Work

Emmanuelle: A Symbolic-Cognitive Agent Grounded in Universal Virtues via the S-Module Architecture

Year: 2025

Type: article

Abstract: We present the Module S, a symbolic artificial intelligence architecture integrating seven universal virtues as a guiding framework for agent behavior and learning. Grounded in a dynamically updated semantic graph, the model aligns low-level reinforcement learning signals with high-level symbolic states, selected and regulated by virtue-based weighting. Emmanuelle, the resulting system, demonstrates sustained alignment, interpretability, and ethical consistency in environments such as CartPole-v1 and LunarLander-v2. This work introduces a new paradigm of moral symbolic alignment, combining the power of neural agents with a cosmically inspired symbolic backbone.

Source: Zenodo (CERN European Organization for Nuclear Research)

DOI: 10.5281/zenodo.15863279

Author: Martin Fernandez, Ismaël

Language: English

Topic: Reinforcement Learning in Robotics

Subfield: Artificial Intelligence

Field: Computer Science

Domain: Physical Sciences

Open Access status: green

OpenAlex ID: W6968860607

FWCI:

Cites:

Cited by:

Related to: