AI-Powered Social Network Analysis
Using LLMs to Explore Bluesky Networks
Workshop Philosophy: AI as your collaborative research partner
Focus on concepts, iterate with AI, build lasting analytical skills
Why AI-Assisted Analysis? 🤖
Traditional Approach
- Memorize code syntax
- Copy-paste examples
- Get stuck on technical details
- Limited to known methods
AI-Collaborative Approach
- Focus on research questions
- Iterate through conversation
- Understand concepts deeply
- Explore novel approaches
Key Insight: The goal is to become a thoughtful analyst who uses AI tools effectively, not a
programmer who memorizes syntax.
Social Networks: Core Concepts 🌐
What is a Social Network?
A mathematical representation of relationships
Components
- Nodes: People, concepts (users, hashtags)
- Edges: Relationships (mentions, co-occurrences)
- Weights: Relationship strength (frequency)
Two Network Types
- Mention Networks: Who talks to whom?
- Hashtag Networks: Which topics cluster together?
AI as Research Partner 💬
Sample AI Conversation:
You: "I have social media data about AI discussions. What should I analyze first?"
AI: "Let's start with understanding your research goals. Are you interested in:
- Influence patterns (who shapes conversations)?
- Topic evolution (how discussions change)?
- Community structure (how groups form)?"
Best Practice: Always start with your research question, not the technical implementation.
Smart Data Strategy 📊
The Three-Phase Approach
Phase 1: Explore & Sample
- Start small (100-1000 posts)
- Understand data structure with AI help
- Refine research questions
Phase 2: Scale & Refine
- Increase dataset size gradually
- A/B test different approaches
- Use AI for alternative strategies
Phase 3: Validate & Document
- Cross-reference with multiple visualizations
- AI-assisted interpretation
- Document methodology
Data Management Best Practices 📁
AI Prompt:
"Help me design a folder structure and naming convention for tracking different experiments, versions, and
findings over time."
Suggested Structure:
bluesky_ai_posts_raw.jsonl # Original filtered data
bluesky_ai_mentions_processed.json # Processed mention network
bluesky_ai_hashtags_v2.json # Hashtag network (version 2)
bluesky_ai_analysis_notes.md # Findings and insights
Key Principle: Version control everything - data, analysis, insights, and AI conversations.
Mention Networks: Social Influence 👥
Ask AI First:
"What does a mention network reveal about online communities? What patterns indicate influence, engagement, or
community structure?"
Key Concepts
- Degree Centrality: Most connections
- Betweenness: Bridges groups
- Clustering: Tight subgroups
- Density: Overall connectivity
What to Look For
- Influential users (high centrality)
- Community clusters
- Bridge accounts
- Isolated nodes
Hashtag Networks: Topic Relationships #️⃣
AI Discussion Starter:
"Explain hashtag co-occurrence analysis and what patterns I should look for in the results."
What You're Learning
- Semantic Clustering: Topics that appear together
- Bridge Topics: Hashtags connecting communities
- Temporal Patterns: How relationships change
- Cultural Insights: What associations reveal
Insight: Hashtag networks often reveal hidden thematic connections that aren't obvious from
individual posts.
Iterative Visualization with AI 🔄
The Design Thinking Process
1. Rapid Prototyping
- Create quick, basic visualizations
- Focus on structure, not aesthetics
- Use AI to identify what's working
2. Pattern Recognition
- AI helps interpret visualizations
- Generate hypotheses about patterns
- Design focused follow-up analyses
3. Refinement & Storytelling
- Craft clear narratives with AI
- Iterate on visual design
- Validate through multiple approaches
Multi-AI Strategy 🎯
Claude
- Technical explanations
- Code review & debugging
- Detailed analysis
ChatGPT
- Brainstorming
- Creative problem-solving
- Research planning
Gemini
- Data interpretation
- Pattern recognition
- Statistical analysis
Pro Tip: Use different AIs for different perspectives on the same problem.
AI-Powered Troubleshooting 🔧
Instead of memorizing errors, learn to:
1. Describe symptoms: "My visualization loads but shows no nodes"
2. Provide context: "Analyzing 500 AI posts, filtered by mentions"
3. Ask for systematic help: "Walk me through debugging each step"
4. Learn from solutions: "Why did this error occur?"
Common Issues & AI Solutions
- Too crowded: "Help me reduce complexity while preserving patterns"
- Unclear results: "Explain this pattern and suggest questions to ask"
- Potential bias: "Help identify biases in my selection methods"
From Analysis to Insight 💡
Moving Beyond Description:
"I found that AI researchers form tight clusters but rarely mention mainstream tech accounts. Help me
interpret what this means for knowledge transfer in the tech community."
AI as Story Editor
"Help me craft a compelling narrative from my network analysis that would interest both technical and
general audiences."
Remember: The best insights come from combining computational analysis with human
interpretation and domain expertise.
Key Takeaways 🎯
1. AI as Thinking Partner
Use AI for conceptual understanding, not just code generation
2. Concept-First Approach
Understand what you're measuring before measuring it
3. Iterative Methodology
Refine through conversation, document everything
4. Ethical Responsibility
Remember that data represents real people and communities
Ultimate Goal: Build analytical thinking skills that will remain valuable as tools evolve
Building Your Research Practice 🚀
Immediate Actions
- Set up AI-assisted environment
- Practice asking analytical questions
- Build first simple network
Medium-term Development
- Develop domain expertise
- Build analytical approach library
- Create reproducible workflows
Long-term Impact
- Contribute to research communities
- Develop novel approaches with AI
- Mentor others in AI-assisted methods
Remember: The skills you learn today will help you adapt to whatever new tools emerge
tomorrow.
Thank You! 🙏
Questions & Discussion
Let's explore how AI can transform your research practice
Remember: The best analysis comes from combining computational power with human curiosity
and ethical responsibility.