Deep Dive Flow: Complete Conceptual and Developer Explanation

Conceptual Framework

What is a Deep Dive?

A Deep Dive is an extension of completed Movement 1 reflections that provides another reflection round for students to re-examine their research topic through specialized conceptual lenses. Rather than starting fresh, deep dives build upon the foundation of a student's completed 4-round Socratic questioning journey by applying focused analytical frameworks.

Pedagogical Philosophy

Deep dives operate on the principle that inquiry can deepen through perspective shifts rather than more information accumulation. Each deep dive category represents a different way of seeing the same research territory:

• Spark of Inquiry: Personal connection and origin stories
• Inquiry as Story: Narrative structure and intellectual tension
• Stakes and Significance: Public relevance and broader implications
• Puzzles and Unknowns: Research process and uncertainty navigation
• Listeners and Lens: Audience awareness and rhetorical strategy
• Production Lab: Invite students to reflect on the act of making as part of their inquiry

The Relationship to Movement 1

Trying out a deep dive requires a completed Movement 1 session (with synthesis) as their foundation, but they aim to generate entirely new insights rather than simply reviewing previous work. The original topic becomes the stable anchor while the deep dive category provides the analytical lens.

From Podcast to Multimodal Communication

While many of the initial subcategory examples are framed for an audio project like a podcast, the Deep Dive architecture is fundamentally media-agnostic. By adapting system instructions, it, like RTS itself, can be a tool for any form of inquiry-based communication. The core principles of shifting perspective, deepening analysis, and structuring a narrative apply equally to:

• Written Forms: Research essays, reflective papers, articles.
• Visual Media: Video scripts, presentation storyboards, data visualizations.
• Oral Presentations: Conference talks, formal presentations, project pitches.

The prompt guiding each subcategory can be easily adapted. For example, a subcategory like "Translating Stakes for Audio" could become "Visualizing Stakes for a Presentation" or "Structuring Stakes in an Argumentative Essay." The goal remains the same: to make the significance of the inquiry clear to a specific audience through a chosen medium.

Movement 1 Extension: The Deep Dive Architecture

Category-Based Recursive Inquiry

After completing initial Movement 1 prompts, students may choose to go deeper into another analytically and rhetorically-focused framing. This launches a 4-round Deep Dive, powered by structured generative AI question generation and reflection capture.

Each Deep Dive uses these database tables to write and read:

• deep_dive_followup_questions_log (for AI-generated questions)
• deep_dive_reflection_rounds (for user responses and metadata)
• deep_dive_synthesis (for AI summary)

Each Deep Dive:

• Runs for another 4 reflection rounds
• Each round includes:
  • Another interpretive summary
  • 7 subcategory-based questions for each deep dive (see below)
  • Student chooses and responds
• Culminates in a category-specific synthesis

Deep Dive Categories and Their Subcategories

Note: The following categories and subcategories provide concrete examples tailored for an audio-based project. This framework is highly adaptable; the core purpose of each question can be modified for different communication modes (written, visual, oral) by adjusting the prompt and handlers.

🔥 Spark of Inquiry

Example JSON - same for each subcategory:

// Schema for Spark of Inquiry structured questions
const SparkQuestionsSchema = {
  type: Type.OBJECT,
  required: ["Interpretive Summary", "Follow-up Questions"],
  properties: {
    "Interpretive Summary": { 
      type: Type.STRING,
      description: "4-5 sentence summary about their spark of inquiry journey" 
    },
    "Follow-up Questions": {
      type: Type.OBJECT,
      required: [
        "The Origin Scene", 
        "The Emotional Core", 
        "The Naive Question", 
        "The Starting Assumption", 
        "The Personal Stake", 
        "The Disciplinary Bridge", 
        "Narrating the Spark"
      ],
      properties: {
        "The Origin Scene": { type: Type.STRING },
        "The Emotional Core": { type: Type.STRING },
        "The Naive Question": { type: Type.STRING },
        "The Starting Assumption": { type: Type.STRING },
        "The Personal Stake": { type: Type.STRING },
        "The Disciplinary Bridge": { type: Type.STRING },
        "Narrating the Spark": { type: Type.STRING }
      },
    },
  },
};

Subcategory	Purpose
The Origin Scene	Reconstruct the vivid moment their interest began
The Emotional Core	Identify the affective engine of their curiosity
The Naive Question	Surface the first unfiltered question they had
The Starting Assumption	Articulate early beliefs they brought into the topic
The Personal Stake	Link inquiry to their identity, background, or values
The Disciplinary Bridge	Reframe the spark using a course concept
Narrating the Spark	Strategize how to evoke this moment in audio

📚 Inquiry as Story

Subcategory	Purpose
The Diagnostic Puzzle	Define a core disciplinary tension or definitional conflict
The Narrative Pivot	Identify a before/after in the scholarly conversation
The Unresolved Core	Position their project in relation to a major unanswered question
Competing Hypotheses as Characters	Narrativize academic debates using scholars as characters
Challenging the Common Assumption	Contrast academic research with public misconceptions
The Paradox of Intervention	Explore tensions between research and real-world application
Mapping Complexity to Audio	Choose a guiding concept to structure complexity for listeners

🌎 Stakes and Significance

Subcategory	Purpose
The "So What?" Question	Articulate the real-world relevance in plain language
The Intended Audience	Define a specific, relevant listening group
The Desired Shift	Clarify the intended belief/thought/action change in the audience
The Public Conversation	Situate research within a current debate using course materials
The Ethical Obligation	Reflect on accountability and ethics in storytelling
The Counter-Argument	Consider stakes from a conflicting perspective
Translating Stakes for Audio	Strategize how to make the stakes audible in the podcast

🧩 Puzzles and Unknowns

Subcategory	Purpose
The Unexpected Discovery	Identify a surprising finding and how it challenged assumptions
The Methodological Gap	Highlight what remains unknowable or unanswerable
The Missing Voice	Surface perspectives or communities absent in the literature
The Productive Confusion	Name a difficult concept they’re still grappling with
The "I Used to Think..." Moment	Reflect on a personal shift in understanding
The Next Question	Define the evolved research question post-inquiry
Narrating the "How"	Plan how to narrate their research process in audio form

🎧 Listeners and Lens

Subcategory	Purpose
The Ideal Listener	Describe the audience’s assumptions before hitting play
The Core Takeaway	Name the one sentence they want to linger in memory
The Desired Feeling	Define emotional/intellectual resonance goal
The Call to Inquiry	Prompt the listener to ask a new question
The Lens of the Discipline	Offer a key concept to shift how the listener sees the world
The Memorable Moment	Identify the emotional or rhetorical peak of the episode
The Opening Hook	Craft the first 30 seconds to earn listener attention

🎙️ Production Lab

Subcategory	Purpose
The Narrator’s Voice	Define the persona and vocal style of the storyteller, linking tone to Spark of Inquiry or Audience Awareness
The Sonic Mood Board	Imagine the overall sound and feel of the episode; use adjectives to shape audience resonance and connect to Stakes
Pacing and Silence	Explore rhythm and pauses as rhetorical tools to highlight key stakes or invite reflection
Showing vs. Telling in Audio	Translate one concept into sound (archival clip, effect, ambient audio) instead of pure narration, deepening inquiry through form
The Key Moment	Identify the most important sentence or idea and experiment with production techniques to make it memorable
The Cold Open	Craft the first 15 seconds to set tone, intrigue, and immediately signal why the story matters
The Simplest First Step	Normalize low-stakes making (phone recording, one music clip, quick edit) as a way to collapse thinking/doing early

Technical Architecture

Core Data Flow Pattern

 Completed Session (first topic flow) → Deep Dive Selection → Category-Specific 4-Round Flow → Synthesis

Database Architecture

The system uses session-based continuity with category-based differentiation:

-- Original session (prerequisite)
rts_sessions (session_id, topic_text, user_id, course_id)
movement_synthesis (session_id, movement_number=1) -- Completion marker

-- Deep dive data (new records, linked to original session)
deep_dive_reflection_rounds (
session_id,           -- Links to original session
category_selected,    -- e.g., "Spark Of Inquiry" 
round_number,         -- 1-4 cycle per category
user_reflection,      -- Student's response
interpretive_summary  -- AI's contextual analysis
)

Handler Trio Pattern

Each deep dive category follows an identical 3-handler pattern:

1. Generate Handler: generate[Category]FollowupQuestions.ts

• Receives: session_id + round_number
• Queries: Original topic + previous deep dive reflections (if round > 1)
• AI Process: Generates interpretive summary + 7 category-specific questions
• Returns: Questions grouped by subcategory + AI metadata

2. Save Handler: save[Category]FollowupRound.ts

• Receives: Reflection data + selected question + AI metadata
• Validates: Session ownership + data integrity
• Stores: Complete round record with full metadata tracking

3. Synthesis Handler: generate[Category]Synthesis.ts

• Triggers: After 4 completed rounds
• Process: Analyzes complete deep dive journey
• Output: Mentor-style synthesis focused on category lens

The Core Pattern: A Recipe for Analytical & Rhetorical Riffs

The true power of this architecture lies in the combination of the Handler Trio, a targeted AI prompt, and the flexible database schema. This trio forms a repeatable recipe for crafting an almost infinite range of analytical and rhetorical "riffs" on a student's core research.

Because the pedagogical logic is encapsulated within the AI prompt—not hard-coded into the application—creating a new Deep Dive category is a content and configuration task, not a major engineering effort. This model allows for huge flexibility and expansion, enabling instructors or curriculum designers to rapidly develop new lenses focused on:

• Specific Methodologies: E.g., a "Historical Context" Deep Dive, a "Quantitative Reasoning" Deep Dive.
• Rhetorical Strategies: E.g., a "Pathos & Persuasion" Deep Dive, an "Audience Analysis" Deep Dive.
• Disciplinary Lenses: E.g., a "Sociological Imagination" Deep Dive, an "Ethical Frameworks" Deep Dive.

⠀This makes the system a dynamic platform for pedagogical innovation rather than a static tool.

UI Architecture

Single Page Component Pattern

Each category uses one comprehensive React component that handles:

• State Management: Round progression, question selection, reflection input
• API Orchestration: Generate → Select → Reflect → Save → Repeat
• Metadata Display: AI transparency with token tracking and thought summaries
• Export Functionality: Complete journey documentation

Key UI Features

Question Randomization: Subcategories display in random order to prevent habituation patterns AI “Transparency”: Real-time display of:

• Token usage per round and cumulatively
• AI “thinking” processes for educational transparency
• Model metadata and computational costs

⠀Journey Tracking: Visual progress indicators showing:

• Current round (1-4)
• Completed rounds with metadata
• Export readiness status

Scaling Implementation Strategy

Cloning Process for New Categories

Step 1: Handler Creation

// Copy and rename the trio:
generateSparkFollowupQuestions.ts → generateInquiryAsStoryFollowupQuestions.ts
saveSparkFollowupRound.ts → saveInquiryAsStoryFollowupRound.ts  
generateSparkSynthesis.ts → generateInquiryAsStorySynthesis.ts

Step 2: Prompt Adaptation

• Update system instructions with category-specific pedagogical goals - test via model’s prompt management tools, like Vertex AI (Google)
• Modify question schemas to reflect subcategory structure
• JSON schema for structuring the output (useful for display and database ledger)
• Adjust synthesis prompts for category lens focus

⠀Step 3: UI Component

// Copy and adapt:
spark-of-inquiry.tsx → inquiry-as-story.tsx
// Update: category constants, API endpoints, route handlers

Step 4: Routing Integration

• Add route: /pages/movement1/deep-dive/inquiry-as-story.tsx
• Update deep-dives dashboard to enable new category

Database Considerations

No schema changes required - the same tables support all categories through the category_selected field. Categories are differentiated by string values:

"Spark Of Inquiry" 
"Inquiry As Story"
"Stakes And Significance"
"Puzzles And Unknowns"  
"Listeners And Lens"
"Production Lab"

Developer Implementation Guide

Adding a New Category: "Inquiry as Story"

1. Create Handlers

• Clone spark handlers, update imports and category constants
• Develop new prompts and replace with inquiry-as-story specific content
• Update API endpoints in handler file names

⠀2. Create UI Component

• Copy spark-of-inquiry.tsx to inquiry-as-story.tsx
• Update category constant and page title
• Change API endpoint calls to new handlers

⠀3. Update Dashboard

• Enable "inquiry-as-story" category in deep-dives page

⠀4. Test End-to-End

• Verify question generation with category-specific content
• Test 4-round completion flow
• Confirm synthesis generation and export functionality

Quality Assurance Checklist

Data Integrity:

• Session ID properly links to original completed session
• Category selection correctly stored and retrieved
• Round progression follows 1→2→3→4 sequence
• AI metadata captured at each step

⠀User Experience:

• Question randomization working per session
• Loading states appropriate for AI generation
• Error handling for network/API issues
• Export functionality generates complete reports

⠀AI Integration:

• Prompts generate category-appropriate questions
• Token usage tracked and displayed educationally
• Thought summaries provide meaningful “transparency” (as much as possible)
• Synthesis reflects category-specific analytical lens

Some System Benefits

For Students

• Analytical Flexibility: Same topic, multiple perspectives
• Depth Over Breadth: Focused exploration rather than surface coverage
• Meta-Cognitive Awareness: Understanding different ways of seeing research problems
• Portfolio Generation: Exportable documentation of “thinking” processes
• Expressive Versatility: Adapts the core inquiry process to various final formats—from podcasts to essays—strengthening transferable communication skills.
• Post Prompt Prompting Exercises: Students analyze AI output and experiment with prompting strategies based on analytical directions they want to explore - They create their own deep dives.

⠀For Instructors

• Modular Assignment Design: Mix and match categories based on course goals
• Media-Agnostic Scaffolding: Tailor assignments for different final outputs (written, audio, visual) using the same consistent inquiry tool.
• Assessment Granularity: Track student engagement with specific analytical frameworks
• Scaffolding Control: Students must complete foundational work before accessing extensions

⠀For Researchers

• Pedagogical Data: How different analytical lenses affect student reasoning
• AI Interaction Patterns: Token usage and engagement across categories
• Longitudinal Analysis: How perspective-taking develops over time

⠀ This architecture creates a scalable, consistent, and educationally coherent system where adding new deep dive categories requires minimal technical overhead while maintaining the pedagogical integrity of focused analytical exploration.

#rts/dev/deepdives