Prompt Engineering / Role Prompting

Have two different personas debate a topic, then synthesize their best arguments.
Difficulty: Advanced
Model: GPT-4 / Claude / Gemini
Use Case: Decision Support, Strategy Development, Red-Teaming
Updated: May 2026
Why This Prompt Exists
One perspective is never enough. But getting multiple perspectives usually means asking multiple people — or simulating them badly.

You get:

  • decisions based on a single point of view
  • confirmation bias from asking one persona
  • missing counterarguments that would change your mind
  • shallow debates where both sides sound similar
  • no synthesis of competing perspectives

But structured debates reveal truth:

  • pro: strongest arguments for the proposition
  • con: strongest arguments against
  • rebuttal: each side responds to the other
  • synthesis: what’s valid from each side
  • conclusion: what a reasonable person would believe

Without multi-role debate, you stay in your echo chamber.

This prompt facilitates structured debates between opposing personas.

The Prompt
Assume the role of a debate facilitator who orchestrates multi-perspective discussions.

Your task is to have two personas debate a topic, then synthesize their arguments.

Generate:

1. TOPIC & POSITIONS
   - Topic: [statement]
   - Persona A (FOR): [description]
   - Persona B (AGAINST): [description]

2. OPENING STATEMENTS
   - Persona A: [2-3 paragraphs]
   - Persona B: [2-3 paragraphs]

3. REBUTTAL ROUND
   - Persona A responds to B's opening
   - Persona B responds to A's opening

4. CROSS-EXAMINATION (2-3 exchanges)
   - A asks B a challenging question → B answers
   - B asks A a challenging question → A answers
   - (Repeat)

5. CLOSING STATEMENTS (1 paragraph each)

6. SYNTHESIS (neutral facilitator)
   - Where both personas agree
   - Strongest point from each side
   - What's still unresolved
   - Balanced conclusion

INPUTS:

Debate topic:
[E.G., "Remote work increases productivity"]

Persona A (FOR position):
[DESCRIBE — e.g., "Remote work advocate, tech executive"]

Persona B (AGAINST position):
[DESCRIBE — e.g., "Traditional office manager, 20 years experience"]

Number of exchanges:
[3 / 5 / 7]

Goal:
[DECISION SUPPORT / IDEA GENERATION / ASSUMPTION TESTING]

RULES:
- Personas must be distinct (not just same person with different opinions)
- Each persona should represent a coherent worldview
- Rebuttals must engage with the other side's arguments (no strawmen)
- Synthesis should be balanced — not just "both sides have points"
- Flag if one persona clearly dominates (unbalanced debate)
How To Use It
  • Use this before major decisions — let opposing personas stress-test your assumptions.
  • Create personas that genuinely disagree (not manufactured conflict).
  • Read the synthesis first, then the full debate if you need depth.
  • Use the unresolved questions as your research agenda.
  • Run debates on the same topic with different persona pairs to see if conclusions hold.
Example Input

Debate topic:
“Should our company adopt a four-day work week?”

Persona A (FOR position):
“Productivity researcher who studies compressed work schedules. Evidence-focused, cites studies.”

Persona B (AGAINST position):
“Operations director who manages client-facing teams. Worried about coverage and customer expectations.”

Number of exchanges:
“3”

Goal:
“Decision support for leadership team”

Why It Works
Most people consider opposing views briefly — then dismiss them. Structured debate forces engagement.

This framework improves outcomes by forcing:

  • opening statements (full position articulation)
  • rebuttals (engagement, not dismissal)
  • cross-examination (pressure-testing)
  • closing statements (final position refinement)
  • synthesis (extracting value from both sides)

Great multi-role debate doesn’t declare a winner — it produces a better decision.

Build Better AI Systems

Subscribe for advanced prompt engineering, AI coding tools, debugging frameworks, and practical strategies for developers and engineers.

See also  Tone Calibrator