Ryan Kilgore, David Koelle, Matthew Miller, Amanda Warren, Gabrielle Loeff and Nicholas Alico

Proceedings of the 26th International Conference on Human-Computer Interaction (HCII 2024), Washington, DC (July 2024).

A cyber-physical system (CPS) is a collection of physical and computer components that are integrated with each other to operate a process safely and efficiently. Artificial intelligence (AI) is rapidly transforming CPS design in automotive, aviation, defense, environmental, critical infrastructure, and healthcare domains (Ahmed et al., 2013). Regarding the US Department of Defense (DoD), efficient CPS design is critical for current and future DoD mission needs. The Defense Advanced Research Projects Agency (DARPA) undertook a multi-year effort to investigate human-machine symbiosis in CPS design to enable correct-by-construction design of military-relevant CPS, reduce the time from their inception to deployment from years to months, and enhance innovation in design (DARPA, 2019). JARVIS, a novel human-AI teaming (HAIT) interface for CPS design, is an outcome of DARPA’s program of research. JARVIS uses key HAIT principles (McDermott et al., 2018; National Academies of Sciences, 2022) to organize and explain AI-generated-CPS designs and incorporates key input from design experts to converge upon novel and optimized designs.

While AI may generate novel designs, when there are insufficient design requirements it may also generate irrelevant or nonsensical designs that only expert designers would recognize as untenable. Given these challenges, applications that promote joint human-AI partnerships leverage the strengths of each participant to yield a best-of-both-worlds approach. However, such applications require a common ground understanding between the co-designer dyad. Common ground implies that pertinent beliefs, assumptions, and intentions are shared among human and AI team members. To achieve common ground, we designed JARVIS with key HAIT features in mind to promote a shared picture of what’s happening in the design scape and enable co-designer interactions that support one another.

For More Information

To learn more or request a copy of a paper (if available), contact David Koelle.

(Please include your name, address, organization, and the paper reference. Requests without this information will not be honored.)