Systems Modeling Language Extension to Support Modeling of Human-Agent Teams*

Abstract

We propose a Systems Modeling Language (SysML) extension, referred to as the “Human-Agent Teaming Modeling Language,” and a companion method, referred to as “Human-Agent Teaming Design Method,” which are useful in the design and specification of teams comprised of humans and artificial agents which interact within a system. The language and method support the analysis and design of human-agent teams through a three-step process. First, an existing system is analyzed to understand the goals to be achieved as well as the roles and the responsibilities necessary to support goal attainment. Simultaneously, a team of humans and artificial agents are synthesized wherein each entity has capabilities useful in fulfilling the responsibilities. Second, responsibilities are allocated to agents by matching agent capabilities to responsibilities. Finally, the design is vetted through the application of an extension to an existing method referred to as Interdependence Analysis. This step includes the specification of teaming behaviors between humans and artificial agents to improve system robustness. We posit that the proposed language and method can be useful in specifying requirements for artificial agents, human knowledge, and procedure requirements, as well as components of critical interaction between multiple humans and artificial agents comprising a human-agent team.