A hypothetical brief from a national ministry, demonstrating the framing, modelling and gaming stages through a problem statement, actors, system model, interactive moves and findings.
This is a demonstration, not a client engagement. The country, the players and the brief are fictitious. The supplied demonstrator identifies just 1 of 48 parameters as sourced; the rest are expert judgement or illustrative values. Its underlying model files and test record are not published here. The numbers illustrate the model’s internal behaviour, not a real-world forecast or advice.
What combination of moves takes the players from street confrontation to a stable, lawful process, without anyone conceding the one demand that cannot be split?
Players are named by function. Each has a short menu of moves: a status quo, a cooperative move and a hard line.
| Player | Wants | Can it commit? | Moves in the game |
|---|---|---|---|
| Authorities | Order | Yes, high capacity | Status quo · restrain from day 2 · escalate with heavier force |
| Oversight body | Continuity and credibility | Medium; internally divided | Stonewall · partial disclosure · full disclosure |
| Court | Legitimacy of the process | Yes | No timeline · sets a disclosure timeline · delays |
| Organisers | Visibility and procedural wins | Low while decentralised | Continue · disciplined, stewarded protest · escalate |
| Opposition | Amplification | Medium | Baseline · amplify the street · take the legal route |
| Leaderless groups | Varied | No | Modelled in the network game, not as a lever |
Mutual cooperation is ranked best for both the authorities and the organisers, while cooperating alone is ranked worst. A player may therefore hesitate to move first when the other side’s response is uncertain.
A court-supervised timeline can make a cooperative first move less exposed, while rising costs of confrontation may make cooperation more attractive.
These are qualitative incentive comparisons. Numerical confidence thresholds require cardinal payoff assumptions and a calculation that are not published with this demonstrator.
A threshold model of who can tip whom into cooperating.
The game layers say which moves are worth testing. They have no clock. The system model below runs those moves through time.
Nine stocks, the feedback loops between them, and every player's moves as dashed lever boxes. The double bar marks a delay that matters. The supplied demonstrator reports matching Python and Vensim implementations; the comparison record is not published here.
Open the standalone causal diagram ↗
Each player's single move, combinations, and shocks laid on top of them. Incident exposure is a 60-day index; lower is calmer.
| Scenario | Incident exposure | Peak enforcement | Grievance, day 60 | Records disclosed |
|---|
"Pipeline-aware" means the authorities account for measures already in motion instead of reacting to the current reading alone.
Five players, three moves each: 243 combinations, all run through the model. Choose a move for each player and see whether the standoff resolves. The thresholds are the analyst's, set before the runs; in a real engagement the decision-maker sets them.
combinations resolve the standoff. None of them uses heavier force by the authorities. Across the 81 combinations where the authorities escalate, average incident exposure is more than double that under restraint.
Every resolving combination pairs a disclosure mover (the oversight body or the court) with a calm mover (the authorities or the organisers). Calm without disclosure leaves grievance in place; disclosure without calm arrives into a spiral.
Restraint raises peak crowd size before it lowers incidents, in 98% of 800 sensitivity draws. A leadership judging restraint by its first week would abandon the move that works.
Holding leaders for about 15 days roughly doubles incident exposure against the status quo, because the decapitation loop removes the one actor who can commit the street to calm.
Robustness: across 800 Latin hypercube draws, varying 39 parameters by ±50%, the cooperative package beats the status quo every time. The result rests most on one structural assumption: incidents rise with enforcement and fall with leadership coherence.
| Test | Work status | Reported result | Evidence or limit |
|---|
These are tests reported in the supplied demonstrator; supporting files are not published here. Agreement between implementations would check the build, not validate the model against observed behaviour.
Every engagement starts with a decision question and visible assumptions. Model access and handover materials are agreed in the project scope.