Evaluating Arms Control Treaty Monitoring System Confidence

Nuclear arms control treaties such as New START have included inspections where an inspecting treaty partner is permitted to employ agreed-upon monitoring and verification (M&V) processes and technologies (system) to agreed-upon items, to confirm treaty goals and objectives are being met by a hosting treaty partner. Historically, M&V activities have resulted in data that the inspecting party has used to count treaty accountable items (TAIs). Those counts were made with some degree of confidence. In this research, we aim to quantify such confidence using mathematical models. To that end, we developed a Bayesian framework assuming prior trust, characteristics of data collected via M&V systems, and variability and uncertainty in the components of such systems and the system overall. Here, we refer to confidence in the statistical sense—where an M&V system is used to make inference and conclusions about treaty compliance and confidence is an assessment of the quality of that conclusion. This is distinct from the confidence assessments and determinations made by the Department of State.  Data from individual processes and technologies within a monitoring and verification system are combined to make inference and draw conclusions, and those data or the conclusions drawn may be combined to reevaluate confidence over time. We developed a discrete event simulation that generates plausible data streams under a given set of conditions to study the impacts of variation and uncertainty on confidence, over a range of conditions anticipated in an as-designed M&V system. Such analyses can be used to compare M&V systems with respect to confidence, over several plausible M&V applications.  This work is performed within the PROACTIVE venture, sponsored by NNSA’s Office of Defense Nuclear Nonproliferation Research Development, NA-22.