A Conceptual Framework for Human Interaction with Intelligent Systems

Development of artificial intelligence technologies calls for novel modalities in human interaction with systems. This paper presents a comprehensive conceptual framework to meet this trend. Five human-autonomy teaming forms were proposed firstly, including fully human control, mostly human control & partially autonomy control, human-autonomy negotiation, mostly autonomy control & partially human control, and fully autonomy control. Correspondingly, five levels of human-autonomy interaction were defined, ranging from Ll to L5. Then, four conceptual architectures of human-autonomy teaming were constructed, denoting traded control, executive control, combined control, and supervisory control. Finally, human-autonomy teaming was instantiated based on the systematic approaches in systems engineering. Use case analysis was carried out to sketch the components and their interactions in a typical form of human-autonomy teaming. Block definition and external interaction analysis were implemented with Systems Modeling Language to clarify the internal and external input-output relations in a formalized way. The framework put forward in this work may contribute to designing hardware architectures of human interaction with intelligent systems.