Skip to content

Confidence Interpretation

AI1SAD confidence is an operational data-quality indicator. It is not a probability of an encounter or attack.

Components

  • Coverage confidence: how many expected source categories are present.
  • Freshness confidence: whether available signals are current enough for the requested lookback.
  • Completeness confidence: how much important information is missing.

Missing Data

Missing weather, SST, vessel, biological, human exposure, sighting, incident, reef, or regional-pack data lowers confidence. The model should report missing data rather than silently assuming normal conditions.

Stale Data

Stale sources reduce freshness confidence. Stale signals should not be allowed to create strong operational alerts unless other current signals support the output.

Practical Reading

  • High confidence: enough relevant source categories are present and current.
  • Moderate confidence: useful for review, but some sources are missing or stale.
  • Low confidence: interpret cautiously and seek local authority or provider data.
  • Very low confidence: insufficient data for strong operational conclusions.