Corrigibility and Nuclear Deterrence A Cognitive Constraint on Escalation, Not a Psychology of Irrationality

Nuclear deterrence is not maintained by force structure alone. It also depends on decision systems remaining able to discriminate signals, hold competing interpretations, revise those interpretations when diagnostic evidence arrives, and withhold irreversible action long enough for that work to finish. Acute stress, time pressure, cognitive load, and machine-mediated warning can change those operations. The changes are heterogeneous and must not be summed into a single distortion score. This article states a limited constraint. Let usable decision time be tau, the time needed to integrate contradiction be tau_c, and the time needed for independent verification of a machine recommendation be tau_v. The ratios rho = tau_c / tau and chi = tau_v / tau are architecture comparators, not biological thresholds. Deterrence remains cognitively viable only while competing interpretations stay revisable and at least one channel of checking can finish before commitment. Crossing that line does not imply nuclear use. It means a safeguard assumed by deliberative deterrence has become insufficient. The decision unit is an organization, not a single nervous system. Near-misses show that a corrupt channel plus stress need not close the margin if a second sensor, a doctrinal prior, or a duty officer still revises. Emerging technologies matter when they raise chi or make revision one-sided toward the frightening hypothesis. The implication is narrow: a deterrent relationship can remain structurally intact while the process required to interpret it becomes temporarily incorrigible.