Towards a Toolchain for Formally Capturing and Monitoring Agent Accountability
摘要
The increasing deployment of AI agents in enterprise systems presents a critical governance challenge: ensuring clear accountability chains between humans and autonomous systems while supporting their dynamic, adaptive behavior. This paper presents a toolchain that uses the ISO-standard ODP Enterprise Language to formally specify and monitor actor responsibility. Building on our previous DSL work for ODP Enterprise Language formalism expression, we develop a software toolchain that enables domain users to define actor responsibilities and accountability against normative concepts expressed using the basic normative concepts of permissions, obligations, prohibitions, and authorizations. The toolchain supports expressing delegation scenarios between humans and agents, and across multiple agents, accommodating the dynamic nature of AI-driven enterprise systems. Our solution leverages the textX DSL development framework to create a fit-for-purpose toolchain that integrates with contemporary enterprise and AI technologies, providing formal foundations for accountable agent deployment in enterprise distributed systems.