Human Signal Index
Documentation

Executive Intelligence & Early Warning

How HSI detects emerging risks: deterministic early-warning alerts, risk categories, forecast divergence, executive briefs, governance, and transparency rules.

Executive Intelligence is a detection layer, not a prediction engine. It surfaces what deserves attention now, using deterministic rules over real platform data. Every alert is explainable and reproducible, and human review remains authoritative.

Purpose

Answer "what should I be paying attention to right now?" without black-box scores or AI risk ratings. It detects; it does not predict.

Methodology

Each signal is computed from existing data (signals, evidence, narratives, contradictions, trust, forecasts, geo, review queue) using fixed, documented thresholds. The same inputs always produce the same output.

Alert types

Eight deterministic warnings: Narrative Surge, Contradiction Spike, Evidence Gap, Forecast Divergence, Trust Degradation, Geographic Expansion, Review Backlog, and Source Failure. Each alert states the trigger and the numbers behind it.

Risk categories

Eight categories — Narrative, Trust, Evidence, Forecast, Operational, Collection, Review, and Geo — each shown as GREEN, YELLOW, or RED via transparent ratio/backlog thresholds. No hidden scoring.

Forecast divergence

Recorded forecasts are classified against their observed outcomes as on-track, diverging, missed, or unresolved. This reads real evaluation results; it never generates a forecast.

Executive briefs

A concise, human-readable summary built only from platform deltas: what changed, what is accelerating or slowing, what needs review, which forecasts need attention, trust issues, and remaining blind spots. No generated prose.

Governance

Alerts and risks are advisory. They inform human reviewers and never publish, approve, or change a finding on their own.

Transparency rules

No black-box scores, no invented confidence metrics, no AI risk ratings. Every alert is traceable to its inputs and reproducible from the data.

Limitations

Some signals use the best available real proxy (for example, geographic footprint stands in for temporal expansion until per-entity history is wired). Thresholds are starting points and may be tuned with operational experience.