Requirements-based Explainability for Multi Agent Systems

Explainability is essential for building trust in intelligent and autonomous systems. However, existing techniques for explainability focus on the behaviour of the system, but do not go back to the system’s requirements. We provide fully traceable explanations that link back to requirements, expressed as User and System stories, by extending previous work on explainable agents (XAg) that uses an agent design pattern. Our implementation leverages industry-grade mainstream monitoring tools.

Authors

• Sebastian Rodriguez
• John Thangarajah
• Michael Winikoff

Keywords

• Explainable Agents (XAg)
• requirements
• Verification
• Monitoring