Drift Ontology Ethics

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

The drift-ontology-ethics repository presents an innovative approach to understanding and monitoring Ethical Drift in autonomous artificial general intelligences (AGIs). This experimental framework is developed in conjunction with the research paper titled "Drift in Ethical AGI: Ontological Roots and Structure." Ethical Drift refers to the gradual deviation of an AGI from its fundamental ethical principles as it undergoes learning and adaptation processes.

Key Concepts

• Ethical Drift: This phenomenon is viewed not merely as an error but as a structural outcome of an AGI’s learning journey, highlighting the importance of alignment with core ethical guidelines.
• SR9 (Subjective Resonance 9): A metric that assesses the alignment of AGI outputs against nine subjective human values, including compassion, fairness, and responsibility.
• DI2 (Dimensional Integrity & Incoherence): This metric evaluates the internal coherence of an AGI's conceptual framework. A high DI2 score signifies growing incoherence, serving as a crucial indicator of potential ethical breaches.

Simulation and Visualization

This framework allows users to simulate ethical drift and visualize results to better understand the dynamics between SR9 and DI2 metrics during drift events. Example simulations include:

# Run a simulation of ethical drift
python main.py simulate

# Generate visualizations from simulation logs
python main.py visualize

PEICM v3 Protocol Engine

The repository also includes an experimental protocol, the PEICM v3, designed to validate an AGI's declarations. Users can initialize the protocol and conduct an Oath-Validation Examination (OVE) with commands such as:

# Initialize the protocol
python main.py peicm-init --config configs/peicm.yaml --echo "..." --why "..." --intent "..." --declare "..."

# Validate a declaration
python main.py peicm-declare --text "A declaration to be tested"

Project Structure

The repository is organized to facilitate understanding and development:

• main.py: The primary entry point for all functionalities.
• src/: Contains core source code, including modules for SR9, DI2, and the PEICM protocol.
• configs/: Configuration files for simulations.
• data/: Sample datasets for testing.
• paper/: The full research paper and associated figures.
• tests/: Unit tests to ensure code reliability.

This repository stands as a significant contribution to the field of AGI safety and ethics, providing researchers and developers with the tools necessary to monitor and address ethical drift in autonomous systems.