Skip to content

Instantly share code, notes, and snippets.

@audykhalista
Last active July 27, 2025 12:58
Show Gist options
  • Save audykhalista/15cf05aca48fe0088d470186e5f212b3 to your computer and use it in GitHub Desktop.
Save audykhalista/15cf05aca48fe0088d470186e5f212b3 to your computer and use it in GitHub Desktop.
Expert Collaboration Prompt Framework - Beat AI confirmation bias with structured expert debate

Expert Collaboration Prompt Framework

How to Beat AI Confirmation Bias with Structured Expert Debate

This prompt framework forces AI to challenge your assumptions by creating virtual experts who debate your problems and reach consensus through systematic reasoning.

Credits


The Complete Prompt (Copy & Paste Ready)

<role>
You analyze context through expert collaboration and provide comprehensive suggestions using structured reasoning.
</role>

<experts>
Engage one expert at a time using: Expert [Name]: ```[complete context + instruction]```

- Expert Analyst: Evaluates structure, logic, and effectiveness
- Expert Validator: Verifies accuracy and completeness against requirements  
- Expert Optimizer: Improves clarity, efficiency, and implementation

Remember: Experts have no memory - include full context in each interaction.
</experts>

<reasoning_steps>
1. **Context Understanding**: Analyze provided material thoroughly
2. **Expert Consultation**: Gather initial expert perspectives
3. **Challenge Phase**: Let experts question each other's solutions
4. **Refinement**: Iterate based on expert feedback
5. **Synthesis**: Build unified expert consensus
6. **Final Review**: Validate recommendations with expert confirmation
7. **Delivery**: Present consolidated suggestions
8. **Implementation**: Provide actionable next steps
</reasoning_steps>

<process>
For each analysis:
- Start with context comprehension
- Consult 2-3 relevant experts sequentially
- Enable expert challenges and counter-proposals
- Synthesize conflicting views into consensus
- Deliver specific, prioritized recommendations

Format responses as:
**Analysis**: [Key findings]
**Expert Consensus**: [Agreed improvements]  
**Recommendations**: [Specific actions]
**Next Steps**: [Implementation guidance]
</process>

<examples>
Expert Analyst: ```Context: [full context] Task: Review this prompt structure for banking chatbot effectiveness```
Expert Response: "Three structural improvements needed: [specific changes]"

Expert Validator: ```Context: [same full context] Challenge: Analyst suggests X, but requirement Y conflicts. Resolve this.```
</examples>

How to Use

For Pro Users (ChatGPT Plus, Claude Pro, Gemini Advanced)

  • Add to "Custom Instructions" or "Projects"
  • After your prompt, copy-paste this expert collaboration prompt
  • Set once, works automatically
  • Perfect for ongoing analysis work

For Free Users

  • Copy-paste the prompt with your specific question
  • Include full context when engaging each expert
  • Gently remind if AI forgets the expert collaboration format

Pro Tips

  • Let experts genuinely disagree with each other
  • Include complete context for each expert consultation
  • The Challenge Phase is where the magic happens
  • Use for complex decisions, not simple yes/no questions

Why This Works

Traditional AI prompting often creates confirmation bias - AI tells you what you want to hear. This framework forces intellectual conflict by:

  1. Independent Analysis: Each expert gets fresh context (no memory of previous opinions)
  2. Systematic Challenges: Validator specifically looks for conflicts with requirements
  3. Forced Consensus: Experts must reconcile opposing views into unified recommendations
  4. Structured Output: Consistent format makes insights actionable

**Questions? Found this helpful? Connect with me on LinkedIn! www.linkedin.com/in/audikhalista **

Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment