Confidence Score Explained

A confidence score is an event-level assessment rooted in Forge's evidence qualification layer. Forge evaluates how strongly the available evidence supports the underlying event or claim — including source independence, corroboration depth, evidence quality, retrieval depth, contradiction signals, and unresolved uncertainty — and applies deterministic constraints when the evidence is thin.

This is event confidence, not publisher credibility. Publisher reputation is one source-quality input, but it is not the verdict. A highly reputable outlet may break a story before independent confirmation exists. In that case the event can remain lower-confidence or capped until corroboration improves. USAI also distinguishes truly independent support from repetition or syndication of the same reporting chain.

Confidence is time-stamped and dynamic. It can rise, fall, or remain capped as reporting changes, contradictions emerge, article retrieval improves, or new evidence alters the event posture. The purpose is to expose uncertainty and change over time, not to manufacture a permanent verdict.

Confidence scoring combines multiple structured inputs: source quality and independence, corroboration depth, article-body retrieval status, evidence directness, contradiction and dispute signals, unresolved questions, and event maturity. Deterministic caps prevent a thin evidence base from being presented as settled simply because one source is prestigious or the story is widely repeated.

Scores may change as independent sources emerge, claims are corrected, source chains are reclassified, contradictions are resolved, retrieval improves, or new evidence materially changes the qualified event posture.

Confidence scores are designed as a starting point, not a conclusion. USAI Media recommends users:

• Compare outputs with primary sources, established reporting, and official information.
• Treat high-confidence scores as a reason to read more carefully — not to stop reading.
• Treat low-confidence scores as a monitoring signal, not a dismissal of the underlying claim.
• Read the source posture, confirmed facts, pending facts, evidence gaps, and what-to-watch indicators alongside the score.
• Use USAI as an intelligence layer that helps structure uncertainty, not as a substitute for primary evidence or accountable reporting.

USAI is built to show users not only what the current confidence posture is, but why it is there and what evidence would cause it to change. The score is useful because the reasoning around it remains visible.