1     Definition 2     Functional Requirements 3     Syntax
4     Semantics 5    Conformance Testing 6     Performance Assessment

Definition

Knowledge – in the context of Theory of Mind (ToM) – is the digital representation of the epistemic state of what an Entity  knows, believes, or is aware of regarding a Proposition, along with supporting evidence, confidence, and temporal validity. Knowledge can be inferred or assigned.

Knowledge enables:

  • Attribution of mental states (knowledge, belief, false-belief, uncertainty) to Entities.
  • Reason about perspective (who knows what, and who thinks others know what).
  • Support compliance and governance by tracking provenance, justification, and validity intervals.

Functional Requirements

1. Representation of Mental States

  • Must represent beliefs, intentions, desires, and goals of agents (human or virtual) in a machine-readable format.
  • Should support uncertainty and confidence levels for inferred mental states.

2. Contextual Awareness

  • Must encode situational context (physical, social, temporal) to interpret mental states accurately.
  • Should allow dynamic updates as context changes.

3. Perspective Modeling

  • Must maintain separate knowledge graphs for each agent’s perspective (what they know, believe, or assume).
  • Should support perspective alignment and conflict detection between agents.

4. Inference and Prediction

  • Must enable inference of the likely next actions or decisions based on current  Entity State.
  • Should predict potential misunderstandings or conflicts and suggest resolution strategies.

5. Directive Integration

  • Must align ToM reasoning with active directives (e.g., collaborative tasks, negotiation goals).
  • Should allow injection of governance constraints into ToM reasoning.

6. Traceability and Governance

  • Must record provenance of mental state inferences (origin module, timestamp, confidence).
  • Should support audit trails for compliance and ethical reasoning.

7. Interoperability

  • Must integrate with other MMC Data Types (Intent, Context, etc.) for holistic reasoning.
  • Should expose APIs for AIMs like Dialogue Management, Action Planning, and Emotion Simulation.

8. Extensibility

  • Must allow domain-specific mental state extensions (e.g., gaming roles, educational goals).
  • Should support multilingual and cultural adaptation.

9. Performance

  • Should provide metrics for inference latency, accuracy, and coverage of mental state modeling.

Syntax

https://schemas.mpai.community/MMC/V2.5/data/Knowledge.json

Semantics

Label Description
Header Knowledge Header
– Standard‑EKN The characters “MMC‑EKN-V” marking the Motivation data type family.
– Version Major version – 1 or 2 characters.
– Dot‑separator The character “.” separating version components.
KnowledgeID Id of this knowledge.
EntityID Entity Identifier.
EntityType Categorisation of the entity. One of: a-user, h-user, process, group, unknown.
Proposition The proposition whose truth/belief is evaluated.
TruthValue Objective truth status of the proposition in the modeled world: true | false | unknown.
Belief Whether the Entity believes the proposition (boolean).
Awareness Whether the Entity is aware that it knows/believes the proposition (meta-cognitive flag).
EpistemicStatus Canonical classification of the epistemic state: known | believed | false-belief | uncertain | unknown.
Confidence Confidence (0..1) in the epistemic assessment. Distinct from Uncertainty model below.
Justifications Array of evidence items supporting the epistemic state. Each item includes type/strength/source/timestamp/note.
Justifications.type Kind of evidence (perception, testimony, memory, inference, document, sensor, other).
Justifications.strength Normalized weight (0..1) indicating evidential support.
Justifications.source URI or identifier of the evidence source.
Justifications.timestamp Time the evidence was observed/recorded.
Justifications.note Free-text remarks about the evidence.
Perspective Who attributes the knowledge to whom and with what certainty.
Perspective.AttributorID Entity ID that asserts/attributes the knowledge state.
Perspective.TargetID Entity ID believed to hold the knowledge/belief.
Perspective.EstimationConfidence Confidence (0..1) in the attribution assessment.
Temporal Temporal framing of observation and validity for this Knowledge instance.
Temporal.ObservedAt Time instant/interval when the knowledge state was observed.
Temporal.ValidityInterval Time interval during which the knowledge state is considered valid.
Trace Audit trail linking this instance to its origin and time, optionally enriched in extension.
Trace.Origin Reference to the AIM instance/process that produced this knowledge record.
Trace.Timestamp Creation/update time for this knowledge record.
Trace.Operation Operation name/action that produced the change (e.g., “ingest”, “infer”) – OPTIONAL EXTENSION.
Trace.ActorID An Entity or process ID responsible for the operation -OPTIONAL EXTENSION.
Trace.Rationale Short explanation of why the operation occurred – OPTIONAL EXTENSION.
Versioning Version/provenance meta-data for this record. Baseline is non-breaking.
Versioning.supersedes ID of the previous Knowledge instance superseded by this one.
Versioning.validityInterval Temporal interval indicating when this version is considered authoritative.
Versioning.subversion Numeric subversion component for internal revisioning.
Versioning.dotSeparator Fixed “.” to document version tokenisation in headers – Optional extension.
Versioning.lineage Ordered list of ancestor IDs for provenance chains – Optional extension.
KnowledgeGraphRef Grounds the proposition/evidence in an external knowledge graph – Optional extension.
KnowledgeGraphRef.graphURI Graph location (URI).
KnowledgeGraphRef.nodeID Primary node identifier linked to the proposition or entity  – Optional extension..
KnowledgeGraphRef.subgraph Subgraph selector or serialized subgraph reference  – Optional extension..
SocialContext Social relations/roles that affect interpretation, authority, or norms.
SocialContext.groups Group memberships relevant to the knowledge state – Optional extension.
SocialContext.roles Roles held by the entity (e.g., moderator, curator) – Optional extension.
SocialContext.authority Authority level or credential relevant to trust/weight – Optional extension.
SocialContext.relations Named relations among entities (e.g., parent-of, collaborator-with) – Optional extension.
TemporalContext Scenario/epoch framing beyond concrete timestamps – Optional extension.
TemporalContext.epoch Epoch label (e.g., “Q4-2025”, “post-event”).  – Optional extension.
TemporalContext.scenario Narrative or situational scenario tag.
Uncertainty Structured uncertainty model complementing scalar Confidence – Optional extension.
Uncertainty.model Type of uncertainty representation: point | interval | distribution | other – Optional extension.
Uncertainty.params Model parameters (schema-free object for extensibility) – Optional extension.
Uncertainty.bySource Per-source confidence breakdown array – Optional extension.
Uncertainty.bySource.source Identifier/label of the source – Optional extension.
Uncertainty.bySource.confidence Confidence (0..1) attributable to that source – Optional extension.
Attribution Credit/licensing/role information for claims/evidence – Optional extension.
Attribution.agentID Entity receiving credit/responsibility – Optional extension.
Attribution.role Role of the Entity with respect to the knowledge (author, curator, reviewer).
Attribution.license License string covering reuse of evidence or text – Optional extension.
Attribution.credit Human-readable credit statement – Optional extension.
Explanations Structured explanations linking evidence to conclusions – Optional extension.
Explanations.text Human/machine-readable explanation text – Optional extension.
Explanations.method Reasoning method used (abductive, deductive, causal, counterfactual, other) – Optional extension.
Explanations.evidenceRef List of identifiers/URIs pointing to supporting evidence items – Optional extension.
Predictions Forward-looking propositions inferred from current knowledge – Optional extension.
Predictions.proposition Proposition schema instance representing the predicted statement – Optional extension.
Predictions.confidence Confidence (0..1) in the prediction – Optional extension.
Predictions.horizon Temporal horizon for which the prediction applies – Optional extension.
Conflicts Contradictions detected across propositions or evidence – Optional extension.
Conflicts.with Identifier of the conflicting knowledge/evidence – Optional extension.
Conflicts.type Conflict type: logical | empirical | policy | other – Optional extension.
Conflicts.status Lifecycle: detected | under-review | resolved – Optional extension.
Conflicts.note Resolution notes or human comments – Optional extension.
AlignmentWithDirective Alignment assessment against a named directive/policy – Optional extension.
AlignmentWithDirective.directiveID Identifier of the directive/policy evaluated – Optional extension.
AlignmentWithDirective.alignment Alignment score (0..1) – Optional extension.
AlignmentWithDirective.note Notes describing rationale or caveats – Optional extension.
Governance Policy hooks: access, retention, consent – Optional extension.
Governance.access Access level: public | restricted | private – Optional extension.
Governance.retention Retention policy (duration or rule) – Optional extension.
Governance.consent Consent state or reference – Optional extension.
DescrMetadata Descriptive metadata for UX and cataloging – Optional extension.
DescrMetadata.title Short title – Optional extension.
DescrMetadata.summary Brief summary/abstract – Optional extension.
DescrMetadata.tags Array of tags for classification – Optional extension.
DescrMetadata.language Language code of textual fields – Optional extension.
DescrMetadata.format Format string for any attached content – Optional extension.

5     Conformance Testing

A Data instance Conforms with MPAI-MMC Motivation (MMC-EMT) if:

  1.  Its JSON Object validates against its JSON Schema.
  2. Any included  JSON Object validates against its JSON Schema.
  3. All Data in the JSON Object:
    1. Have the specified Data Types.
    2. Conform with the Qualifiers signaled in their JSON Schemas.

6     Performance Assessment