Confidence Drift tracks the 7-day trend in average event confidence scores for a crisis. When confidence is rising, it means the pipeline is extracting better-sourced, more precisely geolocated, more corroborated events. When it is falling, either the quality of available information is deteriorating or the underlying events are becoming harder to verify — which itself is a signal worth noting.
Each event in the POLYCRISIS database carries a composite confidence score (0–1) derived from four factors: source quality (wire service vs. blog), geolocation precision (city-level vs. country-level), corroboration (number of independent sources), and recency relative to the event. The system computes a daily mean confidence for each crisis, then fits a linear regression over the most recent 7 days. The slope of that regression, normalised by the mean confidence, is the Confidence Drift score.
Positive drift (above +0.1): Coverage is becoming more authoritative. This is normal following an initial breaking-news phase as wire services catch up with rumours. It can also indicate that previously covert activities are becoming visible.
Negative drift (below −0.1): Authoritative sources are becoming scarcer. This warrants investigation: is it an access blackout (journalists expelled, communications cut), or a genuine reduction in significant activity? Negative drift combined with high Escalation Probability suggests the former.
Derived directly from the confidence column of the POLYCRISIS events table. Updated daily.
One email a week: the biggest scenario-probability moves, new cross-crisis connections, and what to watch next. Free, unsubscribe anytime.