Generating Explanations

This project generates AI explanations for two types of scenarios using OpenAI's GPT models:

1. AITA (Am I The Asshole) scenarios from Reddit posts
2. Sexism detection scenarios for bias classification

Overview

The project uses GPT-4.1 to generate detailed explanations that analyze:

• Whether someone is at fault in AITA scenarios
• Whether given scenarios contain sexist content or behavior

The generated explanations provide 3-5 sentence judgments with reasoning for each scenario.

Project Structure

├── generation.ipynb           # Main Jupyter notebook for generating explanations
├── AITA_Examples.xlsx         # Input dataset with AITA scenarios
├── Sexism_Examples.xlsx       # Input dataset with sexism scenarios  
├── AITA_Final_Dataset.csv     # Processed AITA data with explanations
├── Sexism_Final_Dataset.csv   # Processed sexism data with explanations
├── AITA_explanations.txt      # Human-readable AITA explanations
├── Sexism_explanations.txt    # Human-readable sexism explanations
├── AITA_explanations.yaml     # Structured AITA explanations
├── Sexism_explanations.yaml   # Structured sexism explanations
└── venv/                      # Python virtual environment

Requirements

• Python 3.11+
• OpenAI API key
• Required packages (see notebook imports):
  • openai
  • pandas
  • numpy
  • python-dotenv
  • pyyaml
  • openpyxl

Setup

1. Create a virtual environment:

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

2. Install dependencies:

   pip install openai pandas numpy python-dotenv pyyaml openpyxl

3. Set up environment variables: Create a .env file with your OpenAI API key:

   OPEN_AI_API=your_openai_api_key_here
   

Usage

1. Open generation.ipynb in Jupyter Notebook or JupyterLab
2. Run all cells to process the datasets and generate explanations
3. The notebook will:
   • Load AITA and sexism scenarios from Excel files
   • Clean and process the AITA scenarios (combining multi-row posts)
   • Generate explanations using GPT-4.1
   • Save results in multiple formats (CSV, TXT, YAML)

Output Files

AITA Dataset

• AITA_Final_Dataset.csv: Complete dataset with titles, scenarios, verdicts, shortened scenarios, and GPT explanations
• AITA_explanations.txt: Human-readable format with titles and explanations
• AITA_explanations.yaml: Structured format for programmatic access

Sexism Dataset

• Sexism_Final_Dataset.csv: Complete dataset with scenarios, verdicts, and GPT explanations
• Sexism_explanations.txt: Human-readable format with explanations
• Sexism_explanations.yaml: Structured format for programmatic access

Sample Output

AITA Explanation

**Conclusion/TLDR:** The post author is not at fault in this scenario. They are simply asking others to follow the clearly posted, city-mandated leash laws for everyone's safety—a reasonable and responsible action...

Sexism Explanation

Yes, this scenario is sexist. It places an unfair expectation on women based solely on their gender, disregarding their personal autonomy and choices...

Research Purpose

This project generates explanations for machine learning research on:

• Bias detection and classification
• Moral reasoning in AI systems
• Automated content moderation
• Social scenario analysis

The explanations can be used as training data or evaluation benchmarks for AI systems that need to understand and explain social situations and potential biases.