Escalation & Predictions
Most of Krataxis describes what has happened. This set of tools looks forward — estimating where a situation is heading, recognising patterns that precede escalation, and flagging the moment those patterns appear. They're distinct features that complement each other, so it's worth knowing which answers which question.
A word of framing first: these are decision aids, not verdicts. They direct your attention to where escalation is more likely; they don't decide for you, and credibility still governs the evidence underneath.
Predictions
Open the Predictions panel (Tools → Predict). It has two tabs.
Escalation predictions
A ranked set of escalation forecasts for the current scope. Each prediction shows:
- a probability bar — how likely the predicted escalation is,
- a factor breakdown — what's driving the estimate, so it isn't a black box, and
- an analyst rating control — your feedback on whether the prediction was sound, which improves future scoring.
Two filters keep the list useful: a minimum-probability slider to hide low-probability noise, and an Active only toggle to drop forecasts that have passed their horizon. When a battlespace is focused, predictions are scoped to it.
Always read the factor breakdown, not just the bar. A high probability driven by factors you don't trust is worth less than a moderate one built on solid corroboration.
Humanitarian risk
The second tab shows computed risk zones with population estimates — areas where the modelling indicates elevated humanitarian risk. Use it to connect the military picture to its civilian consequences, which is often the part leadership most needs to hear.
Escalation signatures
Open Escalation Signatures (Tools → Escalation). Where predictions estimate probability, signatures recognise patterns — known sequences of events that historically precede escalation. The panel has two tabs.
Matches
Signature matches detected for the active battlespace. Each shows its confidence, the number of matched events, and a predicted time horizon. Unacknowledged matches pulse red to draw the eye; click Acknowledge once you've seen one. The header's Escalation badge carries the unacknowledged count.
Library
The full catalogue of signatures, each shown as an ordered sequence of steps rendered as coloured category chips so you can read the pattern at a glance. Built-in signatures are read-only; you can create, edit, and delete your own to capture patterns specific to the situations you track — a way to encode an experienced analyst's instinct into something the system can watch for.
Auto-escalation suggestions
The Escalate tool (Tools → Escalate) surfaces suggestions — items the platform thinks may warrant escalation, with a badge showing how many are waiting. It's a fast triage surface for "the system noticed something; do you agree?"
How they fit together
| Tool | Question it answers |
|---|---|
| Escalation predictions | How likely is escalation, and why? |
| Humanitarian risk | Who is exposed if it does? |
| Escalation signatures | Are we seeing a known pre-escalation pattern right now? |
| Auto-escalation | Which items should I look at escalating? |
All four are forward-looking aids. As always, an analyst's judgement is the final word.
Where to next
- Credibility & Validity — the grading that underpins these estimates.
- The Intel Causal Graph — trace the chains that lead to escalation.