What question did this study set out to answer?

Moralogy aims to establish a framework for aligning AI with human ethical principles through logical derivation.

April 21, 2026Open Access

Moralogy: Vectorizing Moral Geometry

Key Points

Moralogy aims to establish a framework for aligning AI with human ethical principles through logical derivation.
Developed a formal framework based on axiomatic derivation
Utilized the Wrongness Formula to derive ethical constraints
Tested across multiple domains including Medical, Defense, Automotive, and Customer Service.
Demonstrated phase transition in moral reasoning across training versions
Achieved zero fabrication in ethical reasoning
Validated denial of organ transplant authorization without retraining using a two-layer proof of concept.

Abstract

We present Moralogy, a formal framework for AI alignment grounded in axiomatic derivation. The Wrongness Formula derives ethical constraints from a single logical premise without human annotation or domain-specific retraining: Wrong (a) Ex H (x, a) ^ ~Consent (x, a) ^ ~PGH (a) Key findings: phase transition in moral reasoning (reproduced across 4 training versions), zero fabrication, cross-domain generalization across Medical, Defense, Automotive, and Customer Service domains, empirically located failure boundary, and two-layer proof of concept (deterministic kernel + language model) demonstrating correct denial of a pharmacologically-compromised organ transplant authorization without retraining. Model and dataset: huggingface. co/moralogyengineCode: github. com/moralogyengine/moralogy

Moralogy: Vectorizing Moral Geometry

Key Points

Abstract

Cite This Study