The Perception Confidence Engine (PCE) Explained
PCE is the system that computes a single 0–1 confidence score representing how clearly an AI system can build a coherent perception of a domain from its available content signals.
The Perception Confidence Engine produces the perceptionConfidenceScore: a single 0–1 composite score attached to every PerceptionGraphSnapshot. It answers one question: given the full content of this domain — its pages, entities, schema, links, and semantic signals — how clearly can an AI system build a coherent, accurate, and retrievable understanding of what this domain is, what it knows, and what it offers?
The Four PCE Inputs
PCE is computed from four measurable signals. Schema density: the proportion of pages with valid, complete schema markup of the correct type — higher schema density reduces AI inference work. Link density per entity: how many internal links connect pages discussing the same entity, higher link density creates stronger entity continuity signals. Heading structure quality: how consistently headings describe their content and maintain logical hierarchy, better heading structure produces cleaner chunk extraction. Content depth: average word count and claim density per page — deeper content provides more entity attributes and factual grounding.
Why PCE Is Stored on the Perception Graph Snapshot
The perceptionConfidenceScore is stored as a column on the PerceptionGraphSnapshot table — not on the audit_scores table. This placement is intentional: PCE is a property of the Perception Graph itself, not an independent audit score. As the Perception Graph is updated (on each audit, as the AI derives new inferences from updated content), the PCE is recomputed and stored with the new snapshot. Historical PCE values remain on historical snapshots.
●PCE ranges: 0.8–1.0 (high confidence — AI systems can build a reliable, complete perception of this domain), 0.5–0.79 (moderate confidence — perception is possible but some entities or topics have gaps), 0.0–0.49 (low confidence — AI systems cannot reliably distinguish or understand this domain's primary entities and offerings).
PCE in the Machine Trust Intelligence Score
PCE feeds into the Machine Trust Intelligence Score as a component of the Entity Authenticity Confidence input. A domain with consistently high PCE across multiple audit snapshots demonstrates stable, coherent AI perception — the AI system's understanding of the domain is not drifting or degrading. This stability is a positive trust signal. A domain with declining PCE across audits is showing semantic drift — the AI system's perception is becoming less coherent over time.
Perception vs Fact Layer — Part 3 of 10