Delayed authority transfer contributes to failover instability.

Necessity claim

A stable failover process requires one clearly authorized coordinator at any moment.

Sufficiency claim

A shared state ledger and deterministic transfer rule are sufficient to prevent conflicting commands under the defined failure modes.

Conflict claim

Rapid local response and centralized coordination currently appear to require incompatible operating policies.

Need claim

Local teams require discretion to prevent immediate harm.

Assumption claim

If local teams may act without prior central approval, system-wide consistency will be lost.

Intervention claim

A bounded local-authority protocol will preserve rapid response while maintaining eventual coordination.

Risk claim

Local authority may permit inconsistent actions across regions.

Measurement claim

Incident duration is an adequate indicator of failover quality.

Normative claim

No improvement in recovery time may be purchased by concealing incidents from affected parties.

A collective claim becomes losable when:

meaningful alternatives may be expressed;
relevant stakeholders may challenge it;
a discriminating probe is possible;
traces survive collective uptake;
adverse evidence may count;
the governing model can be revised;
authorized conduct can change;
warranted revision can persist.

4. Mapping LTP artifacts into collective thinking

The ideal-directed theory treats the Logical Thinking Process as a control-design grammar. It maps the Goal Tree to a provisional ideal, the Current Reality Tree to a model of persistent conduct, the Cloud to a policy conflict, assumptions to hypotheses about operative codes, injections to candidate interventions, and the Future Reality Tree to prospective causal simulation.

In the Collective Thinking Process Protocol, the mapping becomes:

LTP artifact	Collective role
Goal Tree	Provisional representation of the shared ideal, necessary conditions, protected floors, and critical success factors
Current Reality Tree	Shared causal model of recurring undesirable effects across participants, roles, scales, and institutions
Evaporating Cloud	Structured representation of a conflict among legitimate collective requirements and apparently incompatible actions
Hidden assumptions	Claims about forecasts, classifications, values, authority, feasibility, incentives, identity, trust, or interests
Injection	Candidate collective action, governance change, resource allocation, probe, rule, capability, or institutional redesign
Future Reality Tree	Shared theory of how the injection may propagate through the collective and its environment
Prerequisite Tree	Obstacles involving capability, consent, authority, resources, legitimacy, coordination, or sequencing
Transition Tree	Distributed action plan specifying who acts, under what conditions, with what expected intermediate effect
Strategy and Tactics Tree	Multi-level alignment between purposes, strategic commitments, local tactics, and necessary assumptions

The trees do not erase disagreement.

They locate it.

A collective tree may contain:

accepted entities;
contested entities;
minority branches;
unresolved rival explanations;
local perspectives;
alternative Clouds;
conditional commitments.

The smallest adequate representation should be used.

A two-person collaboration may need one Cloud and a Test Contract.

A federation of organizations may require a family of linked trees with separate ownership and cross-scale gates.

5. Three engineering disciplines

5.1 Participation engineering

Participation engineering determines how heterogeneous agents enter the process.

It defines:

identity;
constituency;
standing;
role;
affectedness;
expertise;
authority;
capability;
commitment;
rights of challenge;
rights of appeal;
rights of exit.

Participation engineering protects against the false statement:

“The collective believes X.”

Instead, it permits more precise statements:

X is the current procedural basis for action. Agents A and B endorse X. Agent C accepts X provisionally but disputes its scope. Agent D does not endorse X but commits to action Y under safeguard Z. Agent E objects because protected floor F is threatened.

Participation engineering is the grammar of differentiated membership.

5.2 Governance engineering

Governance engineering supplies the external structure through which collective reasoning becomes reliable and actionable.

It includes:

records;
procedures;
permissions;
agenda control;
decision rights;
evidence custody;
role separation;
quorum rules;
appeal;
review;
safeguards;
resource authority;
enforcement;
retention.

Governance engineering is the institutional implementation of the protocol.

It prevents the collective from assuming that good reasoning alone produces:

legitimate decisions;
executable commitments;
preserved evidence;
durable revision.

5.3 Loop engineering

Loop engineering organizes collective inquiry across time.

It defines:

when observation is open;
when claims enter the shared model;
when conflicts require a Cloud;
when the collective may act;
how test contracts are established;
how evidence returns;
how disagreement is adjudicated;
when re-abduction is required;
how commitments are updated;
when the ideal must be reviewed;
when the collective pauses, exits, or dissolves.

Loop engineering is the temporal control structure of collective thinking.

6. Why the three layers must remain separate

Suppose a collective gives every participant a right to speak, but the governance process permits only leadership to register evidence.

Participation exists, but uptake remains closed.

Suppose evidence is preserved and deliberation is open, but no participant has authority or resources to implement the decision.

Inquiry exists, but executive revision is impossible.

Suppose authority exists, but the loop repeatedly returns every failure to the same favored explanation.

The collective acts, but it does not learn.

The architecture therefore requires:

Participation:
    determines who may contribute, challenge, commit, and withdraw

Governance:
    determines how claims, evidence, decisions, and authority are handled

Loop:
    determines how the process moves from uncertainty to action and revision

No layer can substitute for a zero in another.

Part I — Building the Protocol

7. The Collective Thinking Process Protocol

The Collective Thinking Process Protocol is a reusable constitution for goal-directed collective inquiry and action.

Its purpose is:

To enable heterogeneous agents to form, test, revise, and execute shared models of movement toward an ideal while preserving plurality, affected-party standing, protected needs, explicit commitments, and collective losability.

The protocol governs four distinct objects:

The shared ideal
The shared model
The commitment structure
The action-and-evidence loop

These must not be collapsed.

A collective may share an ideal but disagree about the model.

It may accept a model but lack commitment to act.

It may commit to act while retaining doubts about the model.

It may agree on action for different reasons.

The protocol records these distinctions rather than forcing premature consensus.

8. When the protocol should activate

The protocol should activate when one or more of the following applies:

multiple agents are stakeholders in the same goal;
participants understand the goal differently;
participants bear different consequences;
commitment levels vary;
authority is distributed;
implementation crosses organizational or technical boundaries;
important evidence is controlled by different parties;
a persistent conflict blocks coordinated action;
the collective has experienced repeated failed interventions;
the goal is normatively contested;
local improvement may create cross-scale harm;
trust is insufficient for informal coordination;
the decision is costly, irreversible, or high risk.

The protocol supports operating tiers.

Tier	Collective situation	Required protocol
0 — Informal coordination	Low-risk task with clear goal and authority	Shared objective, assignments, completion check
1 — Local collaboration	Small group with limited disagreement	Charter, claims, commitments, review
2 — Plural inquiry	Competing interpretations or uncertain cause	Claim portfolio, Cloud, Test Contract, evidence ledger
3 — Distributed action	Multiple teams, organizations, or AI agents	Full governance, dependency map, commitment ledger
4 — High legitimacy	Public, safety-critical, coercive, or rights-affecting decision	Independent uptake, affected-party standing, appeal, external review

9. Protocol invariants

9.1 No fictitious collective mind

The protocol must not convert procedural acceptance into universal belief.

It records:

who accepts;
who objects;
who abstains;
who commits;
who is bound;
under what authority.

9.2 No unrepresented affected party where representation is feasible

Agents who bear material consequences must have:

direct standing;
delegated representation;
an independent advocate;
or an explicit explanation for their absence.

9.3 No pivotal claim without visible status

Every pivotal entity and arrow must be marked as:

observed;
proposed;
contested;
in probation;
supported;
qualified;
defeated;
normatively unresolved.

9.4 No consensus by ambiguity

Participants must not be recorded as agreeing merely because they use the same word.

Terms such as:

safety;
fairness;
reliability;
autonomy;
efficiency;
quality;
accountability;
success

must be operationally clarified where disagreement matters.

9.5 No conflation of acceptance and commitment

The following are distinct:

“I believe this claim.”
“I accept this claim as a working basis.”
“I endorse this goal.”
“I authorize this action.”
“I commit resources.”
“I will execute this task.”
“I accept the defined risk.”

9.6 No collective action without attributable authority

Every action must identify:

who may authorize it;
who will perform it;
who bears the cost;
who bears the risk;
who may stop it;
who is accountable for the result.

9.7 No pivotal test without possible collective loss

At least one admissible result must be capable of:

weakening the collective model;
favoring a rival;
changing the proposed action;
or reopening the ideal.

9.8 No consequence silently converted into shared evidence

Selection, classification, attribution, and registration must remain inspectable.

9.9 No majority immunity

A numerical majority does not make a causal claim true, erase a protected floor, or invalidate evidence.

9.10 No dissent without disposition

Substantive objections must be:

incorporated;
tested;
answered;
recorded as unresolved;
or escalated.

They may not simply disappear from the shared record.

9.11 No repeated collective action without changed information

After repeated failure, the collective must revise:

the claim;
the intervention;
the uptake process;
the authority structure;
or the ideal.

9.12 No completed decision without retention

A decision is not complete until:

commitments are registered;
authority is assigned;
resources are available;
monitoring exists;
revision conditions are preserved.

10. Protocol inputs and outputs

Inputs

The protocol may receive:

stakeholder descriptions;
current goals;
institutional mandates;
problem reports;
observations;
testimony;
datasets;
existing policies;
commitments;
contracts;
authority structures;
technical models;
constraints;
prior attempts;
applicable rights and protected floors.

Required outputs

Depending on operating tier:

Collective Ideal Charter;
Participant and Standing Register;
Goal Tree;
Current Reality Tree;
Evaporating Clouds;
Claim Register;
Commitment Ledger;
Authority and Capability Matrix;
Collective Test Contracts;
Evidence Ledger;
Decision and Dissent Log;
Revision Deltas;
Dependency Map;
Transition Tree;
Retention Plan;
Residual Disagreement Report.

11. Recommended protocol package

collective-thinking-process/
├── constitution/
│   ├── collective-ideal-charter
│   ├── protected-floors
│   ├── standing-rules
│   ├── decision-rules
│   └── appeal-and-exit-rules
├── participants/
│   ├── participant-register
│   ├── constituency-map
│   ├── commitment-profiles
│   └── authority-capability-matrix
├── trees/
│   ├── goal-tree
│   ├── current-reality-tree
│   ├── evaporating-clouds
│   ├── future-reality-tree
│   ├── prerequisite-tree
│   └── transition-tree
├── inquiry/
│   ├── claim-register
│   ├── rival-register
│   ├── test-contracts
│   ├── evidence-ledger
│   └── verdict-register
├── governance/
│   ├── agenda-log
│   ├── decision-log
│   ├── dissent-log
│   ├── commitment-ledger
│   └── revision-authority-register
├── revision/
│   ├── revision-deltas
│   ├── dependency-impacts
│   ├── unresolved-surprises
│   └── retention-plan
└── audit/
    ├── closure-audit
    ├── participation-audit
    ├── legitimacy-audit
    └── protocol-evaluation

These may be:

documents;
databases;
signed messages;
issue trackers;
append-only logs;
graph structures;
smart contracts;
shared workspaces;
machine-readable artifacts.

The protocol does not depend on one implementation medium.

12. Collective constitution

A compact collective constitution should state:

COLLECTIVE THINKING PROCESS CONSTITUTION

Collective identifier:
Shared or overlapping ideal:
Operating scope:
Participating agents:
Affected non-participants:
Protected floors:

Who may introduce observations:
Who may introduce hypotheses:
Who may challenge classifications:
Who controls evidence registration:
Who may authorize probes:
Who may stop harmful action:
Who adjudicates disputes:
Who may revise the shared model:
Who may revise the ideal:
Who may commit collective resources:
Who may exit or refuse participation:

Decision rules:
Appeal rules:
Record rules:
Retention rules:
Review date:

The constitution itself remains revisable through a higher-threshold procedure.

13. Collective data model

A platform-neutral participant record:

participant_id: A-17
agent_type: human_team
constituency:
  - regional-operations
standing:
  observation: true
  hypothesis: true
  challenge: true
  ideal_review: true
  stop_authority:
    scope: regional_safety
expertise:
  - incident-response
  - regional-regulation
affectedness:
  direct:
    - operational-load
authority:
  authorize_actions:
    - regional-failover
  commit_resources:
    - on_call_staff
capabilities:
  - execute_local_probe
  - provide_incident_logs
commitment_profile:
  ideal_endorsement: qualified
  working_model_acceptance: contested
  resource_commitment: conditional
  execution_commitment: true
  review_commitment: true
conditions:
  - central-coordinator-available
  - safety-floor-preserved

A machine-readable collective message:

{
  "message_id": "M-441",
  "actor_id": "A-17",
  "constituency": "regional-operations",
  "role": "affected-party-representative",
  "speech_act": "challenge",
  "target_claim": "C-28",
  "content": "The proposed latency metric excludes manual recovery time.",
  "evidence_refs": ["E-101", "E-104"],
  "authority_scope": "advisory",
  "requested_disposition": "revise_measurement_claim",
  "timestamp": "2026-07-13T12:00:00Z",
  "signature": "..."
}

Recognized speech acts include:

observe
classify
propose
support
qualify
challenge
object
abstain
authorize
commit
decline
execute
report
appeal
withdraw
revise

Part II — Participation and Governance Engineering

14. What collective governance must do

The protocol’s conceptual rules are insufficient unless the collective can materially realize them.

Governance must make it possible to:

identify participants and constituencies;
preserve differentiated positions;
allocate standing;
expose authority;
register commitments;
control risk;
preserve raw traces;
separate action from adjudication where practical;
make decisions executable;
protect dissent;
enable appeal;
retain revision;
permit legitimate exit.

Governance turns collective reasoning into accountable conduct.

15. Governance components

15.1 Participant and constituency registry

The collective must know:

who is participating;
whom they represent;
whether they speak for themselves or a group;
what authority they possess;
what consequences they bear;
what evidence they control.

A coordinated group may participate as one agent only when:

its internal authorization is defined;
its representative scope is stated;
internal dissent relevant to the collective is preserved or summarized;
its commitments can be enforced internally.

15.2 Standing matrix

Standing is issue-specific.

An agent may have standing to:

observe;
submit evidence;
challenge a classification;
introduce a rival;
object on protected-floor grounds;
vote;
authorize;
execute;
stop;
appeal;
revise the ideal.

Standing should depend on more than expertise.

Relevant bases include:

affectedness;
legal authority;
responsibility;
epistemic access;
implementation capability;
protected-interest representation.

15.3 Commitment profile

Commitment is represented as a vector, not a single score.

Dimension	Question
Ideal endorsement	Does the agent endorse the stated direction?
Working-model acceptance	Does the agent accept the current causal model?
Procedural acceptance	Will the agent treat the adopted model as a basis for coordination?
Resource commitment	What resources will the agent contribute?
Execution commitment	Which actions will the agent perform?
Risk acceptance	Which consequences is the agent authorized and willing to bear?
Review commitment	Will the agent return for evidence review and revision?
Retention commitment	Will the agent help preserve the resulting change?

Possible values include:

endorsed
provisionally accepted
qualified
conditional
contested
abstained
declined
withdrawn

A collective plan is executable only if the necessary commitment chain is complete.

15.4 Authority and capability matrix

For each proposed action, record:

Function	Agent
Propose
Authorize
Fund
Execute
Observe
Stop
Adjudicate
Revise
Retain

A collective may agree on a plan while lacking one required capability.

That is a Prerequisite Tree obstacle, not a failure of motivation.

15.5 Shared memory

The protocol requires durable records of:

prior versions;
assumptions;
dissent;
commitments;
evidence;
decisions;
reversals;
unresolved questions.

Shared memory should be:

versioned;
attributable;
accessible according to standing;
resistant to silent rewriting;
able to preserve minority interpretations.

A collective that remembers only its final decisions cannot examine how those decisions were produced.

15.6 Evidence custody

Raw traces should be controlled by roles that are not entirely dependent on the success of the intervention.

Evidence custody includes:

collection rights;
retention duties;
privacy constraints;
provenance;
access;
redaction;
cross-scale translation.

The same effective coalition should not silently control:

the action;
the measurement;
the classification;
the official story;
and the decision to revise.

15.7 Agenda governance

Agenda control determines what may become a collective sign.

The protocol must record:

who may introduce an issue;
who may defer it;
what threshold requires consideration;
how affected-party concerns enter;
whether unresolved dissent remains visible.

Agenda exclusion is a form of representational closure.

15.8 Decision procedures

Possible decision rules include:

consent;
consensus;
qualified majority;
simple majority;
delegated authority;
expert adjudication;
bicameral approval;
affected-party veto;
threshold authorization;
federated ratification.

No decision rule is universally appropriate.

The rule must match:

risk;
reversibility;
rights implications;
distribution of consequences;
authority;
urgency.

Empirical claims should not be settled by vote alone.

Normative authorization cannot be derived from evidence alone.

The protocol requires both evidential and legitimate decision pathways.

15.9 Independent challenge

Higher-risk collectives should create an explicit challenge function.

The challenger may:

search for alternative causes;
test negative branches;
audit classification;
inspect excluded evidence;
represent absent stakeholders;
challenge the ideal;
test implementation feasibility.

The challenger is not merely an opponent.

It is a structural mechanism of collective losability.

15.10 Appeal and exit

Participants must know:

what may be appealed;
to whom;
within what time;
with what evidence;
whether action pauses;
whether exit is permitted;
what obligations survive exit.

Exit must not be confused with defeat of the collective claim.

But impossible exit may make apparent commitment unreliable.

15.11 Safeguards and stop authority

Every high-risk action should identify:

protected floors;
monitoring conditions;
stop thresholds;
rollback;
authorized stoppers;
consequences of false stop and failed stop.

Stop authority should be distributed to agents capable of observing relevant harms.

15.12 Resource and enforcement path

A collective decision without:

funding;
access;
labor;
authority;
enforcement;
time;
technical capability

is not yet a collective intervention.

The Prerequisite Tree should expose every missing link.

16. Governance enforcement rules

Rule G1 — Standing gate

A pivotal issue may not be closed before all identified affected constituencies have:

participated;
delegated representation;
declined;
or been explicitly recorded as absent.

Rule G2 — Claim-status gate

No contested pivotal claim may be presented as collective fact.

Rule G3 — Commitment gate

No action enters execution until required agents have made explicit commitments.

Rule G4 — Authority gate

No commitment may bind an agent beyond the authority of the committing representative.

Rule G5 — Test-before-settlement gate

No pivotal hypothesis may be promoted without a linked Test Contract and evidence record.

Rule G6 — Dissent-disposition gate

Completion requires disposition of every substantive recorded objection.

Rule G7 — Repetition gate

After two materially similar failed collective interventions, another attempt requires model revision or explicit higher-level authorization.

Rule G8 — Protected-floor gate

Any authorized observer may trigger review when a protected floor is plausibly violated.

Rule G9 — Dependency gate

A revised pivotal claim triggers review of dependent decisions, commitments, and actions.

Rule G10 — Retention gate

The process may not close until ownership of follow-up, monitoring, and future revision is assigned.

17. Governance modes

Lightweight mode

Suitable for a small, low-risk collective.

Provides:

shared charter;
named participants;
claim list;
action assignments;
review meeting.

Standard mode

Suitable for plural interpretation and distributed authority.

Adds:

Goal Tree;
Cloud;
commitment profile;
Test Contract;
evidence ledger;
explicit verdict;
revision log.

Adversarial mode

Suitable for high-risk or high-power asymmetry.

Adds:

independent evidence custody;
protected dissent;
affected-party representation;
challenger role;
appeal;
stop authority;
external audit.

Federated mode

Suitable for multiple collectives.

Adds:

cross-scale gates;
local autonomy rules;
translation protocols;
dual registration;
federated ratification;
provenance preservation.

Part III — Loop Engineering

18. The Collective Thinking Process Loop

The core loop is:

CHARTER
    ↓
REGISTER standing and commitments
    ↓
OBSERVE plural realities
    ↓
MODEL current reality
    ↓
SURFACE conflicts
    ↓
SELECT pivotal claims
    ↓
CONTRACT collective tests
    ↓
ASSEMBLE authority and commitment
    ↓
ACT
    ↓
ACQUIRE traces
    ↓
PERFORM plural uptake
    ↓
DELIBERATE and adjudicate
    ↓
REVISE
    ↓
PROPAGATE
    ↓
RETAIN
    ↓
REVIEW the ideal and constitution

This should be implemented as a state machine.

19. Loop states

19.1 `CHARTERED`

The collective has:

a provisional ideal;
scope;
protected floors;
decision rules;
review procedures.

Transition to STANDING_REGISTERED.

19.2 `STANDING_REGISTERED`

Participants, constituencies, authorities, capabilities, and affected parties are recorded.

Transition to OBSERVING.

19.3 `OBSERVING`

Participants submit:

observations;
testimony;
metrics;
local models;
uncertainties;
consequences not represented in current records.

No unified causal model is yet assumed.

Transition to MODELING.

19.4 `MODELING`

The collective constructs:

a Goal Tree;
Current Reality Tree;
claim portfolio;
alternative causes;
scale and gate map.

Differences remain attributable.

Transition to CONFLICT_SURFACING.

19.5 `CONFLICT_SURFACING`

Clouds are used where:

legitimate needs appear to imply incompatible actions;
commitment is blocked by different assumptions;
agents disagree about necessary conditions;
authority and autonomy conflict.

Transition to PIVOTAL_CLAIM_SELECTED.

19.6 `PIVOTAL_CLAIM_SELECTED`

The collective identifies the claims whose truth or falsity would most alter:

the diagnosis;
the conflict;
the intervention;
the commitment structure;
the ideal.

Transition to TEST_CONTRACTED.

19.7 `TEST_CONTRACTED`

The collective pre-specifies:

incumbent claim;
rivals;
probe;
consequences;
uptake;
adjudication;
revision consequences;
protected floors.

Transition to COMMITMENT_ASSEMBLY.

19.8 `COMMITMENT_ASSEMBLY`

Required agents state:

authorization;
resources;
execution;
monitoring;
risk acceptance;
stop responsibilities.

If a required commitment is absent, transition to PREREQUISITE_ANALYSIS.

If complete, transition to ACTING.

19.9 `PREREQUISITE_ANALYSIS`

The collective identifies obstacles such as:

absent authority;
insufficient trust;
missing capability;
legal prohibition;
unavailable resources;
unacceptable risk;
unresolved legitimacy.

Intermediate objectives are assigned.

Transition back to COMMITMENT_ASSEMBLY.

19.10 `ACTING`

Agents perform coordinated actions under the Transition Tree.

Transition to TRACES_ACQUIRED.

19.11 `TRACES_ACQUIRED`

Raw consequences are preserved across relevant scales.

Transition to UPTAKE_REVIEW.

19.12 `UPTAKE_REVIEW`

The collective examines:

what was selected;
what was omitted;
how observations were classified;
what attributions were made;
what entered durable memory;
whose interpretation was lost in translation.

Transition to DELIBERATION.

19.13 `DELIBERATION`

Participants compare:

predictions;
rivals;
evidence;
protected floors;
affected-party testimony;
commitment consequences.

Transition to ADJUDICATION.

19.14 `ADJUDICATION`

Permitted verdicts include:

Verdict	Meaning
Supported	Evidence discriminates in favor of the collective claim
Weakly supported	Evidence is compatible but not strongly discriminating
Qualified	Claim survives only within narrower conditions
Weakened	Warrant declines without clear replacement
Defeated	Claim should no longer govern collective action
Rival favored	Another explanation receives stronger warrant
Measurement invalid	Indicator cannot support the intended conclusion
Implementation failure	The agreed intervention was not realized
Uptake compromised	Collective evidence processing is unreliable
Commitment failure	Required participation or resources did not materialize
Authority failure	Action or revision lacked valid authority
Inconclusive	No adequate discrimination was achieved
Second-order surprise	No live model explains the evidence
Normative conflict	Effective action violates or destabilizes the ideal
Legitimacy failure	Decision procedure lacked adequate standing or consent

19.15 `RE_ABDUCTION`

Triggered when no live explanation adequately accounts for the registered evidence.

The collective must widen:

vocabulary;
stakeholder participation;
causal scale;
time horizon;
rival set;
institutional analysis.

The two-crossings theory treats registered surprise as the hinge at which failed inquiry should reopen the space of hypotheses.

19.16 `REVISING`

The collective may revise:

a claim;
a tree;
a measurement;
an intervention;
a commitment;
authority allocation;
uptake rules;
protected floors;
the ideal;
the protocol constitution.

Transition to DEPENDENCY_PROPAGATION.

19.17 `DEPENDENCY_PROPAGATION`

Affected:

claims;
decisions;
commitments;
responsibilities;
contracts;
policies;
delegated authorities;
monitoring systems

are reviewed.

Transition to RETENTION.

19.18 `RETENTION`

The collective embeds the revision through:

rules;
routines;
contracts;
training;
software;
budgets;
records;
incentives;
future review dates;
onboarding;
institutional memory.

Transition to IDEAL_REVIEW.

19.19 `IDEAL_REVIEW`

The collective asks:

Did the action move toward the intended ideal?
Were some participants improved at others’ expense?
Did protected floors hold?
Did affected parties expose a missing condition?
Does the Goal Tree still represent the ideal?
Should the constitution change?

Transition to COMPLETE, MODELING, or DISSOLUTION_REVIEW.

20. Nested collective loops

20.1 Interaction loop

message → interpretation → reply → clarification

Time scale: seconds to days.

Purpose: mutual intelligibility.

Risk: mistaking conversational agreement for commitment.

20.2 Inquiry loop

claim → rival → probe → evidence → verdict

Time scale: hours to months.

Purpose: improve the shared model.

20.3 Coordination loop

decision → commitment → assignment → execution → monitoring

Time scale: days to years.

Purpose: produce coordinated conduct.

20.4 Institutional loop

consequence → public or organizational uptake → policy revision → retained practice

Time scale: months to generations.

Purpose: durable collective learning.

Evidence from one loop is not automatically sufficient for another.

A successful meeting does not establish implementation.

A completed action does not establish ideal progress.

A favorable metric does not establish legitimacy.

21. Loop triggers

21.1 Interpretive surprise

Participants classify the same observation incompatibly.

Response:

clarify vocabulary;
preserve rival classifications;
avoid premature averaging.

21.2 First-order empirical surprise

Evidence conflicts with the active model.

Response:

review uptake;
compare rivals;
revise or qualify.

21.3 Second-order surprise

No live model explains the evidence.

Response:

reopen stakeholder participation;
widen the hypothesis space;
inspect another scale.

21.4 Commitment surprise

An agent who accepted the decision does not execute the expected action.

Response:

distinguish misunderstanding, capability failure, authority failure, incentive conflict, and genuine withdrawal.

21.5 Representation trigger

An affected group or consequence is discovered outside the current model.

Response:

reopen standing;
review the ideal and relevant trees.

21.6 Protected-floor trigger

A plausible threshold violation occurs.

Response:

pause or contain;
perform independent review.

21.7 Capture trigger

One coalition controls the agenda, evidence, classification, and verdict.

Response:

separate roles;
introduce independent custody or review.

21.8 Scope trigger

The action affects agents or systems outside the original charter.

Response:

revise scope;
add standing;
reconsider authority.

21.9 Repetition trigger

The same intervention fails twice without a changed model.

Response:

block repetition;
require re-abduction or escalation.

21.10 Dissolution trigger

The collective no longer possesses sufficient overlapping ideal, authority, commitment, or legitimacy.

Response:

renegotiate charter;
federate;
narrow scope;
or dissolve.

22. Loop termination

The collective loop may terminate as COMPLETE only when:

the agreed scope has been addressed;
pivotal claims have explicit status;
commitments have been performed or formally revised;
affected-party objections have received disposition;
evidence and dissent remain preserved;
dependent decisions have been reviewed;
retention ownership exists;
residual disagreement is recorded;
reopening conditions are specified;
no unresolved protected-floor violation remains.

It may terminate as PAUSED when:

decisive evidence is unavailable;
a protected-floor concern requires review;
required authority is absent;
commitments are incomplete;
a legitimate appeal is active.

It may terminate as DISSOLVED when:

the shared goal no longer exists;
participating agents withdraw essential authority;
the process becomes illegitimate;
coordination costs exceed the value of joint action;
federation or separate action is more appropriate.

Dissolution is not necessarily failure.

It may be the warranted revision of an invalid collective assumption.

23. Preventing collective loop pathologies

23.1 False-consensus loop

Shared words
→ assumed shared meaning
→ action
→ conflict
→ conflict attributed to weak commitment

Repair:

operationally define pivotal terms;
record alternative interpretations;
distinguish semantic disagreement from commitment failure.

23.2 Majority-confirmation loop

Majority adopts explanation
→ majority controls action
→ action generates predicted consequence
→ consequence confirms majority

Repair:

include intervention effects in the causal model;
preserve minority predictions;
use counterfactual or alternative probes.

23.3 Coalition capture loop

Coalition controls agenda
→ controls available evidence
→ controls classification
→ controls verdict

Repair:

distribute agenda rights;
separate evidence custody;
create appeal and external review.

23.4 Commitment theater loop

Agents publicly agree
→ no resources or authority are committed
→ action fails
→ model is blamed

Repair:

use commitment profiles;
require attributable resource and execution commitments.

23.5 Dissent-erasure loop

Objection raised
→ summarized by incumbent coalition
→ nuance removed
→ final record shows consensus

Repair:

allow self-authored dissent;
preserve original challenge;
require disposition.

23.6 Lowest-common-denominator loop

Collective seeks universal agreement
→ removes every contested claim
→ produces vague plan
→ plan cannot guide action

Repair:

permit conditional and federated commitments;
preserve disagreement while selecting bounded action.

23.7 Endless-deliberation loop

New perspective
→ further discussion
→ no decision threshold
→ no action-generated evidence

Repair:

set inquiry budgets;
distinguish reversible probes from irreversible decisions;
use bounded action to acquire evidence.

23.8 Voting-as-truth loop

Claim disputed
→ vote held
→ majority claim treated as fact

Repair:

separate epistemic adjudication from normative authorization;
preserve minority hypotheses.

23.9 Local-success/global-harm loop

One unit improves
→ cost moves to another unit or affected population
→ local metric confirms success

Repair:

map cross-scale gates;
include displaced consequences;
grant affected parties standing.

23.10 Institutional amnesia loop

Collective revises
→ participants rotate
→ records lose rationale
→ old policy returns

Repair:

preserve revision deltas;
update training, rules, systems, and onboarding.

23.11 Authority-without-uptake loop

Leader can act
→ consequences are filtered
→ leader sees only confirmation

Repair:

independent evidence channels;
protected reporting;
direct affected-party access.

23.12 Uptake-without-authority loop

Collective learns
→ no actor can implement revision
→ conduct remains unchanged

Repair:

assign executable revision authority;
revise governance or scope.

Part IV — End-to-End Operating Procedure

24. Step 1: Name the collective ideal

The collective should state the ideal as an orientation, not merely a slogan.

The ideal-directed theory represents an ideal through:

a viable region;
a partial relation of improvement;
protected floors;
a procedure for revising the end.

Example:

The collective seeks a service that remains reliably available during regional failures while preserving local safety, cross-region consistency, transparent incident reporting, and the autonomy necessary for timely response.

Protected floors:

no suppression of incidents;
no unauthorized exposure of personal data;
no forced transfer of legal responsibility;
no local action that creates unacceptable systemic risk.

The charter records where participants differ.

25. Step 2: Audit current reality

Each participant submits observations from its position.

O-1: Central approval delays emergency response.
O-2: Local interventions sometimes conflict across regions.
O-3: Regional teams do not trust central status information.
O-4: Central coordinators receive incomplete local incident reports.
O-5: Incident metrics exclude manual recovery work.
O-6: Local teams are held accountable for outcomes they cannot authorize.

The Current Reality Tree connects the effects without erasing provenance.

A possible model:

Authority and information are separated
        ↓
Local teams cannot act or fully inform central coordination
        ↓
Central decisions arrive late and with incomplete context
        ↓
Local teams create informal workarounds
        ↓
Cross-region consistency declines
        ↓
Central control is tightened
        ↓
Authority and information become further separated

This is a reinforcing conduct loop.

26. Step 3: Surface the conflict

An Evaporating Cloud:

A: Maintain reliable and legitimate system response

B: Enable timely action under local conditions
C: Preserve coordinated system-wide behavior

D: Grant local teams broad discretionary authority
D′: Require central authorization before action

Possible assumptions:

Local discretion necessarily creates unacceptable inconsistency.
Central authorization necessarily prevents inconsistency.
Local teams cannot be trusted with system-wide information.
Central coordinators can act quickly enough.
Responsibility must be concentrated where authority is concentrated.
One uniform policy is necessary for coordination.

Different agents may challenge different arrows.

The Cloud converts personal or institutional opposition into explicit assumptions.

27. Step 4: Identify the pivotal claim

A pivotal claim may be:

C-41: A bounded local-authority protocol can preserve rapid response without producing unacceptable cross-region inconsistency.

Rivals:

R1: Any meaningful local authority produces unacceptable divergence.
R2: The delay is caused by inadequate tooling rather than authority structure.
R3: Local authority works only for a narrow class of incidents.
R4: Central coordination can be made fast enough without delegation.
R5: The apparent conflict is primarily caused by ambiguous accountability.

The selected claim should materially affect the collective’s action.

28. Step 5: Create the Collective Test Contract

COLLECTIVE TEST CONTRACT

Target claim:
    Bounded local authority preserves rapid response without unacceptable
    system-wide inconsistency.

Participating agents:
    Central coordination
    Two regional teams
    Reliability engineering
    Compliance
    Affected customer representative

Rivals:
    R1 through R5

Probe:
    Run a time-bounded pilot in which regional teams may take pre-specified
    emergency actions under shared logging, notification, and rollback rules.

Expected result under incumbent:
    Response begins faster.
    Conflicting actions remain below the defined threshold.
    Central visibility improves.
    Local reporting remains complete.

Defeating result:
    Material inconsistency rises beyond the protected threshold.
    Local teams conceal or fail to register relevant actions.
    Central coordination becomes less capable of recovery.
    Affected-party harm exceeds the agreed floor.

Raw traces:
    Action timestamps
    Authorization events
    Local and central logs
    Rollback events
    Conflicting commands
    Incident duration
    Manual work
    Affected-party reports

Selection rule:
    Include all pilot incidents, including aborted and near-miss cases.

Classification rule:
    Predefine inconsistency, harmful divergence, acceptable local variation,
    delayed response, and incomplete reporting.

Attribution rule:
    Separate effects of authority, tooling, training, incident type, and staffing.

Adjudication:
    Joint panel with independent evidence custodian and affected-party standing.

Required revision:
    Supported: expand cautiously.
    Qualified: narrow incident scope or strengthen safeguards.
    Defeated: withdraw protocol and revisit authority model.
    Normative conflict: suspend and review the ideal.

29. Step 6: Assemble an admissible probe

A collective probe must satisfy:

relevant agents can actually perform it;
materially different outcomes may occur;
commitments are explicit;
risk is contained;
affected parties are represented;
evidence is preserved;
rollback exists;
the result may change future action.

Required commitments:

Agent	Commitment
Central coordination	Recognize bounded local authority during pilot
Regional teams	Execute only authorized action classes and preserve logs
Reliability engineering	Instrument and monitor
Compliance	Review protected-floor conditions
Evidence custodian	Preserve raw traces
Adjudication panel	Return for verdict
Executive authority	Implement or reverse the resulting governance change

30. Step 7: Act and acquire traces

During the pilot, the collective records:

actions taken;
actions considered but not taken;
local context;
central context;
delays;
conflicts;
near misses;
reporting behavior;
unanticipated effects;
deviations from the protocol;
participant withdrawals;
changes in commitment.

Absence of action may itself be a consequence.

31. Step 8: Perform plural uptake

Selection

Did each scale observe the same events?

Were near misses included?

Were non-participating affected parties sampled?

Classification

Was a local deviation classified as:

adaptive discretion;
non-compliance;
necessary exception;
unsafe divergence?

Who controlled the category?

Attribution

Was a good result attributed to local authority when it was actually caused by extra staffing?

Was failure attributed to local teams when central data were incomplete?

Registration

Did local and central records preserve:

provenance;
timestamps;
disagreements;
missing data;
later corrections?

A consequence becomes evidence only through these uptake operations.

32. Step 9: Deliberate and adjudicate

Suppose the pilot shows:

median response begins 40% faster;
no severe cross-region conflicts occur;
three minor divergences occur;
all are corrected through shared logging;
one regional team omits manual workload from its report;
affected customers report improved communication;
central coordinators report greater visibility but increased review burden.

Verdict:

The claim is supported but qualified.

Qualification:

Bounded local authority works for pre-classified incidents when shared state, notification, and rollback are available. It creates additional central review work and requires stronger manual-work registration.

This is not recorded as unanimous belief.

The Decision Log records:

who supports the verdict;
who qualifies it;
who objects;
what conditions govern expansion;
which agents commit to the next step.

33. Step 10: Revise the collective model

Old branch:

Local authority
    ↓
Inconsistent action
    ↓
System instability

Revised branch:

Local authority
AND shared state
AND bounded action classes
AND immediate notification
AND rollback
    ↓
Faster local response
AND recoverable variation

New negative branch:

Expanded local authority
    ↓
More central review work
    ↓
Possible review backlog

New injection:

Automate comparison of local actions with shared incident state while preserving human override.

34. Step 11: Implement through a Collective Transition Tree

Current condition:
    Pilot demonstrated qualified viability.

Need:
    Preserve faster response while preventing review overload.

Action:
    Define three incident classes eligible for bounded local authority.

Responsible:
    Central coordination and regional operations.

Expected effect:
    Eligible actions become predictable and auditable.

Evidence:
    Classification agreement rate and exception frequency.

Next action:
    Implement shared action ledger and automated conflict warnings.

Responsible:
    Platform engineering.

Expected effect:
    Cross-region divergence is detected before harmful execution.

Evidence:
    Warning precision, response latency, prevented conflicts.

Next action:
    Expand to additional regions only after local training and evidence-custody
    readiness are confirmed.

Expected effect:
    Qualified improvement scales without losing uptake integrity.

Each transition step includes:

responsible agent;
authorizing agent;
committed resources;
expected effect;
evidence;
rollback;
next decision point.

35. Step 12: Propagate and retain

Review dependent:

policies;
contracts;
escalation procedures;
accountability rules;
technical systems;
training;
metrics;
budgets;
incident reviews;
public reporting.

Retention mechanisms:

amended authority charter;
shared action ledger;
regression simulations;
recurring review;
onboarding;
protected dissent channel;
annual Ideal Review.

A diagrammatic revision without institutional carryover is not collective learning.

Part V — Collective Arrangements

36. Small-group arrangement

Suitable for:

small teams;
partnerships;
local human–AI groups;
project squads.

Roles may be combined, but the protocol still distinguishes:

facilitator;
model steward;
evidence custodian;
action owner;
challenger;
adjudicator.

A small group should use:

one shared charter;
one Goal Tree;
one commitment ledger;
explicit dissent;
bounded probes.

37. Distributed-agent arrangement

Suitable for:

autonomous AI agents;
cross-functional organizations;
geographically distributed teams.

Each agent communicates through structured messages.

The shared protocol service maintains:

identity;
claim status;
tree versions;
commitment state;
evidence references;
deadlines;
authority constraints.

Agents may operate asynchronously.

No agent may infer silence as consent unless the constitution explicitly defines that rule.

38. Federated arrangement

Suitable for:

alliances;
multi-organization programs;
public–private systems;
groups of groups.

Each local collective maintains:

its own charter;
local Goal Tree;
local commitments;
local evidence.

The federation maintains:

shared ideal intersection;
cross-scale claims;
translation gates;
federated decisions;
provenance;
local reservation rights.

The protocol must preserve the difference between:

local acceptance;
federated acceptance;
binding federation-wide obligation.

Cross-scale gates should record what was lost or transformed during translation.

39. Mixed human–AI arrangement

Human and AI agents may occupy any role allowed by authority and capability.

AI agents may:

construct trees;
maintain claim registers;
identify assumptions;
search for rivals;
monitor commitments;
detect dependency impacts;
summarize evidence;
simulate FRT branches;
challenge inconsistencies.

Human agents should retain or explicitly delegate authority over:

protected-floor definition;
coercive action;
rights-affecting decisions;
binding resource commitments;
legitimacy judgments;
revisions of foundational ideals.

The protocol does not assume that all humans share one kind of judgment or that all AIs lack standing.

Standing is assigned by the constitution and constrained by applicable authority and ethics.

Part VI — Relationship to Existing Collective Decision Methods

40. Consensus inside the protocol

Consensus may be used for:

defining a common ideal;
adopting protected floors;
high-trust reversible decisions;
constitutional rules.

Consensus is not required for every claim.

The protocol permits:

qualified agreement;
conditional commitment;
recorded dissent;
federated action;
bounded experimentation.

Consensus should not be produced by removing every meaningful difference.

41. Voting inside the protocol

Voting may authorize action.

It does not by itself establish causal truth.

A conforming vote record should identify:

decision_id: D-19
decision_type: action_authorization
rule: qualified_majority
eligible_agents:
  - A-1
  - A-2
  - A-3
result:
  authorize: 2
  reject: 1
epistemic_status_of_claim: contested
minority_position: preserved
protected_floor_review: passed
reopening_condition:
  - evidence_threshold_E27

Voting and hypothesis status remain separate.

42. Deliberation inside the protocol

Deliberation is used to:

clarify entities;
expose assumptions;
generate rivals;
compare needs;
interpret evidence;
revise the ideal.

Deliberation should culminate in:

a sharper claim;
a test;
a commitment;
a recorded unresolved issue;
or a reasoned stop.

Discussion without a state transition should trigger process review.

43. Prediction markets and scoring methods

Prediction markets, confidence scores, or forecasts may help compare expectations.

They do not replace:

affected-party standing;
protected floors;
causal inference;
authority;
commitment;
revision.

A collective may score predictions while still preserving qualitative dissent and normative constraints.

44. Multi-agent debate

Structured debate may implement the challenger function.

A debate is useful when:

live rivals exist;
roles are balanced;
raw evidence is shared;
adjudication rules are fixed;
the winning argument can revise action.

Debate becomes theater when the outcome does not alter authority, commitments, or the shared model.

Part VII — Relationship to Governance Engineering

45. Governance engineering as externalized collective integrity

The central role of governance is not administrative efficiency.

It is to prevent one effective code or coalition from controlling:

what may be discussed;
what may be observed;
what the observation means;
what becomes the official record;
whether the result changes conduct.

The formal theory distinguishes epistemic integrity from ethical legitimacy and treats role separation, provenance, pre-registration, appeal, and independent review as mechanisms that support integrity.

Governance engineering externalizes:

memory;
standing;
authority;
commitment;
evidence custody;
adjudication;
stopping;
revision.

This gives the collective a practical capacity to receive an unwelcome answer.

46. Governance controls by closure type

Closure risk	Governance response
Relevant stakeholder absent	Standing review and representation
Relevant harm has no category	Vocabulary expansion and affected-party testimony
One coalition controls agenda	Distributed agenda rights
Action manufactures confirmation	Alternative response or counterfactual probe
Disconfirming state never occurs	Bounded off-path experiment
Different outcomes get same verdict	Predefined plural classification
Evidence is recorded but ignored	Bidirectional criticism and appeal
Verdict lacks power	Assign revision authority and resources
Commitment is vague	Commitment ledger
Representative exceeds mandate	Authority-scope validation
Dissent disappears	Append-only dissent log
Local learning does not compose	Cross-scale provenance and gate audit
Decision improves harmful end	Ideal and legitimacy review
Revision does not persist	Retention ownership and institutional memory

Governance engineering is the material implementation of collective losability.

Part VIII — Relationship to Loop Engineering

47. Loop engineering as the design of collective correction

A collective may have:

intelligent participants;
good data;
open discussion;
formal authority

and still fail because its loop is poorly engineered.

Common loop defects include:

action before claim clarification;
evidence collection after criteria are known to be favorable;
repeated deliberation without probes;
repeated action without re-abduction;
unresolved dissent disappearing between meetings;
commitments decaying between decision and execution;
local success ending the loop before ideal review.

Loop engineering determines:

sequence;
state;
recurrence;
escalation;
memory;
stopping;
reopening.

48. The critical collective loop distinction

There are two distinct collective loops.

Steering loop

What coordinated action moves the collective toward the ideal?

Inquiry loop

What coordinated action makes competing explanations or policies distinguishable?

They often overlap.

But a collective may need to perform a probe that does not immediately optimize the focal outcome.

Examples:

pilot one governance rule;
preserve parallel reporting systems;
temporarily delegate authority;
run two local policies;
invite an external classification;
expose an off-path state.

The ideal-navigation theory explicitly distinguishes steering value from inquiry value.

A collective action may be evaluated as:

[ \text{Collective action value}

\text{expected ideal progress} + \beta \cdot \text{expected discrimination}

\text{cost}

\text{risk}

\text{legitimacy burden} ]

The expression is a decision aid, not a universal utility function.

49. Re-abduction as a collective transition

Ordinary collective processes often implement:

failure
    ↓
renewed persuasion
    ↓
stronger implementation of the same policy

The CTPP requires:

failure of incumbent expectation
        ↓
review measurement and uptake
        ↓
compare live rivals
        ↓
if none explains result:
    widen representation and reopen abduction

Collective re-abduction may require:

new stakeholders;
another discipline;
another scale;
another vocabulary;
historical evidence;
different causal boundaries;
questioning the shared ideal.

50. Collective loop memory

The collective should maintain external state:

current ideal
protected floors
participants and constituencies
standing
claim portfolio
rivals
tree versions
commitment profiles
authority matrix
test contracts
raw traces
uptake records
verdicts
dissent
failed interventions
unresolved surprises
dependency impacts
retention duties
review dates

Collective memory should not depend on:

one facilitator;
one leader;
one AI context window;
one meeting transcript;
one institutional narrative.

Part IX — Evaluation and Meta-Losability

51. The protocol itself must be tested

The relevant contention is:

A collective using the Collective Thinking Process Protocol will produce more corrigible shared models, more reliable commitments, more legitimate action, and more durable improvement than credible alternative collective processes at acceptable cost.

This contention remains in probation.

52. Comparative evaluation

Evaluate at least four conditions:

Condition	Description
A	Ordinary meeting-and-decision process
B	Ordinary process with improved facilitation and records
C	Uptake Integrity Protocol without full collective governance
D	Full Collective Thinking Process Protocol

Optional comparison:

| E | Credible rival, such as consensus governance, deliberative polling, structured decision analysis, or another multi-agent coordination framework |

Cases should vary by:

number of participants;
human, AI, or mixed composition;
degree of goal alignment;
power asymmetry;
uncertainty;
reversibility;
cross-scale impact;
conflict severity;
commitment dispersion.

53. Evaluation measures

Outcome measures

goal progress;
reliability of coordinated execution;
negative-branch incidence;
durability;
recovery from failure;
affected-party outcomes.

Inquiry measures

accuracy of causal models;
number of live rivals;
quality of predicted effects;
rate of appropriate claim revision;
detection of measurement failure;
frequency of re-abduction;
calibration of collective confidence.

Participation measures

affected-party inclusion;
preservation of dissent;
semantic clarification;
rate of substantive objection disposition;
ability to distinguish disagreement from non-commitment.

Commitment measures

percentage of required commitments made explicitly;
commitment fulfillment;
authority validity;
resource sufficiency;
unexplained withdrawal;
retention compliance.

Integrity measures

raw trace preservation;
role concentration;
excluded evidence;
uptake disagreement;
appeal outcomes;
revision authority;
cross-scale provenance.

Efficiency measures

elapsed time;
communication burden;
administrative cost;
delay to bounded action;
rework;
number of unnecessary participants;
decision latency.

54. Defeat conditions

The protocol should be narrowed or rejected if comparative evaluation shows that:

it does not improve collective model correction;
it produces more documentation without better decisions;
simpler commitment and evidence records provide equivalent benefit;
participants cannot reliably use the claim and commitment distinctions;
the protocol systematically privileges formally articulate agents;
affected-party standing becomes symbolic;
AI agents exploit the forms without genuine exposure to defeat;
human participants experience excessive procedural burden;
dissent is recorded but has no causal influence;
collective action becomes slower without compensating risk reduction;
the protocol increases strategic gaming;
federated implementations lose coherence;
the full method performs worse than a credible rival;
benefits require unusually expert facilitation;
the protocol strengthens harmful collective goals more effectively than it revises them.

The protocol is not entitled to define every failure as poor implementation.

Repeated implementation failure may count against the method’s practical adequacy.

Part X — Implementation Roadmap

55. Phase 1: Minimal collective protocol

Implement:

Collective Ideal Charter;
Participant Register;
one Goal Tree;
one pivotal claim;
one commitment ledger;
one review point.

Use with small, low-risk groups.

56. Phase 2: Conflict and evidence support

Add:

Current Reality Tree;
Evaporating Cloud;
rival hypotheses;
Collective Test Contract;
evidence ledger;
permitted verdicts.

57. Phase 3: Authority and dependency

Add:

authority-capability matrix;
Prerequisite Tree;
dependency graph;
revision propagation;
Transition Tree;
retention ownership.

58. Phase 4: Adversarial and affected-party integrity

Add:

independent challenger;
evidence custody;
protected dissent;
affected-party standing;
appeal;
stop authority;
legitimacy review.

59. Phase 5: Federated and cross-scale coordination

Add:

local charters;
cross-scale gates;
federated claim status;
dual registration;
provenance;
translation audits;
local reservation rights.

60. Phase 6: Meta-evaluation

Compare the protocol against simpler and rival approaches.

Remove or simplify elements that do not contribute.

A losable protocol should be capable of becoming smaller.

Part XI — Compact Operating Standard

61. The twelve obligations

A conforming Collective Thinking Process must:

state the shared or overlapping ideal;
identify protected floors;
identify participants, constituencies, and affected parties;
distinguish personal belief, procedural acceptance, and commitment;
make pivotal claims and assumptions visible;
preserve live rivals and substantive dissent;
state possible defeating consequences;
preserve raw traces separately from collective interpretation;
allocate valid authority, capability, and resources;
revise dependent claims and commitments after adverse evidence;
retain warranted revision;
preserve reopening, appeal, and ideal-review conditions.

62. The three-layer standard

Participation

Must provide:

identity;
standing;
constituency;
affectedness;
commitment profile;
challenge and exit rights.

Governance

Must provide:

records;
authority;
evidence custody;
decision rules;
safeguards;
appeal;
revision power.

Loop

Must provide:

explicit states;
conflict handling;
testing;
plural uptake;
re-abduction;
propagation;
termination;
retention.

63. Final architecture

                       COLLECTIVE IDEAL
                              │
                              ▼
                ┌────────────────────────┐
                │  COLLECTIVE IDEAL      │
                │       CHARTER          │
                │                        │
                │ viable region          │
                │ protected floors       │
                │ standing               │
                │ end revision           │
                └───────────┬────────────┘
                            │ orients
                            ▼
                ┌────────────────────────┐
                │ PARTICIPATING AGENTS   │
                │                        │
                │ humans                 │
                │ AIs                    │
                │ teams                  │
                │ organizations          │
                │ federated groups       │
                └───────────┬────────────┘
                            │ contribute
                            ▼
                ┌────────────────────────┐
                │ COLLECTIVE LTP MODEL   │
                │                        │
                │ Goal Tree              │
                │ Current Reality Tree   │
                │ Clouds                 │
                │ Future Reality Tree    │
                │ Prerequisite Tree      │
                │ Transition Tree        │
                └───────────┬────────────┘
                            │ guides
                            ▼
                ┌────────────────────────┐
                │ COMMITMENT AND         │
                │ GOVERNANCE LAYER       │
                │                        │
                │ authority              │
                │ capability             │
                │ resources              │
                │ safeguards             │
                │ decision rules         │
                └───────────┬────────────┘
                            │ enables
                            ▼
                ┌────────────────────────┐
                │ COORDINATED ACTION     │
                └───────────┬────────────┘
                            │ co-produces
                            ▼
                ┌────────────────────────┐
                │ MATERIAL AND SOCIAL    │
                │ CONSEQUENCES           │
                └───────────┬────────────┘
                            │ traces
                            ▼
                ┌────────────────────────┐
                │ PLURAL EVIDENTIAL      │
                │ UPTAKE                 │
                │                        │
                │ select                 │
                │ classify               │
                │ attribute              │
                │ register               │
                └───────────┬────────────┘
                            │ evidence
                            ▼
                ┌────────────────────────┐
                │ DELIBERATION AND       │
                │ ADJUDICATION           │
                │                        │
                │ rivals                 │
                │ dissent                │
                │ affected parties       │
                │ protected floors       │
                └───────────┬────────────┘
                            │ revises
                            ▼
                ┌────────────────────────┐
                │ MODEL, COMMITMENTS,    │
                │ GOVERNANCE, OR IDEAL   │
                └───────────┬────────────┘
                            │
                            └──── returns to collective loop

Conclusion

The Collective Thinking Process Protocol treats collective action as a system of differentiated agents coordinating through losable models, explicit commitments, plural evidence, and revisable governance.

It does not require participants to:

possess identical beliefs;
understand the ideal in exactly the same way;
bear equal consequences;
contribute equal resources;
hold equal authority;
display the same degree of commitment.

It requires those differences to become visible and governable.

As a participation discipline, the protocol distinguishes constituency, standing, affectedness, acceptance, objection, capability, authority, and commitment.

As a governance discipline, it externalizes memory, evidence custody, safeguards, decision rights, appeal, resource commitment, revision authority, and retention.

As a loop discipline, it organizes collective conduct through observation, modeling, conflict surfacing, probation, coordinated action, plural uptake, adjudication, revision, propagation, and ideal review.

Its central operating principle is:

Every significant collective action rests on a provisional model of reality, a distribution of commitments, and a conception of the ideal. The model, commitments, governance arrangements, and ideal must all remain capable of revision.

A mature collective should therefore not merely ask:

Did we reach agreement?

It should ask:

What exactly did we agree upon? Which differences remain? Who is committed to what? What assumptions make the conflict appear necessary? What consequences could defeat the shared model? Who controls how those consequences become evidence? Who possesses the authority to revise collective conduct? What must change in the trees, commitments, institutions, or ideal when reality resists?

Collective Thinking Process Protocol

Integrating the Uptake Integrity Protocol with collective agency, governance engineering, and loop engineering

1. Purpose

2. Collective thinking as semiotic conduct

3. The central object: a losable collective claim

Observation claim

Classification claim

Causal claim

Necessity claim

Sufficiency claim

Conflict claim

Need claim

Assumption claim

Intervention claim

Risk claim

Measurement claim

Normative claim

4. Mapping LTP artifacts into collective thinking

5. Three engineering disciplines

5.1 Participation engineering

5.2 Governance engineering

5.3 Loop engineering

6. Why the three layers must remain separate

Part I — Building the Protocol

7. The Collective Thinking Process Protocol

8. When the protocol should activate

9. Protocol invariants

9.1 No fictitious collective mind

9.2 No unrepresented affected party where representation is feasible

9.3 No pivotal claim without visible status

9.4 No consensus by ambiguity

9.5 No conflation of acceptance and commitment

9.6 No collective action without attributable authority

9.7 No pivotal test without possible collective loss

9.8 No consequence silently converted into shared evidence

9.9 No majority immunity

9.10 No dissent without disposition

9.11 No repeated collective action without changed information

9.12 No completed decision without retention

10. Protocol inputs and outputs

Inputs

Required outputs

11. Recommended protocol package

12. Collective constitution

13. Collective data model

Part II — Participation and Governance Engineering

14. What collective governance must do

15. Governance components

15.1 Participant and constituency registry

15.2 Standing matrix

15.3 Commitment profile

15.4 Authority and capability matrix

15.5 Shared memory

15.6 Evidence custody

15.7 Agenda governance

15.8 Decision procedures

15.9 Independent challenge

15.10 Appeal and exit

15.11 Safeguards and stop authority

15.12 Resource and enforcement path

16. Governance enforcement rules

Rule G1 — Standing gate

Rule G2 — Claim-status gate

Rule G3 — Commitment gate

Rule G4 — Authority gate

Rule G5 — Test-before-settlement gate

Rule G6 — Dissent-disposition gate

Rule G7 — Repetition gate

Rule G8 — Protected-floor gate

Rule G9 — Dependency gate

Rule G10 — Retention gate

17. Governance modes

Lightweight mode

Standard mode

Adversarial mode

Federated mode

Part III — Loop Engineering

18. The Collective Thinking Process Loop

19. Loop states

19.1 CHARTERED

19.1 `CHARTERED`

19.2 `STANDING_REGISTERED`

19.3 `OBSERVING`

19.4 `MODELING`

19.5 `CONFLICT_SURFACING`

19.6 `PIVOTAL_CLAIM_SELECTED`

19.7 `TEST_CONTRACTED`

19.8 `COMMITMENT_ASSEMBLY`

19.9 `PREREQUISITE_ANALYSIS`

19.10 `ACTING`

19.11 `TRACES_ACQUIRED`

19.12 `UPTAKE_REVIEW`

19.13 `DELIBERATION`

19.14 `ADJUDICATION`

19.15 `RE_ABDUCTION`

19.16 `REVISING`

19.17 `DEPENDENCY_PROPAGATION`

19.18 `RETENTION`

19.19 `IDEAL_REVIEW`