A Study on Implementing Justice as a Virtue in Artificial Intelligence : Connecting Josef Pieper’s Structure of Justice to the Artificial Virtue Agent (AVA) Architecture

Key Points

• The central aim is to explore the implementation of justice as a virtue within the architecture of artificial agents.
• Analyzed the limitations of existing AI frameworks regarding the virtue of justice.
• Proposed a reorganization of justice within the agent architecture to enhance ethical functionality.
• Introduced Josef Pieper's three types of justice as a guiding principle for agent design.
• Developed agents that utilize game theory to simulate just actions.
• Demonstrated that separating justice from other virtues enhances its application in AI.
• Featured agents produced outcomes aligned with virtuous behavior in varied game-theoretical contexts.
• Provided evidence supporting the interpretation of prosocial bias in relation to virtue ethics.

Abstract