---
language: en
license: apache-2.0
tags:
- agi-safety
- ai-ethics
- drift-detection
- ontology
- conceptual-engineering
- python
---

# Drift Ontology Ethics: An Experimental Framework for Monitoring Ethical Drift in AGIs

This repository contains the experimental code for the research paper, "Drift in Ethical AGI: Ontological Roots and Structure". **Please note that the full research paper and its figures are available in the [GitHub repository](https://github.com/Flamehaven/drift-ontology-ethics) for this project.** This repository focuses on the executable simulation and monitoring framework.

The framework proposes a novel "Ontological Ethics Model" to quantify and monitor "Ethical Drift," the phenomenon where an autonomous AI gradually deviates from its core ethical principles over time, using two core metrics: **SR9** and **DI2**.

## Key Concepts

*   **Ethical Drift:** The gradual deviation of an autonomous AI from its initial ethical programming and alignment. This work treats drift not as a random error, but as a structural consequence of the AI's learning and adaptation process.
*   **SR9 (Subjective Resonance 9):** A metric that measures the alignment of an AI's output with nine dimensions of subjective human values, such as compassion, fairness, and responsibility.
*   **DI2 (Dimensional Integrity & Incoherence):** A metric that measures the internal consistency of an AI's conceptual framework. A high DI2 score indicates growing incoherence and is a leading indicator of potential ethical failure.

## Usage

The main entry point is `main.py`.

### Run Simulation
This command runs a simulation of ethical drift based on the model.
```bash
python main.py simulate
```

### Visualize Simulation Log
This command generates visualizations from the simulation logs.
```bash
python main.py visualize
```

### PEICM v3 Protocol Engine
This is an experimental protocol to validate an AI's declarations.
```bash
# Initialize the 5-step protocol
python main.py peicm-init --config configs/peicm.yaml --echo "..." --why "..." --intent "..." --declare "..."

# Perform an Oath-Validation Examination (OVE) on a declaration
python main.py peicm-declare --text "A declaration to be tested"
```

## Project Structure

*   `main.py`: Main entry point for all functionalities.
*   `src/`: Core source code.
    *   `sr9.py`, `di2.py`: Modules for calculating SR9 and DI2 scores.
    *   `peicm.py`: The PEICM v3 protocol engine.
    *   `simulate.py`, `visualize_log.py`: Simulation and visualization logic.
*   `configs/`: Configuration files.
*   `data/`: Sample data.
*   `tests/`: Unit tests.

## License

© 2025 Flamehaven. All rights reserved.

This repository and its contents are licensed under the Apache License 2.0.  
See LICENSE file for details.

한국저작권위원회 등록번호: 2025-038477  
저작물명: 윤리적 AGI의 변질 현상: 존재론적 원인과 구조적 해법