Two teams, five sites, ten quarters. Price deterrence, commit to it out loud, then see where the game goes once cheating stops paying.
You run the verification directorate of the Meridian Accord. Pacifica has declared five compute sites; you have two inspection teams and ten quarters. Every quarter Pacifica weighs a diversion against your coverage. Your job is not to catch it. Your job is to make cheating a losing bet.
Stage 1 of 6 · read in
Two teams, five sites
Each declared site is worth a different amount to Pacifica if a diversion runs there undetected (the card shows it as G). A caught diversion costs Pacifica a fine of F = 150 in the same units. Under each card is the site's deterrence threshold: the fraction of quarters you would have to cover it to make cheating there break even. Notice the thresholds sum to barely more than one team. The arithmetic says two teams can deter all five sites. The game is about what it takes to actually collect on that arithmetic.
Meridian Array
G = 25 if a diversion succeeds
p* = 25/(25+150) = 0.14
Halcyon Port Annex
G = 35 if a diversion succeeds
p* = 35/(35+150) = 0.19
Cobalt Ridge Fab
G = 45 if a diversion succeeds
p* = 45/(45+150) = 0.23
Vantage Deepwater DC
G = 55 if a diversion succeeds
p* = 55/(55+150) = 0.27
Aurelia National Lab
G = 65 if a diversion succeeds
p* = 65/(65+150) = 0.30
Ten quarters. Quarters 1 to 5 you place teams by hand and Pacifica studies you. From quarter 6 you may commit to an announced policy. Somewhere along the way, expect the game to move.
Facilitator notes and debrief prompts
- The arithmetic is the anchor. Sum of thresholds = 1.13 teams; budget = 2. Ask the room why anyone ever gets away with anything if deterrence is this affordable. The answers (estimation under secrecy, commitment authority, undeclared sites) are the course.
- Watch for threshold players. Some learners set sliders to exactly p* everywhere and bank the slack. Ask what a zero-EV evader does when a real-world shock adds noise. Slack above the threshold is robustness, not waste.
- The role flip is the empathy beat. Learners who breezed through as inspector often hesitate as Pacifica, then comply. Name it: a well-built regime is boring to be inside.
- Tips split the room usefully. Following a 60 percent tip with both teams is a coverage collapse dressed as decisiveness. Ask who tilted versus who lurched.
- Nightjar is the exit question. Declared-site verification worked, so the game moved. What is the Nightjar of a compute-verification regime: unreported clusters, offshore training, algorithmic efficiency? There is no slider for what you cannot see.
Grounded in: Avenhaus and von Stengel on inspection games; IAEA safeguards practice (randomized interim inspections, timely-detection goals); the mixed-strategy deterrence result p* = G / (G + F). Atlantia and Pacifica are fictional; the arithmetic is not.