Event Intensity measures the severity-weighted volume of crisis events over a rolling 48-hour window. Unlike a raw count, it weights each event by its type severity (nuclear > military > naval > political > humanitarian > news) and by the AI-assigned confidence score for that event. The result is a dimensionless score that rises when many serious, high-confidence events occur in quick succession.
The indicator captures the "heat" of a crisis — not just whether things are happening, but whether they are alarming things happening reliably. A single verified airstrike counts more than ten unverified protest reports.
Each event in the 48-hour window is scored as: type_weight × confidence × recency_decay. Type weights run from 1.0 (news) to 5.0 (nuclear). Confidence is drawn directly from the event record (0–1). Recency decay halves the contribution of events older than 24 hours, and halves again at 48 hours. All event scores are summed and normalised against the 90th-percentile value observed for that crisis over the past 90 days.
Normalisation ensures comparability across crises with very different baseline activity levels — the Iran crisis typically produces far more daily events than the AGI risk tab.
High (above 0.75): Severe, well-documented events are clustering in time. This is unusual and warrants close attention. Sustained high intensity over 72+ hours has preceded significant escalations in the historical record.
Low (below 0.25): The crisis is in a quiet phase — either genuinely dormant or experiencing a news blackout. Distinguish between the two by cross-referencing Signal Velocity and Confidence Drift; a blackout typically causes Confidence Drift to decline simultaneously.
Computed from the events database, updated continuously as new AI-extracted events are inserted. Practical update lag is 15–30 minutes from publication of the underlying news source.
One email a week: the biggest scenario-probability moves, new cross-crisis connections, and what to watch next. Free, unsubscribe anytime.