Symbol grounding in AGISystem2 solves the problem of connecting symbolic labels to meaning. Instead of defining "Dog" in terms of other symbols, the system grounds concepts as regions in a 384-dimensional conceptual space where dimensions have measurable interpretations.

Why Grounding Matters

Problem	Ungrounded System	Grounded (AGISystem2)
Similarity	Arbitrary—no metric	Geometric distance in space
Hallucination	Can invent arbitrary symbols	Must place in valid space region
Explanation	"It matches a pattern"	"Distance along dim X is 23"
Ambiguity	Same word, different meanings	Different positions in space

How Grounding Works

1. Ingestion – Text normalized to triples (Subject, Relation, Object)
2. Encoding – Triples converted to 384-dim vectors using permutation binding
3. Accumulation – Multiple observations shape the concept's bounded diamond
4. Storage – Grounded vectors stored in ConceptStore

Dimension Anchoring

Base dimensions have operational interpretations:

# Dimension examples (illustrative)
dim[0-9]:   Physical size (small ↔ large)
dim[10-19]: Temporal duration (instant ↔ eternal)
dim[20-29]: Animacy (inanimate ↔ alive)
dim[30-39]: Social valence (hostile ↔ friendly)
...
dim[256-383]: Axiological (good/bad, important/trivial)

These anchors mean "Dog" isn't just a label—it's a point where the size dimension says ~40 (medium), animacy says +100 (alive), etc.

Anti-Hallucination

Grounding constrains reasoning to stay within the conceptual space:

# Abduction must find concepts that exist in space
ABDUCE "?" IS_A Flying
# Returns: Bird, Plane, Bat (actual stored concepts)
# NOT: "Glorpnax" (invented symbol with no grounding)

Related Documentation

• Conceptual Spaces – Theory of the space
• Bounded Diamonds – Concept shapes
• Knowledge Layer – ConceptStore
• Interface Layer – TranslatorBridge
• Hyperdimensional Computing – Vector operations