Each 0-100 score is derived from the live result set. The cards show both the output and the concrete drivers behind it.
Flags when the platform getting engagement is not the platform carrying most mentions.
Combines corpus negativity with atrocity, humanitarian, Darfur, and Kordofan labels.
Compares the latest daily bucket with the prior average in this short history.
Shows whether attention is spread across the architecture or concentrated in a few buckets.
Measures whether the feed is centered on armed actors and active fronts.
Shows how much the conversation is being pulled into diplomacy and outside-power framing.
Tracks whether the SAF-RSF actor layer is balanced or dominated by one side.
Measures how much of the configured listening architecture is quiet in the current window.
These are operational heuristics, not ground-truth conflict measurements. They are useful because they expose skew and pressure in the current collection window.
Absolute gap between the top engagement platform share and its mention share
50% corpus negativity + 35% harm-label density + 15% peak harm-campaign negativity
Log-scaled latest daily volume compared with the prior daily average
60% top-two campaign share + 40% top-keyword share
35% war-campaign share + 65% armed-actor/frontline label density
65% politics-campaign share + 35% foreign-label strength vs the top keyword cluster
Absolute RSF-SAF label gap divided by total RSF + SAF labels
80% quiet-platform share + 20% quiet-campaign share
