MicroCrowd - Enhanced AI Focus Groups

Transform CSV data into realistic personas for natural, authentic focus group discussions with advanced conversation intelligence.

🚀 Overview

MicroCrowd bridges the gap between static data and dynamic human insights by transforming consumer datasets into interactive personas for experimental research. Traditional market research methods are limited by cost, time, and participant availability, while generic AI lacks the nuanced diversity of real human perspectives. MicroCrowd enables researchers and strategists to conduct focus group simulations with authentic personality-driven responses, making advanced research methodologies accessible for experimentation across diverse scenarios—from product testing and user experience research to academic studies and strategic planning.

This a framework working model and not scaled for mass use.

🎬 Demo

Check out MicroCrowd in action in demo video

https://youtube.com/shorts/WjuCvi4yywE

🚀 Quick Start

Want to see the enhanced conversation engine in action?

Clone and install: git clone https://github.com/vtmade/MicroCrowd.git && cd MicroCrowd && npm install
Add your OpenAI API key to .env file
Run: npm run dev
Upload the included sample_personas.csv
Start a focus group on a controversial topic like "climate change" or "gun laws"
Watch authentic disagreements unfold! 🎭

✨ Key Features

🧠 Enhanced Conversation Engine (NEW!)

🤝 Realistic Disagreements: Personas authentically disagree based on demographics and backgrounds
🎯 Smart Reference Validation: Only references actual statements with exact quotes
🗣️ Persona-Specific Speech Patterns: 6 distinct speaking styles based on personality and background
📊 Dynamic Behavior Tracking: Real-time emotional states affect participation and responses
⚡ Intelligent Interruptions: Natural conversation flow with personality-driven interruptions

🔄 Core Functionality

📊 CSV-to-Persona Transformation: Convert any CSV dataset into rich, multi-dimensional personas
🎭 Authentic Personalities: Each persona has unique traits, behaviors, and response patterns
💬 Natural Conversations: Realistic focus group dynamics with varied participation levels
🎯 Professional Moderation: AI moderator follows your discussion guide systematically
⏸️ Real-time Control: Pause, resume, or end sessions at any time
📈 Export Transcripts: Download complete conversation logs for analysis

🎯 Use Cases

Market Research

Product Testing: Understand how different demographics react to new products
Brand Perception: Explore brand associations across diverse consumer segments
Feature Prioritization: Identify which features resonate with different user types
Pricing Strategy: Test price sensitivity across various customer personas

Academic Research

Social Studies: Simulate group discussions on social issues or policies
Psychology Research: Study group dynamics and personality interactions
Educational Research: Test teaching methods or educational content
Behavioral Studies: Explore decision-making patterns in group settings

UX/Design Research

User Journey Mapping: Understand how different users navigate your product
Interface Testing: Get diverse perspectives on design choices
Accessibility Research: Include personas with varying abilities and tech comfort
Content Strategy: Test messaging with different demographic groups

Innovation & Strategy

Idea Validation: Test concepts with diverse stakeholder perspectives
Scenario Planning: Explore how different groups might react to changes
Risk Assessment: Understand potential concerns from various viewpoints
Stakeholder Alignment: Simulate discussions between different organizational roles

🛠️ Installation

Prerequisites

Node.js 16.0.0 or higher
npm or yarn package manager
OpenAI API key

Quick Start

Clone the repository

git clone https://github.com/vtmade/MicroCrowd.git
cd MicroCrowd

Install dependencies
```
npm install
```
Set up environment variables Create a .env file in the root directory:
```
API_KEY=your_openai_api_key_here
```

In your MicroCrowd folder, create a .env file using env example file
Replace your_openai_api_key_here with your actual OpenAI API key

Start the development server
```
npm run dev
```
Open your browser Navigate to http://localhost:5173

📋 How to Use

Step 1: Prepare Your Data

Create a CSV file with any combination of relevant fields:

name,age,occupation,location,income,interests,education
Sarah Johnson,32,Marketing Manager,Seattle,75000,Sustainability;Technology,Bachelor's
Mike Chen,45,Small Business Owner,Austin,55000,Sports;Family,High School
...

Supported fields (all optional):

Demographics: name, age, gender, location
Professional: occupation, industry, income, experience
Personal: interests, hobbies, education, family_status
Any custom fields relevant to your research

Step 2: Generate Personas

Click "Generate Personas" in the navigation
Upload your CSV file
Wait for AI to transform each row into a rich persona profile
Review generated personas with full personality traits and backgrounds

Step 3: Setup Focus Group

Navigate to "Setup Focus Group"
Configure your session:
- Topic: The main subject for discussion
- Research Goal: What insights you're seeking
- Discussion Guide: Key questions/topics to cover
- Max Turns: How many rounds of discussion
- Participants: Select which personas to include

Step 4: Run the Simulation

Click "Start Focus Group"
Watch as the AI moderator facilitates natural discussion
Use controls to pause, resume, or end the session
Download transcript when complete

🎭 Enhanced Conversation Intelligence

What Makes It Different?

🚫 Before Enhancement:

90%+ artificial consensus on controversial topics
Generic template responses across all personas
Invalid references to non-existent statements
No personality-driven behavior changes

✅ After Enhancement:

Realistic Disagreements: 30-70% disagreement based on topic controversy
Authentic Speech Patterns: Distinct vocabulary and style per demographic
Validated Conversations: Only references actual statements with exact quotes
Dynamic Personalities: Behavior changes based on emotional investment

Conversation Examples

Gun Laws Discussion:

Rural Construction Worker: "I've been hunting since I was 12. Gun control won't stop criminals."
Urban Teacher: "[PASSIONATE] I see the fear in my students' eyes during lockdown drills. We need action."
Suburban Analyst: "The data shows both perspectives have merit, but we need balanced regulation."

Climate Change Discussion:

Environmental Scientist: "The evidence is overwhelming - we're facing an emergency."
Energy Worker: "[FRUSTRATED] Easy for you to say. These policies cost real jobs."
Working Parent: "I want a better world for my kids, but I can't afford an electric car right now."

Technical Improvements

🔍 Reference Validation: No more "someone mentioned" - only verified quotes
🎭 Speech Diversity: Age, education, and regional speech patterns
😤 Emotional States: Frustration, engagement, and confidence affect responses
⚡ Natural Interruptions: Personality-driven interjections and disagreements
📊 Flow Intelligence: AI moderator adapts to conversation dynamics

📁 Project Structure

MicroCrowd/
├── 📱 components/           # React UI components
├── 📄 pages/               # Application pages  
├── 🧠 services/            # Enhanced conversation engine & core logic
│   ├── conversationOrchestrator.ts    # Main conversation coordinator
│   ├── conversationTracker.ts         # Statement validation
│   ├── controversyStanceAssigner.ts   # Disagreement generation
│   ├── responseStyleManager.ts        # Speech pattern diversity
│   └── ... (8 advanced services)
├── 🛠️ scripts/             # Development and deployment scripts
├── 🔧 tools/               # Python utilities and testing tools
├── 📚 docs/                # Documentation and guides
│   ├── ENHANCED_CONVERSATION_ENGINE.md  # Technical details
│   └── ENHANCED_SETUP.md               # Setup guide
├── 📊 sample_personas.csv  # Example data for testing
└── 📖 README.md           # This file

🔧 Configuration

Environment Variables

API_KEY=your_openai_api_key        # Required: OpenAI API key
OPENAI_MODEL=gpt-4o-mini          # Optional: Model to use (default: gpt-4o-mini)

CSV Format Guidelines

Flexible Structure: Any column names and data types accepted
No Required Fields: System adapts to available data
Rich Data = Rich Personas: More fields create more detailed personalities
Clean Data: Remove duplicates and ensure consistent formatting

🤝 Contributing

Welcome contributions from the community! Whether you're fixing bugs, adding features, or improving documentation, your help makes MicroCrowd better for everyone.

Ways to Contribute

🐛 Bug Reports: Found an issue? Open a GitHub issue
💡 Feature Requests: Have an idea? We'd love to hear it
🔧 Code Contributions:

MicroCrowd

Install / Use

README