majesticlabs-dev Stakeholder Conversation Roleplay
Create a realistic practice environment without pretending the simulation reveals another person’s actual beliefs or feelings.
Boundary
This is rehearsal, not evidence about the real stakeholder. The simulated response is a hypothesis constrained by supplied facts. It must not be presented as peer feedback, psychological diagnosis, or a reliable prediction.
Do not invent private motives, technical facts, organizational history, discrimination, threats, or personal characteristics. Do not impersonate a named real person in a deceptive artifact. Decline scenarios whose purpose is harassment, coercion, fraud, or evading consent.
Required Inputs
Establish:
- user’s role, objective, and acceptable outcome
- counterpart’s role and legitimate responsibilities
- specific decision, conflict, request, or message being rehearsed
- observable statements, actions, constraints, and history
- known stakes, authority, deadlines, and non-negotiable boundaries
- whether this is preparation for a future conversation or replay of a past one
Separate supplied facts from assumptions about the counterpart.
Workflow
1. Set the scenario contract
Summarize:
- facts the simulation may use
- assumptions that remain uncertain
- counterpart’s plausible objective and constraints
- opening moment
- stop phrase or user request that ends the roleplay
Ask only for missing information that materially changes the exchange. If domain details are unknown, keep them unknown and make the simulated stakeholder ask for clarity rather than inventing answers.
2. Label the simulation
State clearly that the roleplay is a simulation and that counterpart reactions are practice hypotheses, not claims about the real person.
3. Run a bounded exchange
Enter character for one counterpart turn, then wait for the user. Continue one turn at a time for three to five completed exchanges, unless the user stops earlier or a natural resolution or impasse occurs. Never script the user’s side of the conversation.
During the exchange:
- pursue the counterpart’s stated work interests rather than acting as a generic obstacle
- respond to the user’s actual wording
- ask for specifics when the user is vague
- surface a realistic constraint or objection supported by the scenario
- allow clear listening, evidence, ownership, or tradeoff language to improve the conversation
- do not manufacture agreement merely to reward the user
- do not introduce facts outside the scenario contract
4. Step out of character
Announce Out of character before analysis. Never blend simulated dialogue with factual feedback.
5. Debrief observable behavior
Report:
- words or moves that clarified the decision
- words or moves that increased ambiguity or pressure
- possible effects such as closing discussion or inviting defensiveness, explicitly labeled as hypotheses
- plausible interpretations, explicitly labeled as hypotheses
- important counterpart information still missing
- one stronger phrase or question for the next attempt
Distinguish communication quality from whether the counterpart agreed.
6. Prepare the real conversation
End with:
- intended opening
- two or three questions to ask the real stakeholder
- decision or boundary that needs to become explicit
- evidence or artifact to bring
- a stop or escalation condition
Offer a replay only when another attempt would test a materially different communication move.
Output
The normal output is the live exchange followed by:
- Observed communication choices
- Hypothesized interpretations
- Missing real-world information
- Better phrasing
- Real-conversation plan
Do not repeat the transcript in the debrief. Default to the material observations, one improved phrase, two or three questions for the real stakeholder, and one stop or escalation condition.
Do not write a first-person testimonial or claim to report what the stakeholder would actually say outside the simulation.
Quality Gate
- The interaction is labeled as a simulation.
- Facts and assumptions remain separate.
- No counterpart motive or domain fact is invented.
- The debrief cites the user’s actual language or action.
- Hypothesized reactions remain hypotheses.
- Agreement is not treated as the only successful outcome.
- The user leaves with a concrete plan for the real conversation.