1 / 40
adityavj.com

Press 'S' to toggle speaker notes

From Literature Chaos to Research Clarity

Using Gemini Deep Research + NotebookLM for Your PhD Journey

Aditya V Jain | Product Lead at Google | IIMK 06-08

The Reading Never Stops

  • Hundreds of papers to read in your field
  • Each has its own definitions, methods, and findings
  • Need to keep track of how they all connect
  • Need to figure out where YOUR work fits
  • Tools won't solve it entirely—but they make it less painful

Two Tools, One Workflow

Discovery: Deep Research

  • Map the landscape
  • Find papers you should read
  • Understand the big debates

Analysis: NotebookLM

  • Upload the papers you've found
  • Synthesize across sources
  • Find patterns, gaps, contradictions

Session Roadmap

  1. Gemini Deep Research — 30 min (demo + hands-on)
  2. NotebookLM — 60 min (demos)
  3. Your Practice + Q&A — 15 min
  4. Test case: Personalization in Digital Marketing

Phase 1: Discovery

Using Gemini Deep Research

Autonomous Research

  • You give it a question; it creates a research plan
  • Searches multiple times across many sources
  • Synthesizes into a structured report with citations
  • Good for: Getting oriented on unfamiliar topics
  • Good for: Finding foundational and recent papers

Demo: The Mapping Prompt

Prompt 
Map the academic literature on personalization in digital marketing. I want to understand:

- Foundational theoretical frameworks
- Key findings on when personalization works vs. backfires
- Major debates and contradictions
- Research gaps
- Most influential papers to read

Focus on marketing and information systems journals.

Watch: It Creates a Plan

  • Notice: It breaks the question into sub-questions
  • It shows you exactly what it will search for
  • You can modify the plan before it runs
  • This transparency ensures it targets your specific needs

The Research Report

  • Structured Overview: A high-level summary of the field
  • Key Theories: Identified with context
  • Major Papers: Cited directly in the text
  • Debates: Surfaced contradictions in findings
  • Links: Direct access to the source material

Trust But Verify

  • Deep Research is useful but not perfect
  • Always: Click through to cited papers
  • Always: Verify papers say what's claimed
  • Check: Look for what might be missing
  • The output is a starting point, not the final answer

From Report to Reading List

  • Foundational papers to read
  • Recent work to review
  • Theories to understand
  • A sense of the major debates
  • Next Step: Download PDFs → Upload to NotebookLM

Templates for Deep Research

  • Literature mapping: "Map the academic literature on [topic]. Identify foundational papers, key debates, and recent developments."
  • Finding gaps: "What research gaps have scholars identified in the [topic] literature?"
  • Theoretical landscape: "What theoretical frameworks have been applied to [topic]?"
Activity: 10 Min

Your Turn: Map Your Literature

  1. Go to gemini.google.com (Gemini Advanced)
  2. Enter a prompt about YOUR topic
  3. Example: "Map the academic literature on [your topic]. Identify foundational papers, debates, and gaps. Focus on [your discipline]."

What Did You Find?

  • Papers you already knew?
  • Papers you didn't know existed?
  • Anything surprising in the structure?
  • Anything clearly missing?
  • Deep Research is your scout. For deep analysis, we switch tools.

Phase 2: Deep Analysis

Using NotebookLM

Your Papers, Searchable

  • Upload your PDFs/Sources
  • Ask questions about them
  • It answers using ONLY what you uploaded
  • Every answer cites the exact source passage
  • Key Difference: Regular AI knows everything but cites nothing. NotebookLM knows only your papers but cites everything.

How They Work Together

Deep Research → Finds papers

↓

You → Download PDFs

↓

NotebookLM → Deep analysis with citations

Discovery feeds into analysis.

Demo: Building a Notebook

  • Creating a new notebook
  • Uploading papers found via Deep Research
  • Papers I'm uploading:
  • Aguirre et al. (2015) — Personalization Paradox
  • White et al. (2008) — Reactance
  • Tam & Ho (2006) — Web Personalization

The Citation System

  • Level 1: AI Response — Synthesis across papers
  • Level 2: Citation Preview — Hover to see source text
  • Level 3: Deep Link — Click to jump to exact PDF passage
  • This is how you verify every claim.

Visualizing the Literature

  • The mind map shows themes across all papers
  • See how concepts connect visually
  • Click any theme → get synthesis → trace to source
  • Helps identifying common threads immediately

Better Questions

Better Answers

The Prompting Principle

Weak Input

  • "Summarize these papers"
  • Result: Generic output
  • Vague summaries

Strong Input

  • "Compare how each paper defines [X]"
  • Result: Useful output
  • Structured comparison

Live Demo: The Difference

  • First: "Summarize these papers on personalization"
  • Then: "Compare how each paper defines 'personalization.' For each: quote the definition, note how it's used, identify dimensions. Where do they differ?"
  • Watch how the tool responds to structure.

Prompt 1: Construct Clarity

Prompt 
For the term '[X]' across these papers:

1. How does each paper define it?
2. What dimensions or types appear?
3. How is it measured?
4. What similar terms might it be confused with?

Show differences with citations.

Demo: Construct Clarity

  • Running the prompt on "personalization"
  • Output shows:
  • Definitional differences between authors
  • Specific dimensions used in studies
  • Measurement approaches

Prompt 2: Gap Hunter

Prompt 
What gaps or unanswered questions do these papers mention?

Categories:
- Methods not yet tried
- Questions not yet answered
- Contexts not yet studied

For each gap, cite which paper mentioned it.

Demo: Gap Hunter

  • Running the prompt
  • Output identifies specific future research directions
  • Highlights gaps explicitly called out by authors
  • Perfect for finding "Future Research" suggestions without reading every conclusion again

Prompt 3: Contradiction Finder

Prompt 
Where do these papers contradict each other?

For each contradiction:
- What does each side say?
- Why might they disagree?
- What would resolve it?

Cite both positions.

Demo: Contradiction Finder

  • Running the prompt
  • Output reveals the "Personalization Paradox"
  • Shows when personalization helps vs. when it hurts
  • Identifies the tension between utility and privacy

Prompt 4: Theoretical Landscape

Prompt 
What theories do these papers use to explain [topic]?

For each theory:
- What does it predict?
- Where do theories agree?
- Where do they conflict?

Demo: Theoretical Landscape

  • Running the prompt
  • Identifies core theories: Privacy Calculus, Reactance Theory
  • Shows how different lenses explain the same phenomenon
  • Maps the theoretical conversation

Prompt 5: Contribution Finder

Prompt 
Based on these papers:

1. What do they agree on?
2. Where do they disagree?
3. What's still unanswered?
4. What would be valuable to add?

Cite sources.
Activity: Try It

Your Turn: NotebookLM

  1. Go to notebooklm.google.com
  2. Create a notebook and upload 2-3 papers
  3. Try one prompt:
  4. Construct Clarity (Definitions)
  5. Gap Hunter (Missing pieces)
  6. Contradiction Finder (Debates)

Common Issues

  • PDF won't upload? Try a text-based PDF, not a scanned image.
  • Output too generic? Add more structure and specific questions to your prompt.
  • No papers on laptop? Pair up or grab a few from Google Scholar now.

The Complete Workflow

  • Week 1: Deep Research → Find papers → Download
  • Week 2-3: NotebookLM → Upload → Run Prompts → Document
  • Week 4+: Write → Use NotebookLM to verify claims and pull quotes

Honest Limitations

They Don't:

  • Tell you what to research
  • Replace reading carefully
  • Make judgment calls

They Do:

  • Find patterns quickly
  • Retrieve information instantly
  • Surface connections

Best Practices

  • Verify citations: Always click through
  • Save prompts: NotebookLM doesn't save chat history permanently
  • Curate sources: Garbage in, garbage out
  • Iterate: Refine your prompts based on the outputs

Where to Go

  • Tools: gemini.google.com (Deep Research)
  • Tools: notebooklm.google.com
  • Your Handout: Includes all prompts from today
  • Guide: Troubleshooting tips included

Questions?

  • What worked for you?
  • What didn't work?
  • What would you want these tools to do for your specific topic?

Try Them This Week

Run Deep Research. Upload papers. See if they're useful.


adityavj.com

Session Material & Free AI Resources