Modelling situated intent for human-autonomy teaming: a human-centric approach

Abstract

Entering an era where humans and synthetic agents are supposed to collaborate and cooperate, adequate models of human intent are crucial for coordinated teamwork. Unfortunately, although there is a need for such models, the concept of intent is ambiguous and approaches to model intent from a human-centric perspective are scarce. Building upon theoretical and methodological foundations, this study aims to address these gaps by presenting a conceptualisation of intent alongside an approach. Specifically, leveraging the six levels of cognitive control outlined in the Joint Control Framework, a provisional model of human intent alongside a defined and operationalised concept is presented. Building on these foundations, a novel approach is proposed. Utilising seven scenario-based interviews, the value of these contributions is demonstrated through an example case in the context of Manned-Unmanned Teaming. It is concluded that intent should be understood as a multi-faceted concept shaped by situated constraints, where intent is formed through a commitment to choices by context-situation and means-end reasoning. It is also concluded that the approach is useful, particularly since it can glean insights from choices considered and committed, both being essential in the design of synthetic teammates’ capability to adapt to their human partner’s agency.