Nuclear Entanglement | Interactive Deterrence

How to read it

You command a conventional campaign against a nuclear-armed adversary, the target. Your aim is conventional: sink ships, suppress missiles, blind sensors. The trouble is that the target’s assets are entangled. The same early-warning satellites, command links, launchers, bases and surveillance networks serve its conventional and its nuclear forces. James Acton calls this entanglement and argues it could turn a conventional war nuclear even if neither side intends it.

The planner has five generic asset categories and four campaign phases. Click or tap cells to strike; click again to cancel. Each category shows its nuclear role, the share of it that serves nuclear forces.
The risk chart stacks each escalation channel’s hazard per phase and draws the cumulative chance of nuclear use against the same war without entangled strikes.
The channel cards show why the risk rises: use-or-lose pressure, misinterpreted warning, the damage-limitation window, and warhead and target ambiguity, all amplified by the fog of war.
The big number in the side panel is the chance that the target uses nuclear weapons by the end of phase 4. Conventional effect (0 to 100) scores the military value of the damage you did: what you gain. Use-or-lose pressure is how hard the losses push the target to use its nuclear weapons before it loses them, and fog of war how much of its warning, command and surveillance picture is gone.
The levers in the side panel apply mitigation measures from the literature, each with its source, and show what they buy and what they cost in conventional effect.
The address bar stores the plan, levers and every parameter, so a link reproduces the setup.

Questions to try

Strike only early warning. The target’s nuclear forces are untouched. Why does the risk still rise?
Delay strikes on command links to phase 3. How much does timing matter compared with the choice of target?
Turn on strike exclusions, then look at the conventional effect. Which categories are hardest to spare?
Lower the target’s confidence in its second strike. Which channel responds, and why that one?

Sources

Sources for each real-world example are listed on its card. This page describes asset categories in general terms only. It names no real locations or units and contains no targeting detail.

Method · The campaign

Each category c has a nuclear role n_c (share serving nuclear forces), damage per phase struck d_c and a conventional value v_c. Damage accumulates: after a phase with a strike, D_c ← 1 − (1 − D_c)(1 − d_c); after a phase without one, a share of it is repaired. Damage to the nuclear-serving share is n_cD_c.

Each channel k sums weighted nuclear damage across categories, X_k = Σ_c w_kc n_c D_c. The weights encode which categories feed which mechanism: early warning and command links drive misinterpreted warning; launchers and co-located bases drive use-or-lose pressure and ambiguity; surveillance drives the damage-limitation window.

Fog. Degraded warning, command and surveillance add up to a fog level F between 0 and 1. The target fills the gap with worst-case assumptions: the use-or-lose, warning and ambiguity channels are multiplied by 1 + φF.

Method · From pressure to probability

Per-phase hazards: use-or-lose h = κ_ciX_ci(1 + φF)(1.5 − s), where s is the target’s confidence in its second strike; misinterpreted warning h = κ_mwX_mw(1 + φF); ambiguity h = κ_waX_wa(0.25 + a)(1 + φF), where a is the share of your weapons that are dual-capable; the damage-limitation window h = κ_dlX_dl, only if the target has that doctrine.

The chance of nuclear use in a phase is 1 − exp(−(h₀ + Σh)), and the campaign total compounds the four phases. Levers scale named terms: separation halves every n_c; exclusions cut nuclear damage by 70% and remove that share’s conventional value; restraint halves misinterpreted warning and cuts ambiguity by a third; communication cuts both by up to 45%, less as fog rises; resilient NC3 halves fog.

Method · What is notional

Every number is notional: nuclear roles, damage rates, channel weights, baseline hazard, lever strengths and the resulting probabilities. They were set so that the model’s ordering matches the arguments in the sources, not fitted to data. No data set exists from which such numbers could be estimated. The model shows which way risk moves and why, not how likely nuclear war is.

What it leaves out. Cyber operations, which Acton treats separately; the attacker’s own entangled assets and the target’s strikes on them; bargaining and war termination; the target’s conventional options short of nuclear use; and learning over the campaign. Acton’s damage-limitation window concerns a state that holds its opponent’s nuclear forces at risk; here it applies to the target only when you switch that doctrine on.

All the real-world examples below the model come from public, sourced reporting and official documents.