A transparent account of what the simulator does and exactly how AI agents are used to produce each negotiation — including what they do not do.
The Diplomatic Simulator replays multi-party diplomatic negotiations. Each delegation in a scenario is played by an AI agent that argues from that government's real negotiating brief. The output — a round-by-round transcript, a tactic analysis, and a convener report — is generated end-to-end by large language model agents. This page documents that process so the results can be read with the right amount of trust.
Every statement, score, and summary on this site is generated by an AI language model role-playing a government. It is notional and educational — not a forecast, not intelligence analysis, and not the position of any real government or official. Scenarios are adapted from academic crisis-negotiation exercises (see Provenance). Treat the simulator as a structured thought experiment, not a source of fact about world events.
Each scenario moves through six stages. Stages 1–5 are the same reusable toolchain (sim/); only the source documents change.
build_session.py, inject.py, build_report.py) stitches the agent outputs into a session record, normalises the self-tagged tactics, and renders the interactive table and this report page. No model is involved here.The core design choice is information isolation. Real diplomacy runs on private information; so does this simulator. When a delegation agent writes its statement for a round, it is given only:
The agent is instructed to stay in role, ground its claims in its brief, protect its secret bottom lines, and — for the analysis layer — self-tag the negotiating tactics it used from a fixed vocabulary (e.g. anchoringconditional-offerred-line-signaledissue-linkagecoalition-buildingappeal-to-law). It writes one statement and nothing else; it does not see other delegations' private reasoning, only their public words.
| Round | Phase | What each delegation does |
|---|---|---|
| 1 | Opening plenary | Headline priorities, framing, opening anchors, coalition signals |
| 2 | Positioning Arctic only / Bargaining | React to rivals by name; defend challenged red lines; probe for trades |
| 3 | Bargaining & coalitions | Table concrete conditional offers and packages; form coalitions; use linkage / side-payments |
| 4 | Closing plenary | Assess progress; state supported understandings and remaining gaps; propose a way forward |
The flagship Arctic scenario runs four rounds; the others run three (opening → bargaining → closing).
After the talks conclude, a single analyst agent reads the whole transcript alongside every delegation's profile and judges how well each side achieved its stated objectives. It outputs:
These numbers are an AI's subjective judgement of an AI-written transcript. They are internally reasoned and differentiated across parties, but they are not measured outcomes and carry no ground truth. Read satisfaction as "how the analyst rated the argument on the page," not as a real assessment of national interest.
What a Monte Carlo simulation is. A single negotiation run tells you one story. But the result depends on chance and on conditions no one controls — a sudden crisis, the mood in the room, who has leverage that week. A Monte Carlo simulation answers a different question: if we ran this many times under many different conditions, what range of outcomes would we see, and how often? The technique — named after the casino — is simple: (1) identify the uncertain inputs, (2) draw random values for them, (3) run the model once for that draw, and (4) repeat many times and look at the distribution of results rather than any single run. Where a point estimate says "the deal fails," a Monte Carlo says "the deal fails in 2 of 8 runs, usually when a hardline shock coincides with a confrontational mood" — which is far more useful for judging how robust or fragile an outcome is.
For each scenario we re-simulate the negotiation many times, and each trial randomizes four exogenous conditions:
An agent simulates the negotiation's outcome under that specific draw — who concedes, which coalitions form, whose red lines hold, and whether a deal is reached — and returns a structured result (deal type, per-party satisfaction, goal status, red-line holds). Across the trials we aggregate the distribution: the mix of deal types, each delegation's satisfaction spread (mean and min–max range), how often each red line survives, and which coalitions recur. A wide spread means the outcome is condition-sensitive; a tight one means it is robust. Results are on the Monte Carlo page.
Reduced-form, and still not a forecast. To keep this affordable, each trial is a single-agent outcome simulation conditioned on the random draw — not a full re-run of the multi-round negotiation. So the Monte Carlo samples the model's distribution over plausible outcomes, not an empirical distribution of real events. The randomized conditions are illustrative, not calibrated probabilities. Read it as "how sensitive does the model think this outcome is to shocks and mood," never as odds on real diplomacy. This run used 8 trials per scenario (32 total).
Model. Every agent in the main negotiations — profile extraction, public brief, delegation role-play, and analysis — is Claude Opus 4.8 (Anthropic), run through Claude Code's agent framework. The lighter-weight Monte Carlo trials use Claude Sonnet to keep that batch inexpensive. There is no fine-tuning and no other provider.
Orchestration. The Arctic scenario was run with direct parallel agent calls. The other three were run as a background workflow that fans out one agent per delegation per round and sequences the phases (setup → profiles → rounds → analysis). Roughly 130 agent tasks produced the four simulations.
The first orchestration run hit a transient network outage that failed about half its agents mid-negotiation. A targeted re-run then completed the missing rounds and analyses cleanly. No gaps were papered over — every published statement is a real agent output, and the run artifacts (profiles, transcripts, analyses) are committed under sim/scenarios/ for inspection.
Reproducibility. The toolchain and every intermediate artifact are in the repository. Given the same source PDFs, the pipeline (sim/extract_pdf.py → profiles → negotiation → sim/build_session.py → sim/inject.py → sim/build_report.py) reproduces a comparable session; because language models are stochastic, exact wording will differ between runs.
The working paper, reference architecture, and convening behind this project were scoped to respect Microsoft's IP, public-disclosure, and conflict-of-commitment obligations. About 70% of the working paper is publishable as-is from public Microsoft references, EEA / x402 Foundation materials, and existing academic literature; material touching Microsoft Traceability Program internals is treated synthetically or routed through CELA and Microsoft Communications review on a published timeline. The Diplomacy Table reference architecture is built on the open-source DTFS / TwinForge stack and is publishable in full.
Scenario packs are adapted from U.S. Army War College International Strategic Crisis Negotiation Exercise (ISCNE) and related academic crisis-negotiation materials — notional exercise content for educational purposes. The simulator, toolchain, and generated negotiations were built under Ethical Tech CoLab. Negotiation content is AI-generated with Claude Opus 4.8.