SEAL (Social Emotional Adaptive Learning)

An AI-powered system that analyzes student performance data and generates targeted intervention plans using advanced language models with detailed EMT-specific scenarios.

Overview

SEAL is a REST API service that helps educators create personalized intervention strategies for students based on their emotional intelligence metrics. It uses Google's Gemini AI to generate contextually relevant and age-appropriate interventions through two specialized assessment tools:

1. EMT Assessment Tool (Q1) - Analyzes class performance data across four EMT areas with detailed scenario-based interventions
2. Curriculum Assessment Tool (Q2) - Provides grade-appropriate emotional learning activities based on skill areas and performance scores

Features

EMT Assessment Tool - Detailed Scenario-Based Interventions

Process and analyze Emotional Measurement Task (EMT) scores across four key areas with specific intervention strategies:

• EMT1: Visual Emotion Matching (Visual-to-visual matching)

  • Emotion flashcard pairs for matching practice
  • Mirror expression practice with emotion cards
  • Digital emotion matching games with progressive difficulty
  • Pattern recognition activities with facial expressions

• EMT2: Situation-to-Expression Connection (Verbal context to visual expression)

  • Story-based emotion discussions with character analysis
  • Scenario cards with emotional contexts for matching
  • Role-playing emotional situations with expression practice
  • Situational emotion analysis activities

• EMT3: Expression Labeling (Visual to verbal labeling)

  • Emotion word wall development and daily practice
  • Expression-label matching games and activities
  • Emotion vocabulary journals with daily entries
  • Vocabulary building through visual-verbal connections

• EMT4: Label-to-Expression Matching (Verbal label to visual expression)

  • Emotion word-to-face games and quick responses
  • Verbal emotion cues practice with audio support
  • Group emotion word activities and competitions
  • Label comprehension through interactive exercises

Curriculum Assessment Tool

Grade-specific emotional learning activities for:

• Emotional Awareness: Understanding and identifying emotions
• Emotional Regulation: Managing and controlling emotional responses
• Anger Management: Specific strategies for anger control and expression
• Supports grades 1, 2, and 5

Intervention Generation

Creates detailed intervention plans including:

• Performance analysis with EMT-specific focus areas
• Targeted strategies based on proven EMT methodologies
• 4-week progressive implementation timelines
• Measurable success metrics
• Specific resource requirements
• Age-appropriate and classroom-ready activities

🛡️ Safety System

Comprehensive content validation to ensure all AI-generated content is appropriate for children:

• Multi-layer safety validation with severity classification
• Child-focused protection against harmful or inappropriate content
• Configurable safety levels for different environments
• Real-time content filtering and violation logging
• See SAFETY.md for detailed documentation

API Features

• JSON schema validation
• Automatic retries with exponential backoff
• Health check endpoint
• Structured logging
• CORS support
• Data privacy compliance (aggregated scores only sent to AI)

Quick Start

1. Clone the repository

   git clone https://github.com/yourusername/seal.git
   cd seal

2. Set up a virtual environment

   python -m venv venv
   source venv/bin/activate  # On Windows use: venv\Scripts\activate

3. Install dependencies

   pip install -r requirements.txt

4. Environment Setup

   Copy the example environment file and configure your settings:

   cp example.env .env
   # Edit .env with your configuration

   Required environment variables:

   GOOGLE_API_KEY=your_gemini_api_key_here
   TOKENIZERS_PARALLELISM=false
   LLM_PROVIDER=gemini  # Optional, defaults to gemini if not set
   

5. Start the API server

   python -m uvicorn app.main:app --reload

   The API will be available at http://localhost:8000

6. Access the API documentation

   • OpenAPI docs: http://localhost:8000/docs
   • ReDoc: http://localhost:8000/redoc

API Endpoints

1. EMT Assessment Tool - Generate Intervention Plan

POST /score

Analyzes class EMT performance data and generates targeted intervention strategies using detailed EMT-specific scenarios.

Example request:

{
  "scores": {
    "EMT1": [65.0, 70.0, 68.0],
    "EMT2": [58.0, 62.0, 60.0],
    "EMT3": [72.0, 75.0, 70.0],
    "EMT4": [63.0, 65.0, 64.0]
  },
  "metadata": {
    "class_id": "C1",
    "deficient_area": "EMT2",
    "num_students": 3
  }
}

The API will automatically select appropriate intervention strategies from the EMT2 scenario set (Situation-to-Expression Connection) including story-based discussions, scenario cards, and role-playing activities.

2. Curriculum Assessment Tool - Generate Learning Plan

POST /curriculum

Provides grade-appropriate emotional learning activities based on skill areas and performance.

Example request:

{
  "grade_level": "2",
  "skill_areas": ["emotional_awareness", "emotional_regulation"],
  "score": 65.5
}

3. Health Check

GET /health

Returns the health status of both assessment tools and services.

Project Structure

seal/
├── app/
│   ├── llm/
│   │   ├── gateway.py              # EMT assessment tool with detailed scenarios
│   │   ├── curriculum_gateway.py   # Curriculum assessment tool implementation
│   │   └── __init__.py
│   ├── prompts/
│   │   ├── intervention.py         # EMT scenarios and intervention strategies
│   │   ├── curriculum.py           # Curriculum intervention templates
│   │   └── __init__.py
│   ├── schemas/
│   │   ├── base.py                 # EMT score and intervention schemas
│   │   ├── curriculum.py           # Curriculum-specific schemas
│   │   └── __init__.py
│   ├── safety/
│   │   ├── guardrails.py           # Content safety validation
│   │   └── __init__.py
│   ├── main.py                     # FastAPI application with both assessment tools
│   └── __init__.py
├── data/
│   └── synthetic_data.py           # EMT test data generation
├── test/
│   ├── compare_llms.py             # Model comparison testing
│   └── results/                    # Test results storage
├── requirements.txt
├── Dockerfile
├── docker-compose.yml
├── example.env                     # Environment configuration template
├── SAFETY.md                       # Safety system documentation
├── DEPLOYMENT.md                   # Deployment instructions
└── README.md

Assessment Tool Details

EMT Assessment Tool (Q1 Implementation)

The EMT Assessment Tool processes emotional measurement task scores and generates intervention plans using detailed, scenario-based strategies. Each EMT area has specific proven interventions:

• Comprehensive scenario library with 3+ detailed strategies per EMT area
• Progressive difficulty levels from basic to complex emotions
• Multi-modal approaches including visual, verbal, and interactive components
• Classroom-ready activities with specific resource requirements
• Measurable outcomes with quantitative and qualitative metrics

Curriculum Assessment Tool (Q2 Implementation)

The Curriculum Assessment Tool provides grade-specific interventions across three skill areas:

• 10 predefined interventions covering emotional awareness, regulation, and anger management
• Grade-appropriate activities for grades 1, 2, and 5
• Structured implementation plans with clear steps and timelines
• Developmental considerations matching cognitive and emotional capabilities

Data Privacy

SEAL is designed with data privacy in mind:

• Individual student scores are accepted by the API but never sent to the AI model
• Only aggregated class averages are included in prompts sent to the LLM
• No personally identifiable information is processed or stored
• All AI interactions use anonymized, class-level data only

Docker Support

1. Build the image

   docker build -t seal .

2. Run the container

   docker-compose up

Testing

Run the model comparison tests to evaluate different LLM performance:

python test/compare_llms.py

Results are saved in test/results/ with detailed performance metrics and response quality analysis.

Future Improvements

1. Add support for additional grade levels
2. Expand EMT scenario library with more detailed interventions
3. Implement real-time assessment data input capabilities
4. Add a user interface for easier interaction with both assessment tools
5. Develop progress tracking and outcome measurement features
6. Add multi-language support for diverse classrooms

License

This project is licensed under the MIT License - see the LICENSE file for details.

Contributing

1. Fork the repository
2. Create a feature branch
3. Make your changes
4. Add tests for new functionality
5. Submit a pull request

For questions or support, please open an issue on GitHub.