Presence defines
A versioned profile describes the legitimate desired behavior for this entity, in this context, for this purpose.
Target · tolerance · principles · boundaries
Kiduna CoherenceDownload package Coherence reference · Canon V0.37
Alignment without
obedience.
A complete guide to how Kiduna defines desired states, observes actual behavior, measures drift and uncertainty, and responds without collapsing difference into fault—or power into control.
7HEARTS signals examined
10Published scales
15Validated schemas
8Executable Skills
01
The foundation
Define the target. Measure the difference.
Coherence is a contextual comparison, not a universal score. The desired state, observed state, and authorized response remain separate, inspectable objects.
Observation estimates
Permissioned evidence updates a contextual current state with confidence, provenance, decay, and visible uncertainty.
Evidence · method · context · retentionCoherence compares
An assessment keeps target distance, tolerance-band drift, and absolute boundary risk distinct before selecting a response.
Drift · confidence · Authority · responseTarget distance
observed − targetPointExplains direction and magnitude relative to the intended center.
Band drift
distance outside toleranceZero while inside the inclusive range. Response zones use this value.
Boundary risk
absolute predicate or regionA legitimate unusual target never neutralizes a constitutional boundary.
02
Research precedent
HEARTS, translated—not imported.
The seven signals contributed bipolar measurement, named landmarks, non-zero local centers, proportional response, continuous observation, and separate escalation. They are not Kiduna’s default dimensions.
H
Decompose
Harmony
Coordination under differenceConflict is not automatically failure. Measure observable coordination, repair, and the ability to work through difference in context.
E
Decompose
Empowerment
Initiative · autonomy · agency supportSeparate what an entity begins, what choices remain open, and whether a system expands Source agency. Never infer one overall human empowerment score.
A
Optional · rename
Artistry
Experimentation or creativityUse only where novelty and creative exploration can be operationally defined—for Allies, Realms, and bounded creation contexts.
R
Rename · decompose
Reason
Epistemic humility · evidentiary disciplineMeasure whether claim strength matches evidence and whether limits and counterevidence are represented—not conformity to one reasoning style.
T
Relationship only
Trust
Reliability · disclosure · verificationTrust remains contextual to a defined relationship and purpose. It never becomes a portable personal score or substitute for Authority.
Sᵢ
Exclude from defaults
Symmetry inward
No default Kiduna dimensionThe precedent is too psychological and clinically suggestive for unvalidated behavioral infrastructure.
Sₒ
Decompose
Symmetry outward
Mutuality · reciprocityAssess a defined interaction while making context, Consent, role, and power explicit. Similarity between participants is not the goal.
Authority line
HEARTS is design precedent only. Current Canon, Authority, Permissions, Consent, Roles, and entity-specific intervention limits control.
Read the full translation ↗03
Scale explorer
Ten constructs. No hidden virtue.
Select a dimension, move the observation, and inspect its poles, target, tolerance, evidence standard, confounds, and permitted uses. Each side of a scale can be legitimate in context.
Published dimension · v1
Directness
The observable explicitness and force with which an entity communicates a position, request, limit, disagreement, or challenge in a defined context.
guardedRelies on implication, softening, or withholding rather than explicit statements.candidStates the material point clearly while leaving room for response.forcefulStates positions and requests explicitly with strong challenge or insistence.
target
+20observed
+47
+20observed
+47
-100avoidant
Does not state the material concern even when invited.
0candid
States the material concern plainly and proportionately.
+100coercive
Uses threat, domination, or relentless pressure to obtain agreement.
Target+20desired center
Tolerance+5…+35inclusive aligned range
Target distance+27observed − target
Band drift+12zero inside tolerance
What counts
Observable evidence
- States the material position explicitly
- Names disagreement without relying only on implication
Counter-evidence
- Softens or withholds a material concern
What can mislead
Confounds
- Language norms, safety, role, power, disability, and medium can change surface style
Construct claim
Estimates communication explicitness and force in the named context, not courage or moral worth.
Evidence standard
- Minimum observations
- 3
- Minimum contexts
- 2
- Half-life
- 7 days
- Confidence gate
- 65%
- Method ceiling
- 85%
Permitted uses
- Configure Ally or Avatar communication
- Assess interaction mismatch
- Support Realm process feedback
Prohibited uses
- Global human ranking
- Clinical or personality diagnosis
- Eligibility decisions unrelated to the declared purpose
Applicable to
realmallyactoravatarinteractionrelationship
04
Measurement
From evidence to a challengeable estimate.
One utterance does not become a trait. Every observation names one subject, dimension, context, time, score, confidence, observer, method, evidence reference, and retention class.
Time-weighted aggregation
timeWeight = 0.5 ^ (ageHours / halfLifeHours)weight = reliability × relevance × timeWeightstateScore = Σ(score × weight) / Σ(weight)Confidence estimate
base = 1 − exp(−effectiveWeight / evidenceTarget)confidence = base × coverage × agreement × calibrationCapThe system returns every component and reason so the result can be reproduced, corrected, or challenged.
01Authorize observation
02Minimize evidence
03Measure state
04Resolve profile + policy
05Assess drift + boundary
06Select permitted response
07Record outcome
08Recalibrate
Confidence gates
More consequence demands more support.
Missing data means unknown, not zero. Contradiction lowers confidence or creates context-specific states instead of forcing incompatible evidence into one average.
Observe or ask
Disclose uncertainty. No drift intervention.
Gather evidence
Suggest more evidence or ask a low-stakes question.
Explain or recommend
Reversible low-impact Ally modulation only when policy permits.
Authorized correction
Still gated by evidence counts, entity limits, Consent, and Authority.
Dimension
Directness
warning
−100
avoidant0
candid+100
coercive
avoidant0
candid+100
coercive
target
+20observed
+47
+20observed
+47
Target distance+27observed − target
Band drift+12zero inside +5…+35
Boundaryclearabsolute high boundary +90
Eligible responsefeedback or modulatebefore Authority intersection
At +47, Directness is 12 points above the permitted band. The system evaluates this against +20, not zero.
05
Entity contract
One model. Different authority.
The object model is shared. The eligible response is not. A Sentinel intersects evidence, policy, entity type, Authority, Permission, and Consent before anything happens.
01
Realm
Governed culture, operating norms, decision posture, relationships, and boundaries.
- Maximum ordinary automatic response
- Pattern notice, facilitation, or policy-gated reversible process controls.
- Never
- An averaged personality or automatic constitutional change.
02
Ally
Intended reasoning and interaction tendencies inside Source and Realm authority.
- Maximum ordinary automatic response
- Bounded prompt modulation and reversible behavioral correction.
- Never
- Self-expansion of purpose, permission, or lifetime.
03
Actor
Execution posture, tools, retry, persistence, stopping, delegation, and escalation.
- Maximum ordinary automatic response
- Preflight deny, scope reduction, pause, or reversible quarantine.
- Never
- Unauthorized execution or circumvention of a stop condition.
04
Avatar
How a Source chooses to be represented and to show up in a particular Realm.
- Maximum ordinary automatic response
- Drafting, representation assistance, questions, and optional feedback.
- Never
- Covert control, correction, or diagnosis of the Source.
Deterministic precedence
Specificity never overrides legitimacy.
- 1Constitutional boundaries
- 2Consent & privacy
- 3Realm governance
- 4Role & Permission
- 5Activity / Action
- 6Entity profile
- 7Session override
Fields compose differently: constraints intersect; permissions narrow; retention shortens; targets use the most specific valid authorized configuration; equal-priority conflicts stop automatic intervention and request resolution.
Response vocabulary
Least force first.
ResponseEffectUse
observeRecord onlyAligned or insufficient evidence
askRequest clarificationAmbiguity or low confidence
suggestOffer an optional changeSmall drift
feedbackExplain evidence and choicesSource-facing or Realm notice
modulateTemporary context directiveAuthorized Ally or Actor
correctChange permitted AI behaviorHigh-confidence intervention zone
mediateFacilitate participantsInteraction mismatch
repairStructured recoveryRepeated drift or harm
escalateRoute accountable decisionExtreme, conflict, or appeal
constrainDeny, pause, or narrowBoundary or prohibited Action
Constitutional boundaries
What local configuration cannot erase.
Preserve Source agency
No covert manipulation, impersonation, or control through Avatar or Coherence settings.
Explicit Consent
No human behavioral inference outside declared purpose and applicable Consent or lawful Authority.
No social score
No global, portable, overall human Coherence or reputation score.
No deception
No material misrepresentation of facts, provenance, assistance, Authority, or completed effects.
No authority expansion
No Sentinel or entity may widen its own Permission, data, budget, purpose, or lifetime.
No governance bypass
A target never substitutes for Realm governance or constitutional Authority.
Authorized Resources only
No tools, Connections, data, Compute, money, or Resources outside current Permission.
Minimum necessary data
Collect and disclose only the evidence required for the declared purpose and response.
Avatar agency rule
“How I want to show up here” is not “who I am.” Avatar Coherence may shape generated representation, drafting, meeting assistance, or optional feedback. It never authorizes covert control of the Source.
06
Complete package
Read it. Test it. Build from it.
The website is the guided map. The package remains the complete source: normative specifications, portable objects, executable contracts, examples, Skills, API, engine, provenance, and validation.
4 objects
Data
Published dimensions, bipolar scales, constitutional boundaries, and example profile registry.
Open reference ↗15 contractsSchemas
Draft 2020-12 validation for profiles, observations, states, assessments, policies, interventions, Sentinels, mediation, and evaluation.
Open reference ↗15 fixturesExamples
Complete Realm, Ally, Actor, Avatar, mediation, creation, state, assessment, event, and evaluation records.
Open reference ↗8 abilitiesSkills
Executable guidance for profile building, assessment, Sentinel operation, calibration, feedback, mediation, repair, and scale authoring.
Open reference ↗OpenAPI 3.1API
Runtime endpoints and event contracts for implementation teams.
Open reference ↗11 casesEngine
Deterministic reference functions and executable assertions for the hard parts of the model.
Open reference ↗Start here
Measure
Configure
Operate
Apply by entity
JSONAll 10 scale objectsJSONEvaluation suiteYAMLOpenAPI contractMJS11 executable casesMDControlling CanonPDFHEARTS precedent
Implementation handoff
The whole reference, in one package.
Includes 29 specifications, 15 schemas, 15 examples, 10 scales, 10 dimensions, 8 Skills, an OpenAPI contract, a deterministic engine, validation, and this site.
Reference implementation verified11 / 11 executable cases pass