In AGISystem2, ontology refers to the factual dimensions of the vector space—dimensions 0-255 that encode what things are, not what they're worth. This separation from axiological dimensions enables unbiased reasoning when needed.
The 384-dimensional vector is partitioned: 256 dimensions for ontology (facts about the world) and 128 for axiology (values and judgments). This separation allows reasoning about facts independently of values.
Ontological dimensions capture what is—properties that can be observed or measured:
| Category | Examples | Typical Relations |
|---|---|---|
| Taxonomic | Species, categories, types | IS_A, HAS_INSTANCE |
| Mereological | Parts, components, members | HAS_PART, PART_OF |
| Spatial | Location, containment, proximity | LOCATED_IN, NEAR |
| Temporal | Sequence, duration, simultaneity | BEFORE, DURING, AFTER |
| Causal | Causes, effects, dependencies | CAUSES, REQUIRES |
The separation enables veil of ignorance reasoning: by masking axiological dimensions, the system can compare concepts purely on facts, ignoring value judgments. This is essential for:
The BiasController manages this separation, automatically applying ontology-only masks when fairness is required.
// Ontology mask: only dims 0-255 active const ontologyMask = new Uint8Array(384); ontologyMask.fill(1, 0, 256); // dims 0-255 = 1 ontologyMask.fill(0, 256, 384); // dims 256-383 = 0 // Pure factual comparison const factualDist = mathEngine.maskedL1(conceptA, conceptB, ontologyMask);