Skip to content

Latest commit

 

History

History
213 lines (170 loc) · 11.3 KB

File metadata and controls

213 lines (170 loc) · 11.3 KB

Geopol Forecaster / Geopol Modeller — Use Case Roadmap

Potential applications of multi-actor LLM simulation beyond the current Iran-Israel conflict scenario. The core engine is domain-agnostic; these use cases require only new scenario YAML, actor clusters, and optional graph configs.

Geopolitical Forecasting & Conflict Simulation

The primary use case. Fast-moving crises where affected populations need actionable predictions.

Scenario Actor Complexity Notes
Iran-Israel war (active) 25+ actors Current focus. Internal dynamics critical.
South China Sea escalation 15-20 actors US, China, Taiwan, Philippines, Japan, ASEAN
Ukraine front-line evolution 15+ actors Military + diplomatic + domestic politics
Korean Peninsula crisis 10-15 actors DPRK, ROK, US, China, Japan
India-Pakistan escalation 10-15 actors Nuclear dimension, Kashmir trigger
Horn of Africa (Ethiopia, Sudan) 15+ actors Multi-ethnic, multi-state, humanitarian

Foreign Policy Simulation

Testing policy options before implementation. A government analyst could define their country's decision-maker as a controllable actor and test different policy choices against the responses of other actors.

Scenario Use Notes
Sanctions impact modelling Test sanctions regimes against target-state responses Economic + political actors
Alliance stress-testing Model how alliance commitments hold under pressure NATO Article 5 scenarios
Treaty negotiation simulation Pre-negotiate by modelling counterpart responses Trade, arms control, climate
Diplomatic recognition scenarios Model cascading effects of recognition decisions Taiwan, Palestine, Kosovo

Worst-Case Scenario Planning

"Pre-mortems" for geopolitical risks. Run the simulation with nature set high and the shock taxonomy configured for maximum disruption.

Scenario Purpose
Strait of Hormuz closure Energy security planning for import-dependent states
Suez Canal blockage (military) Global supply chain stress test
Simultaneous multi-front war Force planning and readiness assessment
Nuclear escalation ladder At what point do actors cross the nuclear threshold?
Cyber-physical attack on infrastructure How do states respond to ambiguous attacks?

Domestic Policy Simulation

The actor-based approach applies to any multi-stakeholder decision environment. Domestic policy involves actors with competing interests, constraints, and red lines — exactly what the simulation models.

Scenario Actors Notes
Noise pollution regulation Municipal council, residents, businesses, enforcement, developers, nightlife industry Classic NIMBY multi-stakeholder
Housing policy reform Government, developers, NIMBYs, renters, banks, construction unions Red lines and coalition dynamics
Public transit planning City planners, commuters, car lobby, environmental groups, budget office Trade-offs and competing priorities
Healthcare system reform Government, hospitals, insurers, doctors' unions, patients, pharma Institutional constraints
Education curriculum reform Ministry, teachers' unions, parents, religious groups, employers Cultural red lines
Climate adaptation planning City, residents, industry, environmental groups, insurers, federal gov Long-term vs. short-term tension

Economic & Market Simulation

Model market participants as actors with strategies, constraints, and information asymmetry.

Scenario Actors Notes
Trade war escalation Governments, industries, supply chains, consumers, WTO Tariff/retaliation cycles
Central bank policy coordination Fed, ECB, BoJ, BoE, emerging market CBs, markets Interest rate game theory
Sovereign debt crisis contagion Debtor state, creditor nations, IMF, markets, neighbours Cascading default modelling
Energy transition dynamics Oil producers, renewables, governments, utilities, consumers Stranded asset risk
Tech regulation Governments, tech platforms, users, advertisers, civil society Regulatory arbitrage
Sanctions impact wargaming Target state, sanctioning coalition, neutral states, markets, smuggling networks Each model as a national economic perspective

Predictive Forecasting & Verification

Structured prediction through adversarial multi-model analysis.

Scenario Actors/Lenses Notes
Election outcome forecasting Polling analyst, economic lens, sentiment analyst, historical pattern analyst Multipolar analytical lenses, not role-play
Ceasefire durability scoring Each stakeholder assesses agreement survivability from their position Post-agreement stress-testing
Treaty compliance monitoring Compliance advocate, violation hunter, adjudicator Adversarial analysis structure
Disinformation campaign attribution Linguistics analyst, network topology, timing analyst, narrative analyst Multi-domain forensic analysis
Climate-security nexus forecasting Climate modeller, livelihood analyst, political analyst, conflict researcher Causal chain modelling
Generational leadership transition Successor factions, old guard, military, external patrons Succession dynamics in opaque regimes

International Institutional Simulation

Modelling formal international bodies with well-defined rules, voting records, and quantifiable preferences. These are particularly amenable to simulation because their procedures are documented and their actors' historical positions are on record.

Highly Modelable Bodies

Body Members Why Simulable
UN Security Council 15 (5 veto) Complete voting records, veto mechanics, narrow mandate
OPEC+ ~23 Quantified production quotas, transparent economic incentives
European Council 27 QMV mathematics, population-weighted votes, VoteWatch data
FOMC 12 voting Published transcripts, dot plots, estimable reaction functions
IAEA Board 35 Nuclear mandate, non-proliferation voting patterns
IMF Executive Board 24 Weighted voting, lending decision frameworks
FATF Plenary 39 Defined scoring criteria, grey/black list decisions

Moderately Modelable Bodies

Body Members Notes
Arctic Council 8 Consensus, geographic focus, climate + sovereignty
Nuclear Suppliers Group 48 Consensus, India membership blocking dynamics
OPCW Executive Council 41 Compliance disputes, Syria precedent
ICJ 15 judges Published voting records and opinions
IPCC Approval Plenaries 195 SPM line-by-line negotiation

Humanitarian & Crisis Response

Simulate coordination failures and resource allocation in crisis response.

Scenario Actors Notes
Refugee flow prediction Origin, transit, destination countries, UNHCR, smugglers Coordination failure modelling
Pandemic response coordination WHO, national governments, pharma, public, media Information asymmetry critical
Natural disaster response National/local government, military, NGOs, affected population, donors Resource allocation
Maritime chokepoint disruption Shipping, insurers, navies, port states, commodity traders Cascading supply chain effects

Democratic Resilience & Constitutional Stress-Testing

Simulate constitutional crises and institutional resilience. Each branch of government, the military, civil society, and the judiciary operate as actors with their own red lines and institutional incentives.

Scenario Actors Notes
Constitutional crisis red-teaming Executive, legislature, judiciary, military, civil society Democratic backsliding scenarios
Post-conflict power-sharing design Ethnic/religious factions, military, external guarantors Iteratively negotiate and stress-test provisions
Electoral dispute escalation Incumbent, challenger, courts, military, street, international observers Contested election dynamics

Ideological Lens Analysis (Panel Mode)

Not role-play but analytical framework simulation. Each "actor" is an analytical perspective applied to the same problem. Draws from the Panel-Of-Claude two-round deliberation pattern.

Framework Set Lenses Use
IR theory Realist, liberal institutionalist, constructivist, Marxist, postcolonial Foreign policy analysis
Economic schools Keynesian, monetarist, Austrian, MMT, development economics Fiscal/monetary policy
Ethical frameworks Utilitarian, deontological, virtue ethics, care ethics Bioethics, AI ethics, policy ethics
Security studies Traditional, human security, critical security, feminist security Threat assessment

Integration Patterns from Related Projects

From AI-Agent-UN

  • Voting simulation: Define a resolution and have actors vote with structured justification. Applicable to any scenario ending in a formal decision (treaty, sanctions vote, policy adoption).
  • Bilateral impact analysis: Post-simulation assessment of how the outcome affected pairwise relationships.

From Peace In The Middle East

  • Chamber structure: Plenary sessions + working groups + back-channels. Applicable to any negotiation scenario.
  • Facilitator agent: Neutral moderator that manages speaking order and enforces procedural rules. Useful for diplomatic simulations.
  • Sub-faction decomposition: The granular actor modelling approach (150+ actors for the Middle East) sets the standard for fidelity.

From Panel-Of-Claude

  • Two-round deliberation: Independent analysis → cross-pollination. Applicable to policy analysis scenarios where you want multiple analytical frameworks rather than role-playing.
  • Tension mapping: Instead of declaring winners, map where genuine value conflicts exist. Better output for policy analysis than narrative adjudication.
  • Motion decomposition: Break complex problems into problem statements, causal claims, and implicit questions before simulation.

From Claude-AI-Conference

  • Agent clustering: Organise large actor sets into thematic clusters. Already implemented via actor cluster YAML.
  • Identity generation: Generate credible fictional identities for actors when simulating generic roles (e.g., "a municipal council member" rather than a named individual).
  • Broadcast mode: Not all simulations need interactive debate. Some scenarios benefit from each actor independently assessing the situation, with synthesis happening post-hoc.

Priority Roadmap

Near-term (current sprint)

  1. Iran-Israel full actor cluster (done)
  2. Scenario subgraph config system (done — config format)
  3. Information asymmetry implementation in core engine
  4. Escalation ladder structured output

Medium-term

  1. Ensemble forecasting (multi-run aggregation)
  2. Adaptive timestep implementation
  3. Exogenous shock injection system
  4. Second conflict scenario (South China Sea or Ukraine)
  5. Domestic policy scenario template

Long-term

  1. Panel-Of-Claude tension mapping mode
  2. AI-Agent-UN voting mode
  3. PITME facilitator/chamber mode
  4. Public API for simulation results
  5. Calibration feedback loop from accuracy grading