2nd/10_Wiki/Topics/Game_Design/Nuclear Deterrence Models.md

id, title, category, status, canonical_id, aliases, duplicate_of, source_trust_level, confidence_score, verification_status, tags, raw_sources, last_reinforced, github_commit, tech_stack

id	title	category	status	canonical_id	aliases	duplicate_of	source_trust_level	confidence_score	verification_status	tags	raw_sources	last_reinforced	github_commit	tech_stack
wiki-2026-0508-nuclear-deterrence-models	Nuclear Deterrence Models	10_Wiki/Topics	verified	self	Deterrence Theory MAD Nuclear Strategy	none	A	0.9	applied	game-design game-theory strategy geopolitics simulation		2026-05-10	pending	language framework game-theory strategic-modeling

Nuclear Deterrence Models

매 한 줄

"매 first strike 의 cost 의 retaliatory 의 unbearable 의 make → 매 attack 의 prevent". Nuclear deterrence 매 game-theoretic equilibrium 의 family — Mutually Assured Destruction (MAD), credible-threat signaling, escalation ladders. 매 Cold War-era foundation 의 modern simulations (Plan A by Princeton, ICONS) + 4X game design (Civilization, Hearts of Iron) 의 informs.

매 핵심

매 Foundational Models

• MAD (Mutually Assured Destruction): 매 second-strike survivability 매 first strike 의 deters.
• Massive Retaliation (Dulles, 1954): 매 any aggression 매 maximum response.
• Flexible Response (Kennedy): 매 graduated escalation — match 의 force level.
• Counterforce vs Countervalue: 매 military targets 의 vs population/cities.

매 Game-Theoretic Lens

• Chicken / Hawk-Dove: 매 brinkmanship — 매 first 의 swerve 의 lose.
• Schelling 의 Focal Points: 매 tacit coordination 매 escalation 의 around.
• Credible commitment: 매 burn-the-boats — 매 retreat option 의 remove.
• Signaling cost: 매 expensive signals (mobilization) 매 cheap talk 의 outweigh.

매 응용

1. 4X game design — diplomacy systems (Civilization VI/VII nuclear units).
2. Wargame simulation — Plan A, Sigma war games (Pentagon, RAND).
3. Geopolitical training — Foreign Service simulations.

💻 패턴

MAD payoff matrix

// Row = US strategy, Col = USSR strategy. Values: [US payoff, USSR payoff]
const madMatrix = {
  cooperate_cooperate: [3, 3],
  cooperate_strike: [-100, 1],
  strike_cooperate: [1, -100],
  strike_strike: [-100, -100],
} as const;

function nashCheck(strategy: keyof typeof madMatrix): boolean {
  // 매 strike_strike 의 not 매 deviating cooperate 의 worse
  return false; // mutual cooperation 매 not Nash 의 single-shot, repeated game 매 stable
}

Second-strike survivability

interface Arsenal {
  silos: number;
  submarines: number;
  bombers: number;
  silos_survival: number;     // ~0.05
  sub_survival: number;       // ~0.95
  bomber_survival: number;    // ~0.30
}

function expectedRetaliation(a: Arsenal, warheadsPerPlatform: number): number {
  return (
    a.silos * a.silos_survival +
    a.submarines * a.sub_survival +
    a.bombers * a.bomber_survival
  ) * warheadsPerPlatform;
}

function isMADStable(us: Arsenal, ussr: Arsenal, warheads: number, threshold = 200): boolean {
  return (
    expectedRetaliation(us, warheads) > threshold &&
    expectedRetaliation(ussr, warheads) > threshold
  );
}

Escalation ladder (Kahn)

const ESCALATION_LADDER = [
  { rung: 1,  name: 'Crisis declaration', kineticForce: false },
  { rung: 5,  name: 'Show of force', kineticForce: false },
  { rung: 10, name: 'Conventional war', kineticForce: true },
  { rung: 15, name: 'Tactical nuclear', kineticForce: true },
  { rung: 20, name: 'Strategic limited strike', kineticForce: true },
  { rung: 44, name: 'Spasm war (full exchange)', kineticForce: true },
];

function escalate(current: number, signal: 'firm' | 'reciprocate' | 'deescalate'): number {
  if (signal === 'firm') return Math.min(44, current + 2);
  if (signal === 'reciprocate') return current;
  return Math.max(1, current - 3);
}

Brinkmanship simulation (Chicken)

interface Player { resolve: number; rationality: number; }

function chickenOutcome(a: Player, b: Player): 'a_yields' | 'b_yields' | 'crash' {
  // 매 brinkmanship — both swerve 의 mutual benefit, 매 neither swerve 의 catastrophe
  const aYieldP = 1 - a.resolve;
  const bYieldP = 1 - b.resolve;
  const r = Math.random();
  if (r < aYieldP * (1 - bYieldP)) return 'a_yields';
  if (r < aYieldP * (1 - bYieldP) + bYieldP * (1 - aYieldP)) return 'b_yields';
  if (r < aYieldP * (1 - bYieldP) + bYieldP * (1 - aYieldP) + aYieldP * bYieldP) return 'a_yields';
  return 'crash';
}

Credible threat (Schelling commitment device)

class Commitment {
  // 매 retreat 의 cost 의 raise 하여 매 threat 의 credible 의 make
  constructor(
    public publicAnnouncement: boolean,
    public alliancesAtStake: number,
    public sunkInvestment: number,
  ) {}

  credibility(): number {
    return (
      (this.publicAnnouncement ? 0.4 : 0) +
      Math.min(0.4, this.alliancesAtStake * 0.1) +
      Math.min(0.2, this.sunkInvestment / 1e9 * 0.05)
    );
  }
}

매 결정 기준

상황	Approach
Symmetric superpowers	MAD + arms control treaties
Asymmetric (one weak)	Flexible response + extended deterrence
Multiparty (3+ nuclear states)	Stability erosion — 매 stochastic escalation risk
Game design (4X / wargame)	Schelling-flavored mechanics — credible-threat units, escalation tracker

기본값: MAD model 의 baseline + flexible response 의 layered nuance.

🔗 Graph

• 부모: Game Theory · Strategic Studies
• 변형: MAD · Massive Retaliation · Flexible Response
• 응용: Civilization Nuclear Units · Hearts of Iron Doctrine · War Game Simulations
• Adjacent: Algorithmic Rhetoric · Alliance (동맹) · 4X 시스템 (4X System)

🤖 LLM 활용

언제: Wargame scenario drafting, doctrine summary, simulation NPC rhetoric generation. 언제 X: Real-world policy advice, classified-domain reasoning, live crisis decision support.

❌ 안티패턴

• Single-strategy doctrine: 매 only-massive-retaliation 매 small-conflict 매 incredible.
• Ignoring second-strike: 매 silo-only arsenal 매 vulnerable, MAD 의 collapse.
• Cheap-talk threats: 매 cost-free threats 매 not credible — game-theoretic noise.
• Ladder collapse: 매 step-skipping 매 stability 의 destroy.

🧪 검증 / 중복

• Verified (Schelling "Strategy of Conflict", Kahn "On Escalation", Princeton SGS Plan A 2019, RAND deterrence literature 2020-2024).
• 신뢰도 A.

🕓 Changelog

날짜	변경
2026-05-08	Phase 1
2026-05-10	Manual cleanup — MAD, escalation ladder, credible commitment patterns