C o h e r e n c e

A diagnostic system for measuring organizational alignment

Every organization has a center — what it says about itself — and an edge — what the world actually sees. Coherence measures the gap between the two.

Center is the narrative an organization projects: press releases, job postings, earnings calls. Edge is what actually reaches the world: customer reviews, employee sentiment, social media, news coverage. The distance between center and edge is where coherence breaks down.

This is not a survey tool. It is not sentiment analysis. Coherence is a structural alignment measurement — built on publicly available data, scored through adversarial validation, and designed to make invisible failure observable.

It is a running system, not a slide deck. Thirteen pipeline runs completed. Structured outputs. Scored findings. A real diagnostic record.

Truth Authority Continuity

Truth

Does the center narrative match edge reality? The gap between what an organization says and what the world sees.

55% weight

Authority

Does the entity's voice carry weight and consistency? Signal strength across sources, earned through showing up.

45% weight

Continuity

Temporal consistency across collection periods. The through-line that connects past claims to present reality.

Phase 2

Overall = Truth × 0.55 + Authority × 0.45

Four stages, each producing structured outputs.

  • Collect
    Gather center sources (press releases, job postings, earnings transcripts) and edge sources (customer reviews, social media, news coverage, employee reviews). All publicly available.
  • Extract
    Agent-based extraction via local language models. Raw sources become structured claims and observations with schema validation.
  • Score
    Truth and Authority scoring with adversarial Skeptic challenge. Each finding is debated before it enters the evidence ledger.
  • Synthesize
    Scored case file with findings, failure modes, field notes, and overall narrative. The diagnostic record.
The Skeptic. Every finding is challenged before it enters the record. Sustained findings become evidence. The rest are discarded. This is how the pipeline earns its own authority.

Diagnostic Signature

Responsibility assigned without decision rights, combined with signal distortion and meaning drift.

Functional Field Notes · Early Signals

  • Signal Misread
  • Responsibility Compression
  • Context Collapse
  • Meaning Drift

Notice before it breaks.

Failure Modes · Breakdowns

  • Responsibility Without Authority
  • Metric Shadowing
  • Escalation Theater
  • Decision Displacement

Name the pattern.

Crosswalk

Signal MisreadMetric Shadowing
Responsibility CompressionResponsibility Without Authority
Context CollapseDecision Displacement
Meaning DriftEscalation Theater
AR-001 · Governing Constraint
"Automation may observe, summarize, and suggest. Automation may not decide."

AR-001 is baked into the pipeline architecture, not bolted on after the fact. The extraction agents observe. The scoring agents summarize. The Skeptic suggests which findings survive. But no automated system makes the final diagnostic call — that stays with a human who carries the context and the consequence.

The Skeptic enforces this structurally. It doesn't approve findings — it challenges them. What survives the challenge enters the record. What doesn't is discarded. The system is designed to resist its own confidence.