Hello everyone,
I’m Marussa, a solo founder working on QERRA-v2.
I’ve created SEMEV-12, a 12-dimensional ethical evaluation framework that is designed to work as a real-time safety layer for AI systems and, in the future, autonomous robots. The 12 core dimensions (Coherence Protection, Family Severance, Survival Instinct, Moral Pressure, Harm Intent, and others) come from deep personal experience and careful observation of real human consequences — not from abstract theory or cultural checklists.
What makes these vectors different is that they focus on universal human invariants that remain relevant across cultures, time, and legislation. They are built to be computable and explainable, so they can actually be used in real decision-making pipelines rather than staying as high-level principles.
The classical version is already a working FastAPI service that analyses text and returns a structured ethical score with clear reasoning. It is currently live on Hugging Face.
I’ve just published the full documentation that explains the origins of each vector, why they are designed this way, and how the classical implementation works.
Full documentation → QERRA-v2-classical/SEMEV-12_Framework_Documentation.md at main · marunigno-ship-it/QERRA-v2-classical · GitHub
I would really appreciate feedback from the ROS community, especially from people working on Behavior Trees, deliberation systems, or robot safety. I’m particularly interested in how such an ethical layer could be integrated into real robotic architectures.
Thank you in advance for any thoughts or suggestions.Best regards,
Marussa Metocharaki
Founder, QERRA-v2