The Source Synthesis Role-Play Handbook (v1.0)

The Source Synthesis Role-Play Handbook is the central blog post of the Source Synthesis Role-play method.

Note: This is an updated version (1.1.1). Read about the changes at the end of the document.

I. Introduction: What is Source Synthesis?

• Brief Definition: Source Synthesis is a method for creating structured dialogues that explore complex issues by synthesizing diverse perspectives from existing source materials.
• Core Principles:
  • Using multiple, carefully selected sources to represent different viewpoints.
  • Creating personas that embody the perspectives found in the sources.
  • Using a moderator to guide a structured conversation between the personas.
  • Focusing on a specific, well-defined topic.
  • Transparency about the constructed nature of the dialogue and the “pre-editing” process.
  • Effective use of Source Synthesis relies on providing structured input to the AI, typically using Markdown.
• Benefits of Source Synthesis:
  • Illuminates the multifaceted nature of complex issues.
  • Fosters critical thinking and encourages deeper engagement with the topic.
  • Generates new insights by juxtaposing different perspectives.
  • Promotes understanding of diverse viewpoints.
  • Provides an engaging and accessible format for exploring challenging subjects.
• The diverse editorial team behind the Handbook.

II. The Source Synthesis Method: A Step-by-Step Guide

1. Topic Selection:

   • Choose a specific, well-defined topic that is complex and benefits from multiple perspectives.
   • Formulate key questions you want the role-play to address. These questions will guide your source selection and the moderator’s line of questioning.

2. Source Selection:

   • Identify diverse, credible, and relevant sources (articles, videos, podcasts, etc.).
   • Seek out a range of viewpoints: When selecting sources, strive for diversity in perspectives. Consider sources that represent mainstream views, as well as those that might be considered “fringe” or less conventional. Thoughtfully evaluate all sources for credibility and relevance, regardless of their position on the spectrum of opinion. The goal is to create a well-rounded and nuanced discussion, not to simply reinforce existing biases.
   • Use AI tools like Perplexity and Google’s Gemini for initial source discovery, but human judgment is crucial for final selection.
   • Prioritize sources that offer strong arguments, clear perspectives, and are accessible to your intended audience.
   • Prioritize sources that are available in or easily convertible to Markdown format. If sources are not in Markdown, plan to convert them as part of the preparation process.

3. Persona Creation:

   • Analyze each source in detail to identify key arguments, underlying assumptions, tone, and style.
   • Create distinct personas that embody these viewpoints. Each persona gets:
     • A descriptive name (can be the name of the source author or a title that reflects their role or perspective).
     • A defined role or perspective within the dialogue.
     • A consistent voice and tone, based on their source(s).
     • Optionally, assign a color or visual cue for easy identification.
     • Formatting (Markdown): Note how you will use Markdown to format each persona’s contributions (e.g., headings, blockquotes, different text styles). This will ensure consistency throughout the dialogue.
     • Use “virtual” if the persona is based on a real person.
   • Casting: Carefully consider which source to use for which persona.
   • Use AI tools for summarization, but human crafting is essential for developing nuanced and believable personas.
   • Embrace the “role-play” aspect: encourage those creating personas to think of it as a form of acting, where they are embodying the perspective of the source through the persona.

• Breaking the Fourth Wall: Consider creative ways to acknowledge the constructed nature of the dialogue, such as introducing a meta-commentator or incorporating information outside the scope of the original source materials.
  * Balance representation: ensure each persona has a fair voice in the dialogue, avoiding domination by any single perspective.
  * Persona Profiles: Create detailed profiles for each persona, including:
  • Name:
  • Role/Perspective:
  • Source(s):
  • Key Beliefs:
  • Motivations:
  • Fears:
  • Potential Biases:
  • Style of Communication: (e.g., formal, informal, humorous, direct, etc.)
    * The Sound of a Persona: When working with audio sources, pay close attention to:
  • Vocal Tone: (e.g., warm, authoritative, hesitant, sarcastic)
  • Delivery: (e.g., fast-paced, slow and deliberate, passionate, monotone)
  • Rhythm and Cadence:
  • Word Choice: (e.g., technical jargon, slang, complex sentences)
  • Use these observations to inform the persona’s written dialogue and, if applicable, to guide voice actors in audio adaptations.
    * Persona Design: If applicable, consider visual representations of the personas. This could involve:
  • Creating avatars or using stock images.
  • Using distinct colors or visual cues to differentiate between personas in the written or audio-visual format.
  • Drawing inspiration from character design principles in game development.

4. Dialogue Structure and Questions:

   • Structure your dialogue outline using Markdown, which will make the generation process much smoother.
   • Divide the dialogue into logical sections, each addressing a specific aspect of the topic or one of your key questions.
   • Outline the flow of the conversation, mapping out the order in which personas will speak and the points they will address in each section.
   • Develop a list of open-ended, thought-provoking moderator questions. These questions should be designed to elicit insightful responses from the personas based on their known views.
   • Format your questions using Markdown (e.g., bolding key phrases, using lists for multiple questions).
   • Include questions that address areas of agreement and disagreement among the participants.
   • Consider a “what if” scenario to explore potential future implications or alternative possibilities related to the topic.

5. Generating the Dialogue: Using Markdown for Structured Input

   • Use AI tools like ChatGPT or Gemini to generate initial drafts of the dialogue.

   • Provide all input to the AI (persona profiles, questions, source excerpts) in Markdown format.

   • Explain why Markdown is crucial: it provides structure, reduces ambiguity, helps the AI understand the relationships between different pieces of information, and leads to more accurate and relevant responses.

   • Emphasize that while some AI models might accept other formats, Markdown is the preferred and most reliable method for Source Synthesis.

   • Prompt the AI to have each persona stay “in character” and accurately reflect their viewpoint. You can also give the AI explicit instructions on tone of voice.

   • Human refinement is crucial: edit the AI-generated dialogue for clarity, coherence, accuracy, and natural flow.
   • Ensure personas remain consistent with their source-based views throughout the conversation.
   • Use transitions to connect ideas and maintain a logical flow between speakers and sections.

6. Crafting Natural Dialogue:

   • Vary sentence structure and length to mimic natural speech patterns.
   • Use contractions and colloquialisms where appropriate for the persona.
   • Incorporate pauses, interruptions, and other conversational elements to add realism.
   • Use subtext and implied meaning to create depth and nuance.
   • Edit the AI output carefully, paying attention to awkward phrasing, repetition, and inconsistencies in tone.

7. Directing the AI:

   • Provide specific instructions to the AI on the desired tone, style, and emotional expression for each persona.
   • Use examples from the source material to guide the AI’s output.
   • Experiment with different prompts and parameters to achieve the desired results.

8. The Q&A Section:

   • Include a Q&A section to allow for audience interaction and further exploration of the topic.
   • Clearly state in the introduction to the Q&A that persona responses will be based on their established viewpoints and logical extrapolations thereof.

• Consider adding an “expert” persona during the Q&A for specialized knowledge or actionable steps. This persona, based on a source with relevant expertise, can provide in-depth information without requiring the original personas to deviate from their established views.
  * Use visual or verbal cues to distinguish between “in-character” responses (directly based on source material) and “speculative” responses (logical extrapolations). For example, a different font color or a phrase like “Based on my perspective, I would imagine that…”
  * Focus on the “why” behind suggestions, not just the “what.” Encourage personas to explain the reasoning behind their answers, drawing on their underlying values and assumptions.
  * Frame questions to elicit responses rooted in the persona’s established perspective.
  * Use the Q&A to generate constructive conflict and explore differing viewpoints.

9. Review and Revise:
   • Proofread the entire dialogue carefully for any errors in grammar, spelling, or punctuation.
   • Fact-check all information presented in the dialogue against the original source materials.

• Ensure links are correct
  * Ensure that all personas remain “in character” throughout the dialogue, consistently reflecting their source-based viewpoints.
  * Seek feedback from others on the clarity, coherence, and effectiveness of the role-play.
  * Revise the dialogue as needed, making any necessary adjustments to improve its flow, accuracy, and impact.

10. Iterate and Refine Using AI:
    • Embrace the Dynamic Nature of AI: Generative AI allows for an iterative process of refinement that goes beyond simple editing.
    • Adjusting the “Dials”: Experiment with:
      • Source Input: Adding, removing, or replacing sources to shift the range of perspectives.
      • Persona Parameters: Modifying persona descriptions, tones, or even their core arguments to explore different facets of their viewpoints.
      • Moderator Questions: Rephrasing or adding new questions to steer the conversation in different directions.
      • “What-If” Scenarios: Introducing alternative scenarios to test the personas’ responses and explore new possibilities.
    • Use Markdown to make these adjustments, ensuring consistent formatting throughout the iterative process.
    • Restructuring Ideas: Use AI to:
      • Re-order dialogue sections.
      • Generate alternative dialogue paths.
      • Summarize and synthesize insights from different iterations.
    • Example: By adding a new source with a more optimistic viewpoint on AI’s impact on jobs, and then regenerating the dialogue, you might see a significant shift in the overall tone and a more balanced exploration of the issue.
    • Caution: While iterating it is important to keep the core principles and the personas intact. They should be modified, but not fundamentally changed. The overall narrative should be consistent. And, keep the Markdown formatting intact.

New Section: Input Formatting: The Importance of Markdown

Markdown is a lightweight markup language that allows you to format plain text using a simple, intuitive syntax. It’s designed to be easy to read and write, even without being rendered into its final formatted form (like HTML). Think of it as a way to add structure and meaning to your text without the complexity of HTML or the clutter of a traditional word processor.

Why is Markdown Crucial for Source Synthesis?

• Structure: Markdown provides clear structural cues (headings, lists, blockquotes) that help the AI understand the organization and hierarchy of information within your input. This is essential for the AI to correctly interpret persona profiles, moderator questions, and source excerpts.
• Clarity: Markdown removes ambiguity. Links are clearly identified, headings are distinct, and code blocks are clearly separated. This reduces the chances of the AI misinterpreting your input.
• Consistency: Using Markdown ensures consistent formatting across all parts of the input (personas, questions, sources), which helps the AI maintain coherence throughout the generated dialogue.
• Efficiency: Markdown is computationally less expensive for LLMs to process than plain text or richer formats. It’s closer to the “native” format of much of the data LLMs are trained on.
• Editability: Markdown’s plain-text nature makes it easy to edit and refine the dialogue iteratively. It’s also ideal for collaboration, as it can be easily tracked using version control systems like Git.

Basic Markdown Syntax:

• Headings: # Heading 1, ## Heading 2, ### Heading 3, etc.
• Lists: * Bullet point or - Bullet point; 1. Numbered list item
• Emphasis: *italics* or _italics_; **bold** or __bold__
• Links: [Link Text](URL)
• Blockquotes: > Quoted text
• Code: `inline code`; “` (triple backticks) for code blocks.

Using Markdown is not optional for Source Synthesis; it’s a core part of the method, ensuring the best possible results from the AI and facilitating the entire workflow.

Example of markdown:

**Persona: Scientist (Dr. Anya Sharma)**

*   **Role:** Expert in renewable energy.
*   **Key Beliefs:** Advocates for rapid transition to renewable energy sources. Concerned about climate change.
*   **Source:** [Link to Article by Dr. Sharma]

**Persona: CEO (Mr. Robert Jones)**

*   **Role:** CEO of a major oil company.
*   **Key Beliefs:** Acknowledges climate change, but emphasizes the need for a gradual transition to avoid economic disruption.
*    **Source:** [Link to Mr. Jones's speech transcript]

**Moderator Question:**

Dr. Sharma, Mr. Jones, what are your perspectives on the feasibility of transitioning to 100% renewable energy by 2050?

**Source Excerpt (Dr. Sharma):**

> "The technology exists today to transition to 100% renewable energy by 2050. The barriers are not technological; they are political and economic."

**Source Excerpt (Mr. Jones):**

> "While we are committed to investing in renewable energy, a complete transition by 2050 is unrealistic and would have severe economic consequences."
---
**Instructions to AI:**

**Generate a dialogue** between Dr. Sharma and Mr. Jones, moderated by a neutral moderator, based on their provided profiles and source excerpts.  Maintain their distinct viewpoints and styles of communication. Explore areas of agreement and disagreement.

III. The Role of “Pre-editing”

• Definition: “Pre-editing” refers to the significant shaping of the dialogue that occurs before the conversation is generated. It encompasses the choices made during topic selection, source selection, persona creation, question design, and scenario selection. Pre-editing also includes the crucial step of formatting all input materials in Markdown.
• Importance: “Pre-editing” is a defining characteristic of Source Synthesis, distinguishing it from spontaneous, unscripted dialogues. It allows the creator to carefully craft the conversation to explore specific aspects of a complex issue.
• Transparency: The constructed nature of the dialogue and the role of “pre-editing” should be made clear to the audience. This can be achieved through a clear introduction or prologue that explains the Source Synthesis method.
• Ethical Considerations:
  • Strive for accurate representation of sources, avoiding any distortion or misrepresentation of their views.
  • Be transparent about the “pre-editing” process and the choices made in shaping the dialogue.
  • Treat sources and any individuals they represent with respect, avoiding caricatures or stereotypes.
• AI-Enhanced Pre-editing:
  • Dynamic “Pre-editing”: Generative AI transforms “pre-editing” from a static, upfront process to a dynamic, iterative one. It’s no longer just about setting the stage; it’s about actively shaping and reshaping the performance in real-time.
  • Real-Time Adjustments: You can make adjustments to sources, personas, questions, and scenarios, and immediately see their impact on the dialogue.
  • Exploring the Possibility Space: AI enables you to explore a wider range of possibilities and perspectives than would be feasible with traditional methods.
  • New Creative Control: Generative AI gives a new kind of creative control. You are in control of the narrative, and can fine-tune it as you go along.
• Generating Constructive Conflict:
  • Use “pre-editing” to create productive conflict and tension within the dialogue.
  • Choose sources that represent genuinely differing viewpoints.
  • Craft questions that highlight areas of disagreement.
  • Use the moderator to encourage debate and explore different perspectives.

IV. Source Synthesis and Game Development: Learning from NPC AI

Creating a Source Synthesis role-play shares similarities with how game developers create and “train” Non-Player Characters (NPCs) using AI. Both involve using AI to generate behaviors and dialogue that feel natural and contribute to a larger narrative. Understanding this analogy can help us appreciate the importance of “pre-editing” in Source Synthesis.

• Key Parallels between “Pre-editing” and NPC Development
  • Illustrates the “Invisible” Work: Just as game developers spend hours designing and programming NPC behaviors, “pre-editing” in Source Synthesis involves a lot of “invisible” work. This includes the careful selection of sources, the crafting of personas, and the framing of questions. These choices shape the final dialogue in profound ways.
  • Highlights the Creative Control: Game developers have a high degree of creative control over their NPCs. Similarly, Source Synthesis creators have significant creative control through “pre-editing,” shaping the narrative and the perspectives presented.
  • Explains the Use of AI as a Tool: Game developers use AI as a tool to create more dynamic and responsive NPCs. Similarly, Source Synthesis uses AI as a tool to generate and refine dialogue, enabling a more iterative and exploratory approach to crafting the conversation.
  • Emphasizes the Iterative Process: Game development is an iterative process. “Pre-editing” in Source Synthesis is also an iterative process, where creators can experiment with different sources, personas, and questions, using AI to generate and evaluate different versions of the dialogue.
  • Introduces the “Gamification” Potential:
    By incorporating elements from dynamic dialogue systems, such as allowing readers to choose between different moderator questions or offering alternative dialogue paths, Source Synthesis could become even more engaging and interactive.

Read more about NPC:s on Wikipedia.

V. Adapting Source Synthesis for Audio – Optional

• Voice Acting:
  • Use voice actors to portray the personas, bringing the dialogue to life.
  • Carefully select actors whose voices match the tone and style of the personas.
  • Direct the actors to convey the appropriate emotions and nuances in their delivery.
  • Utilize the “expert” persona in the Q&A for specialized knowledge or actionable steps.
• Sound Design:
  • Incorporate sound effects and music to enhance the narrative and create a more immersive experience.
  • Use sound to differentiate between personas, locations, or sections of the dialogue.
• Editing for Spoken Dialogue:
  • Edit the written dialogue to make it more natural and conversational when spoken aloud.
  • Pay attention to pacing, rhythm, and flow.
  • Use pauses and other nonverbal cues to add depth and meaning.
• Vocal Cues and “Speculative” Markers:
  • Use distinct vocal cues or sound effects to indicate when a persona is moving beyond their source-based statements into more speculative territory.
• Directing the AI:
  • Provide specific instructions to the AI on the desired tone, style, and emotional expression for each persona.
  • Use examples from the source material to guide the AI’s output.
  • Experiment with different prompts and parameters to achieve the desired results.

VI. Enhancing Interactivity – Optional

• Reader Polls: Allow readers to vote on which persona they agree with most, or on which argument they find most persuasive.
• Branching Dialogue Paths: Create alternative dialogue paths based on different moderator questions or persona responses. Allow readers to choose which path to follow.
• “Choose Your Own Adventure” Style: Structure the role-play like a “choose your own adventure” game, where reader choices determine the direction of the conversation.
• Interactive Q&A: Allow readers to submit questions to the personas in real-time or through a comment section.
• Gamification: Introduce game-like elements, such as points, badges, or leaderboards, to incentivize participation and engagement.

VII. Using NotebookLM for Source Synthesis Workflows (Optional)

NotebookLM, Google’s AI-powered research tool, can be a valuable asset in creating Source Synthesis role-plays. Here’s how:

1. Source Gathering and Organization: NotebookLM allows you to upload various source materials (documents, web pages, transcripts) into a single “notebook.” This centralizes your research and makes it easily accessible.
2. Source Analysis: Use NotebookLM’s features (summarization, Q&A, source guides) to analyze your sources and identify key arguments, perspectives, and potential points of conflict for your personas.
3. Persona Development: Create notes within NotebookLM to develop your persona profiles. You can even use NotebookLM to “interview” your sources and extract information relevant to each persona.
4. Dialogue Generation (Initial Draft): While NotebookLM isn’t designed for generating full dialogues in the Source Synthesis style, you can use it to generate initial drafts of individual persona responses or sections of the dialogue. You would then need to significantly refine and structure this output (using Markdown!) to create the final role-play.
5. Source Tracking: NotebookLM will help you to keep the sources and citations intact.
6. Iterative Refinement: Use NotebookLM’s features to explore different “what if” scenarios and refine your dialogue. For example, you could add new sources to the notebook and see how they impact the AI’s responses.

Important Note: While NotebookLM can be a helpful tool, it’s not a replacement for the core Source Synthesis process. Human judgment, careful planning, and meticulous editing (using Markdown!) are still essential. NotebookLM is best viewed as an aid to the process, not an automated solution.

VIII. Example: Brief Outline of a Source Synthesis Role-Play

• Topic: The Impact of Social Media on Mental Health
• Sources:
  • Article from a psychologist on the negative effects of social media on self-esteem.
  • Blog post from a social media influencer promoting the positive aspects of online communities.
  • Research paper on the correlation between social media use and anxiety in teenagers.
• Personas:
  • The Psychologist (Dr. Smith): Concerned about the negative impacts of social media, particularly on young people.
  • The Influencer (Alex): Believes social media can be a positive force for connection and self-expression.
  • The Researcher (Sam): Presents data and findings on the complex relationship between social media use and mental well-being.
• Sample Dialogue:

**Persona: Scientist (Dr. Anya Sharma)**

*   **Role:** Expert in renewable energy.
*   **Key Beliefs:** Advocates for rapid transition to renewable energy sources. Concerned about climate change.
*   **Source:** [Link to Article by Dr. Sharma]

**Persona: CEO (Mr. Robert Jones)**

*   **Role:** CEO of a major oil company.
*   **Key Beliefs:** Acknowledges climate change, but emphasizes the need for a gradual transition to avoid economic disruption.
*    **Source:** [Link to Mr. Jones's speech transcript]

**Moderator Question:**

Dr. Sharma, Mr. Jones, what are your perspectives on the feasibility of transitioning to 100% renewable energy by 2050?

**Source Excerpt (Dr. Sharma):**

> "The technology exists today to transition to 100% renewable energy by 2050. The barriers are not technological; they are political and economic."

**Source Excerpt (Mr. Jones):**

> "While we are committed to investing in renewable energy, a complete transition by 2050 is unrealistic and would have severe economic consequences."
---
**Instructions to AI:**

**Generate a dialogue** between Dr. Sharma and Mr. Jones, moderated by a neutral moderator, based on their provided profiles and source excerpts.  Maintain their distinct viewpoints and styles of communication. Explore areas of agreement and disagreement.

IX. Conclusion

Source Synthesis offers a powerful and adaptable method for exploring complex issues in a nuanced and engaging way. By synthesizing diverse perspectives from existing sources, creating dialogue-based role-plays, and embracing the iterative potential of AI, we can foster deeper understanding, critical thinking, and new insights. Experiment with the method, adapt it to your needs, and contribute to more informed and productive dialogues about the challenges and opportunities of our time.

---

Version History:

• v1.1.1 (Current): Clarified instructions for handling Markdown code examples within the handbook to prevent formatting issues.
• v1.1:
  • Core Method Updates:
    • Integrated Markdown as a core requirement for input formatting.
    • Added guidance on using “virtual” prefixes for personas based on real individuals.
    • Added guidance on “breaking the fourth wall” within role-plays.
  • New Sections:
    • Added “Input Formatting: The Importance of Markdown” section.
    • Added “Using NotebookLM for Source Synthesis Workflows” section.
  • Handbook Revisions:
    • Updated the Step-by-Step Guide to reflect the importance of Markdown.
    • Clarified persona representation options.
• v1.0 (Original): Initial release of the Source Synthesis Role-Play Handbook.