Energize an event across eight city domains, then test responses. Each response shows what it improves and what it makes worse, tagged by how many steps removed the effect is.
Ranked by how much total harm each removes on its own, starting one step after the event. Apply several to see them combined.
Each link says how much a one-tic move in the source pushes the target per step. Physical links are better grounded than social ones; low-confidence links are marked in red. Changes rerun everything.
Starter model, not calibrated. Weights are reasoned assumptions to be tuned against past incidents and exercise data. Use it to compare responses and surface knock-on effects for planning and exercises; it does not forecast.