Open Autonomous Intelligence Initiative

Depends on: OAII Base Model — Common Types v0.1; OAII Base Model — World v0.1.1; OAII Base Model — Event v0.1.1; OAII Base Model — Signal v0.1; OAII Base Model — Sensor v0.1

1. Purpose

The Knowledge object represents structured, retained, and interpretable information derived from Signals, Events, and Sensor Knowledge within a World. Knowledge enables continuity, learning, and comparison over time without collapsing into continuous surveillance or opaque global models.

In the OAII Base Model, Knowledge is explicit, scoped, revisable, and accountable.

---

2. Definition (Normative)

Knowledge is an abstract object that:

• MUST be scoped to a specific World;
• MUST be derived from Signals, Events, Sensor Knowledge, or combinations thereof;
• MUST be explicitly represented and referenceable;
• MUST support revision and versioning;
• MUST NOT imply diagnosis, intent, or psychological state by default.

Knowledge is not raw data, a live sensor feed, or an implicit model state.

---

3. Scope and Non‑Scope

In Scope:

• Baselines, norms, and learned patterns
• Aggregated or summarized Events
• Reference models used for comparison
• Sensor Knowledge promotion into durable form
• World‑scoped historical context

Out of Scope:

• Continuous surveillance archives
• Global or cross‑World learning without mediation
• Medical or clinical knowledge claims
• Undocumented or opaque internal model state

---

4. Core Responsibilities

The Knowledge object:

• Persists information beyond the lifetime of individual Signals or Events;
• Supports comparison between current observations and historical context;
• Maintains interpretability through explicit structure and metadata;
• Enables continuity across World evolution;
• Supports accountability via traceability and revision history.

---

5. Required Attributes (Abstract)

A conforming Knowledge object MUST expose the following attributes using shared types from Common Types v0.1.

5.1 Identity and Context

• knowledge_id : string
• world_ref : WorldId

5.2 Knowledge Classification

• knowledge_type : string
  Describes the role (e.g., baseline, routine, threshold set, reference pattern, summary).

5.3 Source and Provenance

• source_refs : string[1..N]
  References contributing Signals, Events, Sensors, or prior Knowledge objects.
• creation_time : ParameterSet

5.4 Validity and Confidence

• confidence : Confidence
• validity_constraints : ParameterSet (optional)

5.5 Privacy and Access

• privacy_class : PrivacyClass
• access_class : AccessClass

---

6. Knowledge States (Normative)

A Knowledge object MUST support at least the following states:

• PROVISIONAL: recently created, subject to change
• ACTIVE: valid and in use
• REVISED: updated due to new evidence
• DEPRECATED: retained for continuity but no longer authoritative
• INVALIDATED: withdrawn due to error or context change

Knowledge state MUST be explicitly tracked and visible.

---

7. Knowledge Mechanism Bundle

Knowledge formation and update MUST be represented using a mechanism bundle:

• knowledge_mechanism_id : MechanismId
• knowledge_mechanism_params : ParameterSet
• knowledge_mechanism_result : ResultSet (optional)

This bundle documents how Knowledge was derived or updated.

---

8. Relationships to Other Objects

• Signals contribute raw evidence
• Sensor Knowledge may be promoted into Knowledge
• Events may be aggregated or summarized into Knowledge
• World scopes meaning and validity
• Policies may reference Knowledge for evaluation
• Agents may reason over Knowledge
• Logs record Knowledge lifecycle changes

Knowledge MUST NOT bypass Events to directly drive action without policy mediation.

---

9. Methods (Normative, Abstract)

To standardize behavior, a conforming Knowledge object SHOULD support the following abstract methods:

9.1 derive_knowledge

• method_params : ParameterSet (source references, derivation rules)
• method_result : ResultSet

9.2 revise_knowledge

• method_params : ParameterSet (new evidence or context)
• method_result : ResultSet

Requirements:

• MUST update knowledge state appropriately.
• SHOULD preserve revision history.

9.3 evaluate_validity

• method_params : ParameterSet
• method_result : ResultSet (validity assessment)

---

10. Edge‑Primary Constraints

In Edge‑primary systems, Knowledge:

• MUST be storable and usable locally;
• SHOULD support bounded retention and aging;
• MAY be selectively synchronized across Worlds or systems;
• MUST respect privacy and access constraints at all times.

---

11. Failure and Degradation Modes

Knowledge degradation may include:

• stale baselines
• overfitting to outdated routines
• insufficient evidence coverage

Systems MUST surface uncertainty rather than forcing false authority.

---

12. Notes for Implementers (Non‑Normative)

• Sensor Knowledge is a specialized, often transient form of Knowledge; promotion to durable Knowledge should be explicit.
• Knowledge objects should remain small, interpretable, and replaceable.
• Aggregation intervals and retention policies should be World‑specific.
• Knowledge exists to support Events and Policies, not to replace them.

---

Status: Draft v0.1 (Normative Object Definition)