Soppia Framework

A Structured Prompting Framework for the Proportional Assessment of Non-Pecuniary Damages in Personal Injury Cases

Overview

Soppia (System of Proportionally and Ponderously Intelligently Applied Guidance) is a structured prompting framework designed to enhance consistency, transparency, and auditability in the quantification of non-pecuniary damages. By integrating weighted legal criteria with Large Language Models (LLMs), Soppia reduces judicial noise and cognitive biases while maintaining human oversight and explainability.

Key Features

• Transparent Methodology: Every decision is traceable and auditable through explicit criteria
• Dual-Logic Scoring System: Correctly models both aggravating and mitigating circumstances
• Weighted Criteria: Reflects the relative importance of different legal factors
• Adaptable Framework: Can be customized for different jurisdictions and areas of law
• AI-Augmented: Leverages LLMs while maintaining interpretability and human control

Case Study

The framework is demonstrated using the 12 criteria for non-pecuniary damages established in Brazilian Labor Law (Art. 223-G, CLT), but the methodology is applicable to any legal system.

Repository Contents

soppia-framework/
├── README.md                    # This file
├── paper/
│   ├── soppia_paper.pdf        # Full academic paper
│   └── main.tex                # LaTeX source for arXiv submission
├── prompts/
│   ├── soppia_full_prompt.md   # Complete prompt template
│   └── soppia_simple_prompt.md # Simplified version
├── examples/
│   ├── example_case_1.md       # Sample case analysis
│   └── example_case_2.md       # Sample case analysis
└── LICENSE                      # MIT License

Quick Start

Using the Framework

1. Choose your LLM: Soppia works with GPT-4, Claude, Gemini, or other advanced LLMs
2. Load the prompt: Use the full prompt from prompts/soppia_full_prompt.md
3. Provide case facts: Input medical reports, testimonies, and evidence
4. Review the analysis: The AI will generate a detailed, step-by-step assessment

Example Usage

User: [Provides case facts about a workplace injury]

Soppia: 
1. CASE SUMMARY
[Analysis of the provided facts]

2. CRITERIA ANALYSIS
Criterion I - Nature of protected legal interest: [Score: 4/5]
[Detailed justification]
...

3. FINAL CALCULATION
- Total weighted score: 58.5 points
- Classification: SEVERE
- Suggested compensation: 15-20× monthly salary

4. CONCLUSION
[Summary and recommendation]

Methodology

The framework consists of four main components:

1. Defined Criteria: Legally relevant factors derived from statutes, case law, or doctrine
2. Calibrated Scoring: 1-5 point scale with dual logic (direct/inverse)
3. Weighting System: Reflects relative importance of each criterion (0.5× to 2.5×)
4. Classification Structure: Maps scores to severity levels and compensation ranges

Applications

• Judges: Structure and justify judicial decisions
• Lawyers: Assess case strength and manage client expectations
• Companies: Proactive risk management and liability assessment
• Academics: Legal analysis and research

Adaptability

The framework can be adapted for:

• Different jurisdictions: Replace criteria with local legal factors
• Other areas of law: Consumer protection, environmental law, tort law
• Custom weighting: Calibrate weights based on local precedents

Academic Paper

The full academic paper is available in the paper/ directory and has been submitted to arXiv. The paper provides:

• Theoretical foundations (noise reduction, XAI, prompt engineering)
• Detailed methodology
• Complete weighting justifications
• Discussion of applications and limitations

Citation

If you use this framework in your research or practice, please cite:

@article{araujo2025soppia,
  title={Soppia: A Structured Prompting Framework for the Proportional Assessment of Non-Pecuniary Damages in Personal Injury Cases},
  author={Araujo, Jorge Alberto},
  journal={arXiv preprint arXiv:XXXX.XXXXX},
  year={2025}
}

Author

Jorge Alberto Araujo
Magistrate and Legal AI Researcher

License

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

Contributing

Contributions are welcome! Please feel free to submit issues, suggestions, or pull requests to improve the framework.

Acknowledgments

This work builds upon the foundational research of:

• Kahneman, Sibony, and Sunstein on noise in human judgment
• Cynthia Rudin on interpretable machine learning
• The broader XAI and legal AI research communities

Disclaimer

This framework is designed as a decision support tool and does not replace professional legal judgment. All outputs should be reviewed and validated by qualified legal professionals.

---

Version: 1.0.0
Last Updated: October 2025
Status: Active Development