Is It a Nuke? | Interactive Deterrence

How to read it

Many missiles can carry either kind of warhead. James Acton calls the resulting uncertainty warhead ambiguity, and adds two relatives: destination ambiguity, when the observer cannot tell where a maneuvering weapon is headed and may wrongly conclude it is headed for its own territory, and target ambiguity, when a strike on conventional forces looks like a strike on nuclear ones because the two are entangled. China’s DF-26 intermediate-range ballistic missile is one example of a missile that can carry either warhead. This tool puts you in the seat of the state that detects the launch.

The Decision box in the side panel names the response with the lowest expected cost: the cost of each response if the truth is no attack, conventional or nuclear, weighted by how likely you think each one is. All costs are notional units; lower is better.
The belief chain starts from your priors and adds one piece of evidence per row: a second sensor, the launch site and the trajectory. Colors split your belief between no attack, a conventional strike and a nuclear one.
The cost chart weighs each response against that belief, and the strip below it shows the thresholds where the best response changes.
The campaign map runs the same reasoning over every launch a campaign could produce and asks how often a conventional strike is met with nuclear weapons. You read evidence through the entanglement you assume; the map varies the entanglement that is real.
The four tiles under the map count outcomes over many launches at the actual entanglement: how often a conventional campaign draws a nuclear launch, how false alarms are answered, and how many real nuclear attacks you would ride out.
The address bar stores every setting, so a link reproduces the setup.

Questions to try

In peacetime with one sensor, how high must the prior on a real attack go before waiting stops being best?
Set the assumed entanglement well below the actual level. Which trajectory produces the most wrongful launches, and why?
Switch off launch before impact. What is left of the danger, according to the model and to Acton?

Sources

James M. Acton, “Is It a Nuke? Pre-Launch Ambiguity and Inadvertent Escalation,” Carnegie Endowment for International Peace, 9 April 2020. carnegieendowment.org
James M. Acton, Silver Bullet? Asking the Right Questions About Conventional Prompt Global Strike (Carnegie Endowment for International Peace, 2013), on warhead, destination and target ambiguity. carnegieendowment.org (PDF)
James M. Acton, “Escalation through Entanglement: How the Vulnerability of Command-and-Control Systems Raises the Risks of an Inadvertent Nuclear War,” International Security 43, no. 1 (2018): 56–99. doi:10.1162/isec_a_00320
James M. Acton, “Why Is Nuclear Entanglement So Dangerous?” Carnegie Endowment, 23 January 2019. carnegieendowment.org
U.S. Department of Defense, Military and Security Developments Involving the People’s Republic of China 2024, p. 102, on commingled missile forces: “it may not be clear what the PRC launched until it detonates”; p. 87 on the DF-26, which “can conduct nuclear, precision conventional, and maritime attacks.” archived PDF
Caitlin Talmadge, “Would China Go Nuclear? Assessing the Risk of Chinese Nuclear Escalation in a Conventional War with the United States,” International Security 41, no. 4 (2017): 50–92. doi:10.1162/ISEC_a_00274
CSIS Missile Defense Project, “DF-26,” Missile Threat, last modified 23 April 2024, on the DF-26’s nuclear and conventional warheads. missilethreat.csis.org
Sources for each historical case are listed on its card.

Method · Updating

Three hypotheses: no attack (F), a conventional launch (C), a nuclear launch (N). Priors: Pr(real) and Pr(nuclear | real), so the prior is F = 1 − Pr(real), C = Pr(real)(1 − Pr(nuclear | real)), N = Pr(real) Pr(nuclear | real).

A second, independent sensor divides the weight on F by its likelihood ratio. Launch site and trajectory are treated as independent given the true state, and each multiplies the weights by Pr(evidence | state). The posterior is the normalized product. The multipliers shown are the nuclear-to-conventional likelihood ratios.

Entanglement. With entanglement e, a share of conventional launches comes from mixed sites (0.8e + 0.05) and heads for dual-use targets (0.05 + 0.45e) or for co-located forces that also serve your nuclear mission (0.02 + 0.25e); nuclear launches also shift to mixed sites. As e rises, site and target say less about the warhead. The observer computes likelihoods at its assumed e; outcomes are generated at the actual e.

Method · Deciding

Each response has a cost in each state (editable in the panel). The expected cost of a response is the belief-weighted sum; the model picks the lowest. In the nuclear state, waiting and conventional retaliation add a deterrent loss scaled by one minus survivability: forces that would be destroyed before they could respond. Launch on warning avoids that loss but is catastrophic if the attack was conventional or not an attack at all.

Campaign map. For each cell the model enumerates every site and trajectory, weights them by how often a conventional launch produces them at the actual entanglement, and sums the probability that the chosen response is launch on warning. The campaign figure is 1 − (1 − r)^K for K conventional launches, assuming independence.

Method · What is notional

The model is this tool’s own stylized construction. Acton’s work describes the ambiguities in words; it contains no Bayesian or decision-theoretic model, and none of the formulas above come from the sources. Every number is notional: priors, likelihoods, costs, the sensor weight and the campaign size. They were chosen so the model behaves as the sources describe, not estimated from data, and no public data exist to estimate them. The model is a teaching device about how beliefs and entanglement interact. It does not describe any real warning system or decision procedure.

What it leaves out. Time pressure and staged decisions; Acton’s central concern, pre-launch ambiguity, which can last for days (the two pre-launch cases show it); signaling and statements that could clarify a launch; and the adversary’s reaction to your response. Acton notes that post-launch ambiguity could spark a nuclear response only for states able to launch before weapons detonate, which the posture switch represents.