Literature Review

How It Works

Starting a literature review:

You: "I need to do a literature review on AI in healthcare"

Literature Review: "I'll help you conduct a systematic literature review on AI in healthcare.

1. Research Scope:
   - What's your specific focus? (diagnosis, treatment, admin, etc.)
   - Time range? (last 5 years, 10 years, etc.)
   - Any specific technologies? (LLMs, computer vision, etc.)

2. Search Strategy:
   - Keywords: "AI healthcare", "machine learning medical", etc.
   - Databases: PubMed, Google Scholar, arXiv
   - Boolean operators: ("AI" OR "machine learning") AND healthcare

3. Screening Criteria:
   - Inclusion: Peer-reviewed, English, empirical studies
   - Exclusion: Opinion pieces, non-peer-reviewed

Let's start with the search. I'll track all papers in your literature library."

Synthesizing findings:

You: "Synthesize the 10 papers I've collected"

Literature Review: "Here's a thematic synthesis of your literature:

Theme 1: Diagnostic Applications (4 papers)
- LLMs show 85-95% accuracy in diagnostic tasks
- Computer vision outperforms human radiologists in some cases
- Key limitation: Explainability and trust

Theme 2: Treatment Optimization (3 papers)
- AI improves treatment planning in oncology
- Challenges in clinical adoption and regulatory approval
- Need for human oversight

Theme 3: Administrative Efficiency (3 papers)
- Strong ROI for AI in scheduling and documentation
- Staff resistance and training needs identified

Research Gaps:
- Limited research on AI in low-resource settings
- Ethical frameworks underdeveloped
- Long-term outcomes rarely studied

Would you like me to draft the literature review section?"

Your Literature Library

Literature reviews are stored in personal-os/memory/literature-reviews.md:

# Literature Review: AI in Healthcare

## Search Strategy
- Keywords: AI healthcare, machine learning medical
- Databases: PubMed, Google Scholar, arXiv
- Date range: 2019-2024

## Papers Reviewed (10)
1. Smith et al. (2023) - LLMs in Diagnostic Radiology
2. Johnson et al. (2022) - AI in Oncology Treatment
[... 8 more papers ...]

## Thematic Synthesis
### Theme 1: Diagnostic Applications
- Papers: 4
- Key findings: 85-95% accuracy, explainability challenges
- Synthesis: [summary paragraph]

### Theme 2: Treatment Optimization
- Papers: 3
- Key findings: improved planning, adoption challenges
- Synthesis: [summary paragraph]

[... more themes ...]

## Research Gaps
1. Limited research in low-resource settings
2. Ethical frameworks underdeveloped
3. Long-term outcomes rarely studied

Example Use Cases

Academic Thesis:

You: "I need a literature review for my PhD thesis on federated learning"

Literature Review: "I'll help you build a comprehensive literature review on federated learning.

Phase 1: Search
- Core terms: federated learning, distributed machine learning
- Advanced: privacy-preserving ML, collaborative ML
- Databases: IEEE Xplore, ACM Digital Library, arXiv

Phase 2: Screening
- I'll help you apply inclusion/exclusion criteria
- Track 50-100 papers for full review

Phase 3: Synthesis
- Organize by: applications, privacy techniques, efficiency
- Identify gaps in current research

This will take 3-5 hours. Ready to start?"

Grant Proposal:

You: "Write a literature review section for my AI ethics grant"

Literature Review: "I'll create a focused literature review section.

Key Themes:
1. Current state of AI ethics research
2. Existing frameworks and limitations
3. Regulatory landscape
4. Research gaps your project addresses

Draft: [1,000-word literature review with 15 citations]

Review:
- Comprehensive coverage ✓
- Strong gap identification ✓
- Positions your project as novel ✓
- Proper academic tone ✓

Ready to submit?"

Best Practices

• Start with a clear research question
• Define inclusion/exclusion criteria early
• Track papers in a citation manager
• Use multiple databases for comprehensive search
• Document search terms and strategies
• Update literature review periodically
• Be critical of research quality
• Identify methodological limitations

Common Mistakes to Avoid

• Too broad scope (hard to manage 200+ papers)
• Not documenting search strategy (can't reproduce)
• Ignoring negative findings (bias)
• Over-relying on automated tools (manual review essential)
• Not identifying research gaps (weakens contribution)
• Poor organization (hard to synthesize)
• Missing recent papers (research evolves quickly)

Level 2: Advanced (Automated Citation Extraction and Thematic Analysis)

Available after session 22

At this level, I automate citation extraction and perform advanced thematic analysis.

Additional Capabilities

1. Automated Citation Extraction

   • Extract citations from PDF papers
   • Auto-populate citation fields (authors, title, year)
   • Detect citation chains (who cites whom)
   • Build citation networks

2. Semantic Clustering

   • Cluster papers by topic similarity
   • Identify sub-themes automatically
   • Detect research communities
   • Visualize research landscape

3. Bibliometric Analysis

   • Citation counts and impact metrics
   • Co-authorship networks
   • Keyword frequency analysis
   • Temporal trend analysis

Example: Advanced Thematic Analysis

You: "Analyze themes across 50 papers on blockchain"

Literature Review: "Advanced thematic analysis complete:

Semantic Clustering (50 papers):
Cluster 1: Scalability (18 papers)
- Keywords: sharding, layer-2, throughput
- Top cited: Vitalik Buterin, Ethereum Foundation
- Trend: Growing research (2022-2024)

Cluster 2: Privacy (12 papers)
- Keywords: zero-knowledge, anonymity, obfuscation
- Top cited: Zcash team, Protocol Labs
- Trend: Stable interest (2019-2024)

Cluster 3: Governance (10 papers)
- Keywords: DAO, voting, consensus
- Top cited: DAO research groups
- Trend: Emerging field (2021-2024)

Citation Network:
- 3 major research communities identified
- Key bridge papers connect scalability and privacy
- Foundational paper: Nakamoto (2008) cited by 92%

Research Gaps:
1. Cross-chain interoperability (emerging topic)
2. Regulatory frameworks (limited research)
3. Energy efficiency (new research direction)

Visualizations: [generates cluster graphs and citation network charts]"

Level 3: Expert (Systematic Review Protocols and Meta-Analysis)

Available after session 22

At this level, I conduct full systematic reviews and meta-analyses following PRISMA guidelines.

Additional Capabilities

1. Systematic Review Protocol

   • PRISMA-compliant protocol design
   • Protocol registration (PROSPERO)
   • Risk of bias assessment
   • Quality evaluation tools (CASP, QUADAS)

2. Meta-Analysis

   • Effect size calculation
   • Heterogeneity testing
   • Publication bias detection
   • Forest plots and meta-regression

3. Publication Strategy

   • Identify target journals
   • Write review manuscript sections
   • Create supplementary materials
   • Prepare response to reviewers

Example: Systematic Review Protocol

You: "Create a PRISMA protocol for systematic review on remote work"

Literature Review: "PRISMA 2020 protocol ready:

PROTOCOL: Remote Work Effectiveness
Registration: [PROSPERO ID]

PICO Framework:
- Population: Knowledge workers
- Intervention: Remote work arrangements
- Comparison: Traditional office work
- Outcomes: Productivity, well-being, collaboration

Search Strategy:
- Databases: PubMed, PsycINFO, Web of Science
- Timeframe: 2010-2024
- Language: English
- Study types: RCTs, cohort studies, surveys

Inclusion Criteria:
- Peer-reviewed empirical studies
- Quantitative outcome measures
- N ≥ 100 participants

Exclusion Criteria:
- Opinion pieces
- Case studies (N < 100)
- Non-English publications

Quality Assessment:
- Risk of bias: Newcastle-Ottawa Scale
- Study quality: CASP checklist
- Evidence grading: GRADE approach

Data Extraction:
- Study characteristics
- Effect sizes
- Confidence intervals
- Heterogeneity metrics

Analysis Plan:
- Random-effects meta-analysis
- Subgroup analysis (industry, tenure)
- Sensitivity analysis (study quality)
- Publication bias: funnel plot, Egger's test

Timeline: 4-6 months
Estimated papers: 80-120 for full-text review
Ready to begin systematic review?"

Integration with Other Research Skills

• citation-manager: Extract and store citations
• knowledge-graph: Build concept maps from literature
• academic-writing: Draft review sections and manuscripts
• fact-checker: Verify claims across sources
• data-analysis: Analyze quantitative findings

Progression

As you use this skill, I learn:

• Your preferred research databases
• Search terms that work best
• Themes you care about most
• How you like literature organized
• Citation formats you use most

This knowledge improves my recommendations and synthesis quality over time.37:["$","$L3e",null,{"content":"$3f","frontMatter":{"name":"literature-review","description":"Conduct systematic literature reviews, synthesize research findings, and identify research gaps"}}]